Active Galactic Nuclei in Void Regions
aa r X i v : . [ a s t r o - ph ] O c t ApJ in press; submitted on May 11, 2007
Preprint typeset using L A TEX style emulateapj v. 08/22/09
ACTIVE GALACTIC NUCLEI IN VOID REGIONS
Anca Constantin
Department of Physics, Drexel University, Philadelphia, PA 19104
Fiona Hoyle
Department of Physics and Astronomy, Widener University, Chester, PA 19013
Michael S. Vogeley
Department of Physics, Drexel University, Philadelphia, PA 19104
ApJ in press; submitted on May 11, 2007
ABSTRACTWe present a comprehensive study of accretion activity in the most underdense environments in theuniverse, the voids, based on the SDSS DR2 data. Based on investigations of multiple void regions,we show that Active Galactic Nuclei (AGN) are definitely common in voids, but that their occurrencerate and properties differ from those in walls. AGN are more common in voids than in walls, butonly among moderately luminous and massive galaxies ( M r < −
20, log M ∗ /M ⊙ < . L [OIII] < ergs − ). Void AGN hosted by moderately massive and luminous galaxies are accreting at equal or lowerrates than their wall counterparts, show lower levels of obscuration than in walls, and similarly agedstellar populations. The very few void AGN in massive bright hosts accrete more strongly, are moreobscured, and are associated with younger stellar emission than wall AGN. These trends suggest thatthe accretion strength is connected to the availability of fuel supply, and that accretion and star-formation co-evolve and rely on the same source of fuel. Nearest neighbor statistics indicate that theweak accretion activity (LINER-like) usually detected in massive systems is not influenced by the localenvironment. However, H ii s, Seyferts, and Transition objects are preferentially found among moregrouped small scale structures, indicating that their activity is influenced by the rate at which galaxiesinteract with each other. These trends support a potential H ii → Seyfert/Transition Object → LINERevolutionary sequence that we show is apparent in many properties of actively line-emitting galaxies,in both voids and walls. The subtle differences between void and wall AGN might be explained by alonger, less disturbed duty cycle of these systems in voids.
Subject headings: cosmology: large-scale structure of the universe – galaxies: active—galaxies:emissionlines— methods: statistical INTRODUCTION
The regions that are apparently devoid of galax-ies (Kirshner et al. 1981) and clusters (Einasto et al.1980), the voids, are arguably the best probes of theeffect of the environment and cosmology on galaxy for-mation and evolution. If, as suggested by the standardcosmological paradigm, structure in the present-day uni-verse formed through hierarchical clustering, with smallstructures merging to form progressively larger ones,galaxies in the currently most underdense regions mustbe the least “evolved” ones, as they must have formedat later times than those in the dense regions. There-fore, void and cluster galaxies must follow different evo-lutionary paths. Disturbing processes like stripping andharassment, that operate preferentially in crowded envi-ronments, should occur rarely in voids. Studies of theproperties of the void galaxies, in contrast to those inrelatively crowded regions, or walls, should provide someof the strongest constraints for distinguishing the intrin-sic properties, which characterize a galaxy when it is firstassembled, from properties that have been externally in-duced, over the whole history the universe: the “natureversus nurture” problem. Statistically significant conclusions regarding the dis-tinctness of the void galaxies relative to those in denserregions, the “walls” hereafter, emerged only recently,with the advent of large surveys such as SDSS and 2dF.Such data, and in particular SDSS, offered for the firsttime the possibility to find and analyse both photomet-rically and spectroscopically, large samples of extremelylow density regions (i.e., δρ/ρ < − . h − Mpc, Rojas et al. 2004, 2005), and allowedfor accurate estimates of the void galaxy luminosityand mass functions (Hoyle et al. 2005; Goldberg et al.2005). These studies show that void galaxies are fainter,bluer, have surface brightness profiles more similar tothose of late-type systems, and that their specific star for-mation rates are higher than those in denser regions. Themass and luminosity functions are found to be clearlyshifted towards lower characteristic mass and faintermagnitudes ( M ∗ ). Moreover, the faint end slopes of thewall and void luminosity functions are very similar whichsuggests that voids are not dominated by an excess pop-ulation of low-luminosity galaxies. Consistently, no sig-nificant excess in the amount of dark matter is apparent.This means that, although largely devoid of light, the Constantin et al.most underdense regions conform to a galaxy formationpicture which is clearly not strongly biased.All these peculiarities demonstrate that the cos-mological evolution of void systems is different fromthat of those living in environments of average cos-mic densities. Given the tight correlations betweenthe mass of black holes ( M BH ) and the dispersionsand the masses of the galactic bulges within whichthey reside (Magorrian et al. 1998; Ferrarese & Merrit2000; Gebhardt et al. 2000; Marconi & Hunt 2003),one would then expect that the growth of massive BHsin galaxy centers (and therefore the accretion processwithin active galactic nuclei, AGN), also differs amongdistinct environs. Extension of environmental studiesof AGN properties to extreme regions like cosmic voidsis thus crucial to understand the co-evolution of galax-ies and their central BHs. Moreover, while there isgeneral agreement that the growth of black holes mustbe closely related to galaxy assembly (e.g., Silk & Rees1998; Kauffmann & Haehnelt 2000; Begelman & Nath2005), there is no consensus as to how exactly accre-tion and star formation are coupled. The void galaxiescould be, arguably, the best test-bed for understandingwhether these processes are synchronized, or precede oneanother, and whether feedback from the actively growingBHs facilitates star formation (e.g., by dynamically com-pressing gas clouds through radio jets), or suppresses it(e.g., by blowing away the gas).To date, studies of the spectral properties of thevoid AGN remain limited to individual voids, e.g., theBootes void (Kirshner et al. 1981), permitting the iden-tification of only a few AGN among only a few dozenvoid galaxies (Cruzen et al. 2002). Quite surprisingly,such investigations find that the AGN fraction and theiremission-line properties are similar in voids and in theirfield counterparts. Moreover, their associated stellarpopulations appear to share similar characteristics in thetwo extreme environs. The conclusions of these studiesare based on small number statistics and do not howeverexclude the hypothesis that the void emission-line activ-ity, whether originating in star-formation or accretion,could be connected with, e.g., filaments within voids;such structures would provide local environs similar tothose in the field. The present SDSS samples of voidsand void galaxies offer us for the first time the possibil-ity to test and observationally constrain such ideas.It is important to note that previous investigations ofthe environmental dependence of nuclear activity in therelatively nearby universe do not reach the extreme spa-tial densities representative of voids. For example, inKauffmann et al. (2004), the lowest density regions in-clude over 25% of the galaxies, which is more than 3times more galaxies than the true void regions encom-pass. Their conclusions are interesting, and an extensionof such an investigation at truly low densities is clearlydesirable. In particular, it is important to quantify thedegree to which the finding that, at fixed stellar mass,twice as many galaxies host strong-lined AGN in low-density regions than in high, extends to cosmic voids.Our work provides such an analysis.We employ in this work the most accurately clas-sified samples of voids identified within SDSS todate, that yield ∼ void galaxies. Motivated byour recent study on the AGN clustering phenomena (Constantin & Vogeley 2006), which shows that thereare differences in the large scale structure of active galax-ies, and that their clustering amplitude correlates withtheir strength or rate of accretion, and possibly withthe availability of fuel, we compare void and wall ac-tive galaxies of different types as classified based on theiremission-line properties. Through such a comparison weaim to understand: 1) how the large and small-scalestructures influence accretion onto their central blackholes, and 2) to what degree AGN activity is triggeredby interactions or mergers between galaxies. To answerthese questions, we investigate the occurrence rate of dif-ferent types of spectrally defined AGN, and how theiraccretion activity relates to their associated black holemass, the mass and the age of their associated stellarpopulations, host morphology, brightness, and nearest-neighbor distance.We organize the paper as follows. In Section 2 wepresent the void and wall sample selection, and the spec-tral classification we use in defining various types of ac-tively emitting galaxies. We compare and discuss theAGN occurrence rate in voids and walls, both globallyand at fixed host properties in Section 3. We examinein Section 4 the accretion rates, the fuel supply, and theproperties of the associated star-formation in void andwall actively line emitting systems, while in Section 5 wediscuss potential differences in their small scale environ-ments. We summarize our findings in 6, and discuss thepossible implications on the nature of the power sourcesin the low luminosity AGN, the AGN–host connection,and current models of galaxy formation. In particular,we show empirical evidence for a possible evolutionary se-quence that links different types of strong line-emittinggalaxies defined based on their spectral characteristics.Throughout this work, unless otherwise noted, we as-sume Ω m = 0 .
3, Ω Λ = 0 .
7, and H = 100 h km s − Mpc − . THE DATA
We employ for this study 10 void galaxies identifiedfrom the large-scale structure sample10 , as described inthe NYU Value-Added Galaxy Catalog (Blanton et al.2005), which is a subset of the SDSS Data Release 2 thatStrauss et al. (2002) describe in detail. This sample cov-ers nearly 2000 deg and contains 155,126 galaxies. Tech-nical details about the photometric camera, photometricanalysis and photometric system employed by SDSS canbe found in Gunn et al. (1998), Lupton et al. (2002),and Fukugita et al. (1996) and Smith et al. (2002), re-spectively. Hogg et al. (2001) describe the photomet-ric monitor, Pier et al. (2003) present the astrometriccalibration, and Eisenstein et al. (2001), Strauss et al.(2002) and Blanton et al. (2003) discuss the selection ofthe galaxy spectroscopic tiling. An overview of the SDSScan be found in York et al. (2000).The following subsections summarize the method ofidentification of void and wall galaxies, and their generalproperties. We present them in terms of emission-lineactivity and their classification in various species thatreflect various degree of contribution from nuclear ac-cretion and star formation. Our spectral analysis usesmeasurements of absorption and emission line fluxes andequivalent widths (EW) drawn from a catalog built byGN in Voids 3the MPA/JHU collaboration . In this dataset, the lineemission component is separated and subtracted fromthe total galaxy spectrum based on fits of stellar popu-lation synthesis templates (Tremonti et al. 2004). Torelate the central BH accretion activity to the host prop-erties, we also use in this analysis stellar masses of galax-ies and two stellar absorption-line indices, the 4000˚Abreak strength and the H δ A Balmer absorption-line in-dex, as calculated by Kauffmann et al. (2003c). A de-tailed analysis of many of these properties, and their rela-tion to the AGN phenomenon in particular, is presentedin Kauffmann et al. (2004).
The Void Galaxy Sample
Voids galaxies were identified using a nearest neighboranalysis to find galaxies that reside in regions of den-sity contrast with δρ/ρ < − . h − Mpc. Wall galaxies are then those objects forwhich δρ/ρ ≥ − .
6. The void galaxy selection procedureis described in detail in Rojas et al. (2004). Briefly, thismethod uses a volume-limited sample ( z min = 0 . z max = 0 . M r < − . z max , the galaxies with fewer than three neighborsin the volume limited sample, within a sphere of radius7 h − Mpc, are flagged as belonging to voids while therest of them are classified as wall galaxies. This proce-dure yields a sample of 1,010 void galaxies and 12,732wall galaxies. Objects that lie close to the edge of thesurvey have been discarded as it is impossible to accu-rately count the neighbors if they are not observed.Note that the density around void galaxies ( ρ vg ) ishigher than the mean density of a void (¯ ρ void ) becausegalaxies are clustered and the few void galaxies tend to lieclose to the edges of the voids. Our choice of density con-trast and nomenclature is consistent with studies of voidsin other three-dimensional samples (Hoyle & Vogeley2002, 2004) where individual void structures were iden-tified using an objective voidfinder algorithm thatmeasures void sizes, average densities, and density pro-files. This procedure uncovers voids with typical radiiof 12.5 h − Mpc that fill 40% of the Universe and have amean density δρ/ρ < − .
9. The average density aroundthe few galaxies in voids is typically δρ/ρ < − . h − Mpc (the typical underden-sity for voids is δρ/ρ ≈ − . The Spectral Classification
We identify and classify accretion sources andother types of active systems in the void and wallgalaxy samples based their emission line proper-ties. Following, e.g., Baldwin, Phillips, & Terlevich(1981), Veilleux & Osterbrock (1987), andConstantin & Vogeley (2006), we employ a set offour line flux ratios to distinguish between sys-tems in which ionization is dominated by accretion publicly available at (Brinchmann et al. 2004) and/or starlight: [O III ] λ β , [N II ] λ α ,[S II ] λλ α , and [O I ] λ α . Thus, weselect from the parent sample of galaxies a subsetof strong emission-line sources that show significantemission in all six lines used in the type classification(H α , H β , [O III ],[N II ], [S II ], and [O I ]), and a setof passive objects that show insignificant, if any, lineemission activity. An emission feature is considered tobe significant if its line flux is positive and is measuredwith at least 2 σ confidence.To classify the actively line emitting objects we employthe criteria proposed by Kewley et al. (2006). Throughtheir semi-empirical nature, the Kewley et al. (2006)separation lines match well with objects’ locations inthree diagnostic diagrams which combine pairs of the sixlines mentioned above. Figure 1 illustrates these def-initions for cases where the line fluxes are constrainedto > σ accuracy, separately in voids and walls. Thegalaxies that show strong star-formation activity, the H ii galaxies, are those that lie below the Kauffmann et al.(2003c) curve (the dashed line) in the [O III ]/H β vs.[N II ]/H α diagram, and remain confined to the leftwing of the point distribution in the other two diagramsas well. Objects situated above this curve but belowthe Kewley et al. (2001) theoretical “maximum star-formation” line (continuous black line) are defined asTransition objects (Ts, or Composites). Galaxies whoseline flux ratios place them above the Kewley et al.(2001) curves are those where the AGN component isconsidered to be dominant. The AGN are separated intoSeyferts (Ss) and LINERs (Ls) by a diagonal (continuousblue line) that represents the best fit to the location ofthe minimum of the distributions in pairs of line ratiosdefining the [O III ]/H β vs. [S II ]/H α , and [O III ]/H β vs.[O I ]/H α diagrams (Kewley et al. 2006).Even though we consider these criteria as the best ba-sis for categorizing the line-emitters, we test our resultswith respect to the Ho et al. (1997a) definition, for thesake of comparison with previous works. Figure 1 showsthe Ho et al. (1997a) separation lines as as (green) dot-ted lines. There are some important differences be-tween these two classification methods, some of which arebriefly discussed by Kewley et al. (2006). Although theLINER samples remain at least 90% consistent in thesetwo different definitions, a significant number of Seyfertsand Transition objects change class. The Kewley et al.criteria classify quite a few Seyferts that would be definedby Ho et al. either as H ii s or Ts based on their low ion-ization level, gauged by their [O III ]/H β ratio. The num-ber of objects classified as Ts significantly increases withthe Kewley et al. (2006) method. However, many ofthese systems are clearly dominated by starlight, even iftheir emission embodies a possible accretion component.Constantin & Vogeley (2006) show that a large fractionof Ts defined based only on the [N II ]/H α diagram heav-ily populate the H ii locus in the other two diagrams.The [S II ]/H α and [O I ]/H α line flux ratios are more sen-sitive to the type of ionization than [N II ]/H α , and there-fore better discriminators between accretion and stellardominated excitation processes; hence, the AGN defini-tions based on the [O III ]/H β vs. [N II ]/H α diagramshould be regarded with caution. For the void galaxysample, however, the separation between LINERs andSeyferts is practically independent of the chosen classifi- Constantin et al.cation method; the sample is very small in size and thefraction of border-line objects is relatively low.We note here that LINERs are not unambiguouslyconsidered to be AGN. Mechanisms alternative to ac-cretion onto a black hole, like photoionization by hot,young stars (Filippenko & Terlevich 1992; Shields 1992;Barth & Shields 2000), clusters of planetary nebula nu-clei (Taniguchi, Shioya, & Murayama 2000), or shocks(Dopita & Sutherland 1995), have proved relatively suc-cessful in explaining the optical spectra of these sources.Particularly ambiguous in their interpretation are theTransition sources. It is not clear whether these objectsare a ”composite,” with a LINER/Seyfert nucleus sur-rounded by star-forming regions, or not. Shields et al.(2006) and Constantin et al. (2007) show exampleswhere this model does not work. Thus, we consider thatamong the narrow-line emitting objects only Seyferts arebona-fide AGN, and advise more caution in generalizingthe term AGN. INCIDENCE OF AGN ACTIVITY
We present here the results of a comparative anal-ysis of both photometric and spectroscopic propertiesof the void and wall galaxies. We consider separatelyvarious types of emission-line activity, whether domi-nated by star-formation, accretion, or a mix of them.In this section in particular, we compare the rate of oc-currence/detection of different types of emission-line ac-tivity in voids and walls, both globally and at fixed hostproperties.
Global Frequencies
We present in Table 1 the percentages of strong line-emitters (all 6 lines are detected) and subclasses of ob-jects relative to the whole void and wall galaxy samples.We list these numbers for definitions involving strictlyhigh statistical significance in the line flux measurements(greater than 2 − σ level). We find that strong line-emission activity is clearly more common in voids thanin walls. This difference seems to be accounted for bythe difference in the rate of occurrence of H ii -type emis-sion. Objects of other types of spectral activity showequal or lower frequency in the most underdense re-gions relative to the crowded ones. That the fractionof H ii galaxies in the void regions (32 . . ii s are less clustered than othergalaxies (Constantin & Vogeley 2006).The systems whose spectra indicate the presence ofan accreting black hole as an ionization source showa somewhat mixed behavior. While Seyferts seem tobe equally represented in voids and walls (1 . . − σ . Constantin & Vogeley (2006) show that LIN-ERs cluster more strongly than Seyferts, so the weakerrepresentation of LINERs in the void regions is not sur-prising, and supports these results. The Transition ob- jects show an intermediate behavior between LINERsand Seyferts, being slightly less common in voids thanin walls. These trends are somewhat expected based ontheir clustering properties as well, although the signifi-cance of this comparison is only at the 1 . − σ level.It is important to note that these trends are nearly in-dependent of the classification criteria. The drop in thefraction of LINERs from walls to voids is only slightlylarger using the Ho et al. (1997a) classification, from1 .
8% (18 out of 996 void galaxies) to 4 .
4% (552 outof 12153 wall galaxies), confirming thus that this typeof activity is not frequent in the lowest density regions.Even for the Transition objects, whose definition variesthe most between Ho et al. (1997a) and Kewley et al.(2006), with the latter allowing for more objects that areeither classified as H ii s or remain unclassified by thefirst method, the fractions show the same trend, beinghigher in walls, and do not change by more than 5%.Thus, the results are consistent with those based on theKewley et al. (2006) definition, and show that AGN ac-tivity is generally less common in voids than in walls. Frequencies at fixed Host Properties
Because void and wall galaxies are characterized byquite different distributions in their morphologies andluminosities (Rojas et al. 2004), any possible variationin the AGN properties, including the frequency of theiroccurrence, could be related to such morphological dif-ferences. To examine these issues for our particular sam-ples, we present in Table 2 the median values for host r -band absolute magnitudes ( M r ) and host concentrationindices ( C as a proxy for the morphological type) forvoid and wall line-emitting objects of different spectraltypes. Void hosts are clearly less luminous than their wallcounterparts, by ≈ . C values) hosting actively line-emittinggalaxies in voids than in walls. We will thus emphasizehere a comparison of various emission properties at fixedhost brightness.Figure 2 illustrates the fractions in which Seyferts,Ls, Ts and H ii s are detected in systems characterizedby different M r and C values. We also show here how C and the stellar mass (log M ∗ /M ⊙ ) are distributed as afunction of M r for all these kinds of objects. Note thatthe stellar mass and the intrinsic brightness of galaxiescorrelate very well in both voids and walls. Thus, ourinvestigation of AGN fractions and other emission-lineproperties at fixed brightness apply as well to the cor-responding bin in stellar mass. Throughout the paper,we will interchangeably characterize galaxies by stellarmass or absolute magnitude. The distribution of C asa function of M r is generally flat given the errors, andthere is no significant difference in the C values of voidand wall systems at any given brightness. LINERs arean exception: the ones in walls show earlier type hostsat larger brightness while the ones in void exhibit theopposite trend, giving rise to significant differences in C between void and wall systems of a given M r .When fractions are compared at fixed luminosity and C = R /R , where R and R are the radii from the centerof a galaxy containing 50% and 90% of the Petrosian flux measuredin r -band. GN in Voids 5fixed concentration index some interesting trends standout. The most striking feature is that the objects thatshow signs of AGN activity, i.e., the Seyferts, Ls, and Ts,are clearly underrepresented among the most luminousor massive ( M r < −
20, log M ∗ /M ⊙ > .
5) hosts invoids. Among moderately bright, medium mass galax-ies ( M r ≈ − .
5, log M ∗ /M ⊙ ≈ . ii s,their frequencies among void galaxies are clearly higherthan in wall galaxies, at almost all host brightnessesand morphologies. This result agrees well with previousfindings of higher specific star formation rates in voidgalaxies relative to their wall counterparts (Rojas et al.2005), with the general tendency of actively star-formingsystems to be more common in the most underdenseregions than in other environs, and in particular withthe fact that H ii s are less clustered than other galaxies(Constantin & Vogeley 2006).It is noteworthy that Seyferts appear to favor differentkinds of hosts in voids and walls. In voids, they con-centrate almost exclusively in galaxies with M r ≈ − − . & M r . − . M r . − .
5. On the other hand,where void and wall Seyfert galaxies overlap in bright-ness, Seyfert-like activity is more frequent in underdenseregions than in more crowded environments. Finally,void Seyferts seem to prefer earlier type galaxies thanthose living in walls (there are no void Seyferts in galax-ies with C . . M r ≈ − C ≈
3) seem more frequent thanthe wall ones, but they are similarly common in void andwall hosts of late-type morphologies. This is interestinggiven that void emission-line systems are generally ab-sent among early-type hosts.There are only three actively line-emitting void galax-ies at M r ≈ −
21. One object is spectrally classified as aT, while the other two remain unclassified because theydo not comply simultaneously to the Seyfert-LINER sep-aration criteria in the diagrams involving [S II ] and [O I ],and thus have properties that are somewhat intermediatebetween a Seyfert and a LINER. Even if these two objectsare both Seyferts or both Ls, there is no statistically sig-nificant rise in the fractions of void AGN correspondingto this brightness range. The error bars of these frac-tions remain rather large; with either 2 Ls or 2 Seyfertsin this absolute magnitude range, their fraction in voidswould be 0 . ± . Frequencies and the Strength of Accretion Activity
That void and wall galaxies host accretion activityto a different degree should not come as a surprise.Kauffmann et al. (2004) showed that at fixed galaxymass, the low-density regions host twice as many AGNas the high density ones. We find here, however, thatsuch a trend extends to voids only for galaxies that aremoderately bright and moderately massive ( M r ≈ − < log M ∗ /M ⊙ < . M ∗ , in both voidsand walls, it is clear that there is no statistically signif-icant surplus of any type of AGN in the most massivevoid galaxies relative to their wall counterparts.For a more direct comparison with Kauffmann et al.(2004) results, we also examine the fractions in whichvoid and wall galaxies of a given stellar mass host AGN.Using the exact definition of AGN samples employed byKauffmann et al. (2004), we show these fractions in Fig-ure 3, separately for strong and weak [O III ] line emit-ters. It is readily apparent that AGN activity at alllevels is equally frequent among massive void and wallgalaxies (log M ∗ /M ⊙ > . M ∗ /M ⊙ < . M ∗ /M ⊙ > .
5) galaxies with stronglines ( L [OIII] > erg s − ) is 9% in walls and only 11%in voids, which is far from the 100% difference found byKauffmann et al. (2004). On the other hand, for thelog M ∗ /M ⊙ < . III ]luminosities in objects of high and medium stellar massranges, separately for void and wall galaxies. Among themassive galaxies, only the low-accretion (log L [OIII] /ergs − ≈
37) systems appear marginally more frequent invoids than in walls, while all levels of AGN activity seemmore common in voids than in walls among less massivehosts.To summarize, the environmental dependence of therate of occurrence of the AGN phenomenon varies withthe properties spanned by their hosts. Our compara-tive analysis conducted at fixed host properties showsclearly that nuclear accretion is more frequent in voidsthan in walls among moderately massive and luminous( M r ∼ −
20, log M ∗ /M ⊙ < .
5) hosts while it remainssimilarly common among massive, luminous ( M r . − M ∗ /M ⊙ > .
5) void and wall galaxies. Theseresults show that accounting for host galaxy properties,i.e., luminosity and mass, is essential for understandingthe environmental behavior of AGN activity; previousfindings of roughly constant fraction of galaxies hostingAGN across a wide range of environments (Miller et al.2003) are justified because they are based on investiga-tions of bright galaxies ( M r ≤ −
20) only.
Incidence of radio activity
Constantin et al.Scrutiny of potential differences in radio activity be-tween wall and void galaxies is of great interest for un-derstanding the environmental dependence of accretionactivity as radio observations overcome obscuration. TheFaint Images of the Radio Sky at Twenty centimeters sur-vey (FIRST; Becker,White & Helfand 1995) offers theadvantage of providing high angular resolution and sen-sitivity, and therefore is well suited for extracting infor-mation regarding galactic nuclear activity.We cross-matched the SDSS void and wall galaxy sam-ples described in Section 2 with the FIRST catalog ofsources with flux densities exceeding 1 mJy at 1.4 GHz.The global census of SDSS void and wall galaxies thatshow radio activity at this level is surprisingly differentin voids and walls. Table 3 lists the results of cross-matching statistics and a comparison of the νL ν (1.4GHz) measurements per galaxy type. Only 16 (1.6%)of void galaxies are brighter than 1 mJy at 1.4 GHz andthey are all strong line emitters. Within walls, the frac-tion of objects with flux densities exceeding this limit issignificantly higher: 308 objects (or 2.5%), about 80%of them showing line-emission activity. These calcula-tions indicate at greater than 2- σ that radio activity inthe centers of galaxies is less common in voids than indenser environments.This trend is not equally shared by galaxies of differ-ent spectral properties. We note that none of the voidLs are detected in FIRST while a fraction as high as 7%of the wall Ls appear as FIRST sources. Seyferts andTs show very similar, if not identical detection rates inFIRST in voids and walls; their corresponding fractionsof FIRST detections account for 15%–20%, and 9%, re-spectively. Radio active H ii s are slightly more com-mon in walls than in voids. While the difference in theFIRST detection rate between all void and wall galaxiesis statistically significant, the variation of radio activityper spectral type remains ambiguous, particularly for thevoid AGN where the number statistics remain low. If asingle void LINER were detected in FIRST, the fractionsof void and wall Ls with radio activity would become in-distinguishable within the errors (5% within voids versus7% within walls). The level of radio activity also showsdifferences between the void and wall systems, with thewall ones being clearly more luminous at 1.4 GHz. Aper type comparison remains however pointless becauseof the very small number of void galaxies with measur-able nuclear radio emission. We present, for reference,the νL ν (1.4 GHz) luminosities in Table 3 as well. Notealso that none of the void galaxies considered here wouldbe considered radio loud; the highest radio luminosityamong void galaxies is νL ν (1.4 GHz) = 1 . × ergs − , or L ν (1.4 GHz) = 3 × W Hz − .These measurements are consistent with previous in-vestigations of the large scale structure of radio sources,but reveal potential new features as well. In partic-ular, the fact that radio activity is enhanced in wallregions relative to voids is in line with evidence thatradio loud galaxies are locally restricted to the super-galactic plane, avoiding the voids (e.g., Shaver & Pierre1989). However, that the radio emitting H ii systemsare less common, and probably weaker in voids than indenser regions, is at odds with previous indications of adiminution in the star forming activity with increasing local density, for both radio and optically selected star-forming galaxies(Lewis et al. 2002; G´omez et al. 2003;e.g., Balogh et al. 2004; Christlein & Zabludoff 2005;Best 2004; Doyle & Drinkwater 2006). Note that whilethese studies examine the overall star-formation activ-ity within galaxies, either through galaxy colors or H α equivalent widths, our comparison refers to this kind ofactivity in centers of galaxies only. Thus, taken at facevalue, the discrepancy shown by our measurements mightindicate that core star-formation activity shows some-what opposite environmental dependence trends relativeto that occurring in the envelope; in lower density re-gions, star-formation activity seems weaker in the corebut stronger in the envelope of galaxies. That color gra-dients might also depend on galaxy density, probably asstrongly as star formation rate does, in the sense thatgalaxies in rarefied regions exhibit more bluer envelopesthan those in crowded environs (Park et al. 2007), ap-pear to support this idea.A possible explanation for this difference may be thatthe radio emission associated with nuclear star-formationactivity is not entirely due to stellar activity, but it in-cludes an additional component. A possible such contri-bution may be an AGN that is heavily obscured, or itsemission-line characteristics are swamped into the lightof the host galaxy or the circumnuclear star-forming re-gions; note that the definition of the H ii class relies on optical line emission properties, and thus is not sensitiveto obscured accretion. That AGN show possibly strongerradio emission in walls than in voids might thus suggestthat the enhanced radio detection and emission in thewall H ii s comes from an accretion source that is heavilyobscured. Although not detected in optical wavelengths,such an AGN would be revealed at radio frequencies.This picture seems supported by other studies; e.g., hy-drodynamical simulations of galaxy evolution processes,through merger events, predict that for a large fractionof the accretion time, the AGN (quasar) is heavily ob-scured, and that the intrinsic quasar luminosity peaksduring an early merging phase, but it is completely ob-scured in optical (e.g., Hopkins et al. 2006), and galaxyspectral analyses indicate that much of the BH growthoccurs in AGN with high amounts of dust extinction(e.g., Wild et al. 2007). We explore this idea furtherin the next sections where we compare accretion rates,obscuration level, and stellar properties of void and wallsystems of different optical classifications. NUCLEAR ACTIVITY AND HOST PROPERTIES
We explore here common traits and differences in therelation between nuclear emission-line activity and thehost star-formation that void and wall galaxies of variousspectral types may reveal. We gauge the properties of thevoid and wall emission-line galaxies by means of nuclearaccretion rate, amount of fuel, and properties of theirhost stellar population. The parameters used in quanti-fying these characteristics employ emission-line measure-ments and calculations described by Kauffmann et al.(2003a,b). Our analysis is in some ways similar toKauffmann et al. (2004), however, we concentrate herespecifically on the least populated regions in the uni-verse, the voids. This work constitutes the first in-depthscrutiny of Seyferts, LINERs, Transition objects, andH ii s in these extreme environments.GN in Voids 7 Accretion and Availability of Fuel
Following Constantin & Vogeley (2006), we investi-gate the accretion power in the wall and void activegalaxies in terms of both L [OI] and L [OIII] . While the[O III ] λ I ] λ ii s, [O III ] is one of themost prominent features while [O I ] is hardly measur-able. The [O I ] and [O III ] line fluxes are not simply ascale down (or up) of one another, the [O I ]/[O III ] lineflux ratio is significantly higher in AGN than in otherline-emitting systems (Heckman 1980). Thus, particu-larly for H ii and Transition objects, a comparison of the L [OI] and L [OIII] based estimates of central accretion ac-tivity is an effective way of assessing the validity of thesemeasures. Section 4.2 discusses this issue in detail. Asproxy for the accretion rate we measure the Eddingtonratio as quantified by log ( L [OI] / σ ) and log ( L [OIII] / σ ),similar to Kewley et al. (2006). We gauge the availabil-ity of fuel through estimates of the amount of intrinsicobscuration, expressed in terms of Balmer decrementsH α /H β , and the density of the emitting gas n e , as givenby the [S II ] λ λ M r , compare in voidand wall galaxies of various spectral classifications. It isinteresting to note the different environmental behaviorthat AGN hosted by bright ( M r < −
20) and moderatelybright ( M r & −
20) galaxies exhibit. Among the mod-erately bright hosts, both LINERs and Seyferts tend tohave lower accretion rates in voids than in walls, while inbright galaxies, on the other hand, the small fraction ofvoid AGN activity (almost exclusively LINER-like) ex-hibits higher accretion rates. Another interesting featurerevealed by this plot is a potentially close relation be-tween the rates of accretion (i.e., Eddington ratios) andthe surrounding gas and obscuration: lower accretionrates in void AGN are invariably associated with lowerBalmer decrements and gas densities, while higher accre-tion rates appear whenever the Balmer ratios are higher.While the statistical significance of each of these trends isonly of order 1- σ , there is a quite obvious phenomenolog-ical continuity: these statistically-independent measure-ments systematically follow the pattern expected if theaccretion rate correlates with the amount of dust and gasin their hosts, and thus probably with that of materialavailable for accretion.The boost in accretion in the LINERs inhabitingbrighter galaxies is an interesting finding and it is worthinvestigating its origin. Greater availability of accret-ing material can surely trigger such behavior. However,the origin of such an enhancement is not obvious, giventhat their fainter counterparts exhibit the opposite effect.One possibility would be that these systems have morematerial available for accretion simply because matteris more abundant here. Objects with stronger gravita-tional potential fields should be more capable of keepingthe matter inside, against galactic winds built by thethermal pressure from supernovae, while low-mass sys-tems would eject their gas. Thus, a look at the mass of their inner black holes ( M BH ) should shed some lighton this issue. Interestingly, the comparison of their stel-lar velocity dispersions ( σ ∗ ), as a measure of their M BH (since M BH ∝ σ ∗ , e.g., Ferrarese & Merritt 2000; Geb-hardt et al. 2000) shown in Figure 4 reveals that thevoid LINERs hosted by brighter galaxies have systemati-cally smaller mass BHs than similar wall galaxies. Thus,these particularly active accreting void systems do nothave, as conjectured, higher binding energies, and con-sequently, they are not more powerful in cradling thematerial available for accretion. Another sensible way toproduce the excess of material (i.e., dust) in or aroundthe LINERs hosted by bright void galaxies is star for-mation. We explore this idea in the next two sections,both by comparing the [O I ] and [O III ] line emission andthrough measures of the age of the stellar population inthese sources. [O I ] vs. [O III ] Because the [O I ] and [O III ] line emission mecha-nisms are differently affected by star-forming and accre-tion activities, the differences we see in these parame-ters between void and wall line-emitting galaxies maysuggest some interesting connection between the twosources of ionization. Although [O
III ] is a very com-mon emission feature in star-forming galaxies, for themost luminous systems (log L [OIII] /erg/s & L [OIII] ob-jects, which are LINERs (Constantin & Vogeley 2006;Kewley et al. 2006), any other sample characterizedby log L [OIII] /erg/s < ii -like ionization than an accreting power source(Constantin & Vogeley 2006). On the other hand, [O I ]is extremely weak in systems whose ionization is domi-nated by star-formation, thus [O I ] emission is an impor-tant indicator of excitation due to accretion. The reasonis that the [O I ] emission line arises preferentially in azone of partly ionized hydrogen, which can be quite ex-tended in objects photoionized by a spectrum containinga large fraction of high-energy photons, i.e., originatingin accretion, but is nearly absent in galaxies photoionizedby young, hot (OB) stars. Therefore, more vigorous star-formation activity would cause stronger [O III ] emission,but should barely affect [O I ]. Correspondingly, the [O I ]emission is expected to be enhanced in systems whereaccretion is stronger.Interestingly, the analysis of the [O I ] and [O III ] lineluminosities that we present in Section 4.1 shows thatthese two parameters compare remarkably well, reveal-ing exactly the same trends: both L [OI] and L [OIII] areslightly lower in voids than in walls in approximatelythe same amounts (0 . . . L [OIII] is, however, generally larger than that in[O I ], thus diluting the significance of the trends in log L [OIII] with M r . The larger scatter in L [OIII] is proba-bly caused by the contribution of star-forming activity tothis line. Such a contribution is expected to be even moreenhanced in void galaxies as they exhibit higher specificstar formation rates than their wall cousins (Rojas et al. Constantin et al.2005). Note also that for the LINERs hosted by voidbright galaxies, which show also evidence for youngerstellar populations (see Section 4.3), the enhancementin L [OIII] is larger than that in L [OI] ; such a differencesignals again contribution by star-forming activity to the[O III ] line emission. Thus, particularly for void galax-ies, L [OIII] must be used with caution when employed inestimating the accretion rate. Relation to Star Formation
A variety of mechanisms have been suggested by whichstarbursts and AGN are physically related. An inter-esting idea is that the collapse of very massive starscan give rise to the seed black hole, which can be fedby stars either directly through tidal capture or indi-rectly through gas shed via stellar mass loss. Over time,the BH and its host galaxy grow in size, but proba-bly not in lock-step. Because more massive galaxiesare less likely to harbor young stars (Kauffmann et al.2004), and have more massive BHs (Magorrian et al.1998; Gebhardt et al. 2000; Ferrarese & Merrit 2000),it is believed that the mass of the BH might have directconsequences for star birth. Hydrodynamic simulationsof galaxy mergers including black hole growth and feed-back (e.g., DiMatteo et al. 2005; Hopkins et al. 2005;Croton et al. 2006) demonstrate how energetic outflowscaused by gas falling onto the central BH might self-regulate both BH growth and star formation in a galaxyby heating the available gas and blowing it out, as previ-ously predicted by e.g., Silk & Rees (1998), or heatingit up to temperatures that suppress clumping and col-lapsing into stars. Further star-formation is thereforerepressed as long as M BH remains above a critical value(e.g., Schawinski et al. 2006).The regions least prone to galaxy-galaxy interactionsor other disturbing events, the voids, are some of the besttest beds of this scenario. We proceed in this section witha comparative analysis of the stellar population ages forvarious types of actively line-emitting galaxies in voidsand walls. The strength of the 4000˚A break ( D δ absorption line (H δ A ) pro-vide constraints on the mean stellar age of a galaxy andthe fraction of its stellar mass formed in bursts over thepast few Gyr (Kauffmann et al. 2003a). For the sakeof comparison with previous work (Kauffmann et al.2003b, 2004; Kewley et al. 2006), we use the measure-ments of the stellar characteristics derived and describedby Kauffmann et al. (2003a) and Brinchmann et al.(2004). Figure 5 shows the results of this comparison,in bins of M r .Globally, the void systems exhibit younger collectionsof stars than those in walls: on average, D δ A is signifi-cantly lower (by ∼
30% in Seyferts, and ∼
60% in Ls) inthe more crowded environments than in the underdenseones. This comparison supports previous indications ofsystematically younger galaxies in lower-density regions(Kauffmann et al. 2004). Here we find that these trendsextend to the most extreme underdense environments aswell. We also show here that such differences are due tothe different range of host properties spanned by the sam-ples that are compared: the younger mean stellar agesof void systems is a consequence of the general shortage of massive bright galaxies (i.e. the usual hosts of oldstellar populations) in the underdense regions. The shiftin the void luminosity function toward lower brightnessgalaxies is nicely documented in Hoyle et al. (2005).At fixed host brightness, the properties of the centralstellar populations of void and wall galaxies are statisti-cally similar, however, there are exceptions. The bright-est ( M r < −
20) void hosts of LINERs stand out as in-teresting peculiarities, as they are clearly associated withyounger stellar populations (larger H δ A , smaller D M r ≥ −
20) void hosts ofLINERs, which show weaker accretion activity and sig-nificantly less obscuration, also show signs of older stel-lar emission (lower H δ A values). All these trends in-dicate that the strength of AGN activity is correlatedwith the age of the associated stellar population, in thesense that AGN activity is stronger when the surround-ing stars are relatively young (few Gyr old, correspondingto D . .
6, and H δ A & D & .
8, and H δ A . THE SMALL SCALE ENVIRONMENT: NEARESTNEIGHBOR STATISTICS
A key point in investigating AGN in voids is under-standing at what scale, if at all, the environment affectsaccretion activity. As shown by Constantin & Vogeley(2006), different types of galactic activity prefer differentlarge scale structures; LINERs are most clustered, H ii sare the least clustered, Seyferts and Ts are less clusteredthan Ls but more clustered than H ii s. Dark matterdensity fluctuations are however important at all scales.In fact, because the fractional density fluctuations arelargest on the smallest scales, any effect that depends ondensity will appear to be most strongly correlated withsmaller than with larger scales. One would thus expectthat the small scale climate also influences the strengthand type of accretion or star-formation, and even the linkbetween them.The rate of interaction is supposedly reduced in voids,and thus, an analysis of the effects of small scale environson galactic activity is particularly instructive. In partic-ular, it is of interest to find out whether objects display-ing a certain type of activity are located in the highest orlowest density substructures within the most underdenseenvirons, and whether AGN are associated or not withsmall groups within voids. To investigate these issues wemeasure the distance to the third nearest neighbor ( d )as a density indicator, and the distance to the first near-est neighbor ( d ) as a measure of the probability ofGN in Voids 9having close encounters. We measure d separately forthe volume limited sample, which thus counts only thebright ( M r < − .
5) neighbors, and for the whole (fluxlimited) sample, which adds fainter galaxies. We notehere that a small number of close neighbors are misseddue to the 55 ′′ minimum fiber separation; this fiber col-lision issue is however unlikely to affect significantly ourcomparative analysis of near neighbor statistics of voidand wall samples as the effects should be nearly identicalfor all samples involved, and thus should cancel out.Table 4 lists the median values in these parametersfor both void and wall systems of various spectral types,and Figure 6 illustrates how these parameters compareat fixed M r . Because the definition of voids and walls em-ploys d , all void galaxies have larger d than those inwalls. As expected, d varies little among void galax-ies , the median values are consistent with each otherwithin 1 − σ . It is interesting to note however an ap-parent tendency of increasing d with host luminosity(Figure 6); this suggests that, in voids, more luminousand therefore massive systems are less grouped than thefainter, less massive ones.Among the wall galaxies, d shows significant dif-ferences among objects of different spectral types, andtrends that are opposite to those exhibited by void sys-tems: LINERs have the closest 3rd neighbor (with anaverage of 3 . ± . h − Mpc), and thus the highest den-sity, while the H ii s populate less dense regions withinwalls (their average d ∼ . h − Mpc). Seyfertsand Ts show very similar values ( d ∼ . h − Mpc),which are intermediate between LINERs and H ii galax-ies. These trends persist even when the samples are splitin bins of M r , and it is pretty clear that d decreaseswith M r (Figure 6). Such findings are consistent withthe difference in clustering that Constantin & Vogeley(2006) reveal, with LINERs being the most clustered ac-tive galaxies, Seyferts showing lower clustering amplitudeand H ii s being the systems that are the least clustered.Note however that the clustering analysis involves auto-correlations of each type of these emission-line objects,while the nearest-neighbor analysis cross-correlates eachtype of object with the full population of galaxies.Distance to the nearest galaxy, d , may be impor-tant in assessing whether or not interactions are impor-tant in triggering nuclear activity in galaxies. The d measurements generally reinforce the behavior shown by d but also show some peculiar trends. When mea-sured within the volume limited sample, d indicatesthat, in voids, LINERs and Seyferts are closer to theirbright neighbors than the H ii s while Ts show a some-what intermediate behavior. Kolmogorov-Smirnov testsare consistent with clear differences in the distributionsof H ii s and LINERs, (the probability that the d dis-tributions of these two samples share the same parentpopulation is Prob HII − L = 0 . The fact that we only see small variation in d in voids is notsurprising. In these regions, we are measuring distances > h − Mpc, and hence, we are looking at quite large smoothing scaleswhere the rms density fluctuations are smaller than on the scaleswhere d is measured in walls. ERs have the closest neighbors, both faint and bright,while H ii s are the farthest away from any galaxy. Thesetrends suggest that actively star-forming systems preferthe densest sub-regions in voids and the most rarefiedneighborhoods within crowded environs, while the sys-tems in which accretion is strong enough to be detectedis generally associated with a grouped small-scale envi-ronment.The comparison of these distances in bins of M r , asshown in Figure 6, may explain these particular trends.For all three measures of density, and in particular for d , the differences between void and wall galaxies ofgiven brightness is greater for the more luminous objects:the brighter a galaxy is, the more isolated it seems to bein voids, while the most “connected” in walls. Alongthese trends, H ii s, which are generally fainter thanother types of objects, occupy more crowded void sub-environments but more rarefied regions in walls, whileLINERs, which inhabit generally bright hosts, are themost isolated in voids and the most grouped in walls.That more luminous galaxies appear to prefer morecrowded regions within walls is consistent with previousresults that more luminous systems are more clustered(on large scale). However, their tendency to be more iso-lated within voids is a new and surprising finding. Thisobservation suggests that while the small scale environ-ment may influence the intrinsic properties of their in-habitants (e.g., M r ), the dynamics that determine thelarge scale structure (e.g., expansion versus contraction)are equally important as they influence the likelihood forinteraction, and thus the type of galactic nuclear activity.It is important to note that both d and d addressthe embedding of the AGN hosts and other galaxies indark matter halos, however, at different scales: d , oforder 1.5 Mpc h − in walls and about three times higherin voids, examines neighbors that may lie within the samedark matter halo, while d , which is ∼ − inwalls and ∼ − in voids, measures the largescale environment traced by multiple halos. Uncertaintyin the d measure arises because of (1) the fiber colli-sion problem and (2) peculiar velocities along the line ofsight. Both effects will tend to increase d by removinga few close neighbors and by spreading virialized systemsalong the line of sight, respectively. Thus, the d val-ues in walls, where d is small enough to be affected byboth systematics, may be overestimates. In voids, thesesystematic effects go in the same direction, however arelikely to be smaller, because the density is much lower.The d measurements are likely to be affected by thesebiases in a similar manner, however much less. Thesearguments indicate thus that our conclusions can in factbe stronger. In particular, the differences in d couldcertainly not be caused by systematic effects, and aregenerally underestimated. Moreover, the comparison be-tween types of active galaxies in voids and walls shouldbe hardly affected, as the spectral classification and nearneighbor statistics are completely independent analyses.To summarize, the youngest, actively star-forming,fuel-rich systems, i.e., the H ii s, appear in weaklygrouped environments, that are the densest regionswithin voids and the rarefied ones within walls. On theother hand, the most common type of AGN activity (i.e.,LINER-like), which is relatively feeble and associatedwith old massive evolved hosts, shows a peculiar pref-0 Constantin et al.erence for the most clustered wall substructures and theemptiest among voids. This trend argues against a majorrole played by close encounters or mergers in triggeringor sustaining their activity. Interestingly, the most activeAGN (Seyferts and Ts) prefer the relatively low den-sity wall sub-environments and the somewhat groupedvoid ones; this is a behavior that is intermediate betweenthose manifested by LINERs and H ii s. CONCLUSIONS AND DISCUSSION
Summary of Results
We use the largest sample of voids and void galaxiesyet defined to investigate the galactic nuclear accretionphenomenon in the most underdense regions, in relationto their more populous counterparts. By employing spec-troscopic and photometric data based on the SDSS DR2catalog, and in particular measurements available in theGarching catalog, we conduct a comparative analysis ofvoid and wall systems of different radiative signature.We find that all types of Low Luminosity AGN ex-ist in void regions. However, their occurrence rate andintrinsic properties show variation from their wall coun-terparts. The differences between the wall and void AGNseem to be driven by the properties of their hosts, whichare correlated with (or governed by) their small scale en-vironment. Following is a summary of our main results:(i) Among moderately bright or fainter galaxies (10 20 galaxies, have the tendency to accrete at lowerrates than in those in walls. This behavior seems relatedto the fact that the void systems show less obscurationand, perhaps, less dense emitting gas. That the stellarpopulations associated with void and wall AGN are sim-ilarly aged suggests that fuel might be equally availablefor accretion in void and wall galaxies of similar prop-erties (i.e., M r ), but that fuel is less efficiently driventowards the nucleus in void galaxies.(iii) The few void AGN hosted by bright, massivegalaxies ( M r . − 20, log M ∗ /M ⊙ > . 5) are LIN-ERs that show peculiarly higher accretion rates, largeramounts of obscuring matter and more recent star-formation than in their wall counterparts. These par-ticular systems reinforce the general trends other objectsshow: higher accretion rates are invariably associatedwith younger stellar populations and higher obscuration.These trends suggest that the amount of obscurationcould be a measure of the available fuel for both starformation and accretion.(iv) The radio activity of line-emitting galaxies appearsboth less frequent and weaker in voids than in walls.Were we able to support these differences with statis-tically significant measurements, they would imply thatcentral radio activity in wall systems, including H ii s, ismore pronounced because it builds on contributions fromaccretion that remains optically obscured and thereforeundetected. (v) Nearest neighbor statistics show that the type ofemission-line activity is correlated with the small-scalelocal environment. The star-forming regions (the H ii s),populate the most crowded sub-regions of voids whilepopulating relatively sparse regions in walls; both areenvironments where low-mass galaxies recently formed.The weakly active galaxies (LINERs) live within the clus-ters in walls but the most rarefied regions in voids. Thisfinding is puzzling and suggests that these systems, whichare generally old, were probably not aware of their envi-ronments when they formed. Actively accreting systems(Seyferts and possibly the Ts) inhabit intermediate re-gions, which are relatively dense galaxy neighborhoodsin voids but are of average density in walls.(vi) These correlations among the type and strengthof galactic nuclear activity, incidence rates of differenttypes, and their small and large scale environments,suggest an H ii → S/T → LINER evolutionary scenario inwhich interaction is responsible for propelling gas to-wards the galaxy centers, triggering star formation andfeeding the active galactic nucleus. The H ii → S/T → LINER EvolutionarySequence Figure 7 illustrates how various intrinsic and hostproperties of actively line-emitting galaxies follow this H ii → S/T → LINER sequence. The early stages of such ob-jects manifest themselves as H ii , as the accreting sourceremains heavily embedded in dust. As the star-burstfades in time, the dominance of the Seyfert-like excita-tion in systems of generally small but actively accretingblack holes becomes more evident. Successive evolutionreveals aging stellar populations associated with objectsspectrally classified as Transition objects, that are stillshowing signs of accretion, followed by LINERs, whosestars are predominantly old and whose accretion onto al-ready grown-up BHs is close to minimal. Note that thisH ii → S/T → LINER progression is very similar in wallsand voids. The lower accretion rates and the higher fre-quency of actively accreting systems in void versus wallgalaxies of similar properties indicate a potential delayin the AGN dominance phase within voids. Thus, voidAGN progress through the H ii → S/T → LINER sequencemore slowly, while the sequence is similar for both voidand wall galaxies. This picture fits well the observedproperties of each type of galaxy nucleus: • That H ii -type of activity is significantly more fre-quent in voids than in walls suggests that theirvoid-like environments, in which they have closer(both 1st and 3rd) neighbors than other types ofobjects have, are essential in triggering their activ-ity. In other words, close encounters that produceeither harassment, or major and/or minor merg-ers may be an important cause for igniting bothaccretion and star formation. • Seyferts’ environments in both voids and walls areintermediate between those of H ii s and LINERs,regardless of their brightness. If the Seyfert-likeactivity is triggered by interactions, probably thesame ones that turn on the H ii s, there must bea time lag between the onset of the star-burst andwhen accretion becomes dominant, or simply ob-servable. Such a time interval corresponds to aGN in Voids 11period of aging of the stellar population (as seen inthe differences in D δ A between H ii sand Seyferts), when the post starburst fuel be-comes increasingly available for accretion. More-over, this progression develops relatively uninter-rupted in voids, and therefore, possibly, at a slowerpace than it would in walls where close encountersor other types of interactions can either accelerateor terminate it. • Void and wall Ts are barely distinguishable in theirphysical characteristics. In both voids and walls,their nn-statistics and intrinsic and host proper-ties are intermediate between those of LINERs andH ii s, for any given range of M r . Their BHsare apparently growing (their accretion activity isstronger than that of Ls but weaker than that ofSeyferts), their fuel supply seems plentiful (theyare found in some of the most obscured systems),and they are associated with (quite massive) stel-lar populations that are generally younger thanthose of LINERs, but older than the majority ofstar-forming systems. It is among Ts howeverthat the most massive void accreting BHs are ob-served; this might suggest that, in the proposed H ii → S/T → LINER sequence, massive void galaxiesreach the low accretion rate (i.e., LINER) phaselater than in walls. • Whether Seyferts or Transition objects are first inthis sequence remains ambiguous. Both their in-trinsic properties and nn-statistics are very sim-ilar, and remain intermediate between those ofH ii s and LINERs. The H α /H β Balmer decre-ments and the d are the only parametersthat show a “jump” in the otherwise smoothH ii → S → T → LINER sequence manifested by otherproperties of these systems. Further investigationsof these objects should address the differences be-tween them in terms of a possible evolutionary pro-gression. • Although we do not provide any quantitative esti-mates of the time spent in or during the variousphases we propose here, this this analysis showsthat both void and wall galaxies follow the samecycle. The AGN evolution does not affect the grav-itational environment. To the contrary, it is theenvironment that sets the time scale for evolutionalong such a sequence; it seems to take longer tomarch through the different phases in voids than inwalls, but the physics is the same. The large scaleclustering is consistent with this picture: LINERsare now more clustered because objects in dense re-gions underwent the H ii → S/T → LINER evolutionmore quickly; the higher rate of galaxy-galaxy in-teractions speeds up the way AGN proceed throughthe sequence. Hosts whose central regions are nowin the H ii phase will always be less clustered thancurrent LINERs.Although far from being complete, this proposedevolutionary sequence is engaging and offers a com-prehensive picture for the co-evolution of AGN and their host galaxies. The broad idea that merg-ers trigger star formation and that the AGN ap-pears afterwards, in fact shutting off the star forma-tion because of feedback, has been discussed previ-ously in the literature. For example, N-body simu-lations by Byrd et al. (1986) and Hernquist & Mihos(1995) showed that interactions drive gas toward the nu-cleus and can produce intense star formation followedby an AGN. More recent state-of-the-art hydrodynami-cal models (e.g., DiMatteo et al. 2005; Springel et al.2005; Hopkins et al. 2005; e.g., Hopkins et al. 2006)show that during mergers, the BH accretion peaks con-siderably after the merger started, and after the star-formation rate has peaked. However, whether earlybright quasars and later, dimmer AGN obey similarphysics needs still to be addressed. The H ii → S/T → Lsequence that this study reveals, based on the smoothalignment of several of their spectral properties, may bethe first empirical evidence for an analogous duty cyclein high redshift bright systems and in nearby galaxieshosting weak quasar-like activity.This scenario can also accommodate the rather in-conclusive findings regarding the role of mergers in ac-tivating AGN: their hosts do not show evidence forbars (e.g., Mulcheay & Regan 1997; Laine et al. 2002)or disturbances caused by galaxy-galaxy interactions(e.g., Malkan et al. 1998), exhibit morphologies verysimilar to those of field galaxies (de Robertis et al.1998a,b), and pair counting in both optical and IRremain inconclusive as possible excess of companionsare sometimes found (Dahari 1984; Rafanelli et al.1995), but not always (Fuentes-Williams & Stocke 1988;Laurikainen & Salo 1995). Moreover, Schmitt (2001,2004) show that claims of evidence of interactions in theliterature could be attributed to selection effects. TheH ii → S/T → LINER cycle suggests that the majority ofAGN might be detected only a certain period after theinteraction, allowing time for the starburst to fade andfor the BH accretion to gain strength.One might argue that that other forms of evolutioncould also exist for these emission-line galaxies, as op-posed to this simple progression. We note that this pro-posed sequence does not imply that every HII galaxy atthe present epoch must necessarily become a LINER; itis certainly possible that some systems go through H ii -only phases, or L-only phases, which might not be partof the larger progression.We would however like to emphasize that thetimescales necessary to transform from one galaxy typeto the next are quite reasonable. At a first glance,this is somewhat surprising given the relatively largerange in BH masses (2 × M ⊙ for H ii s, to 2 × M ⊙ for LINERs, inferred from σ ∗ ’s assuming, e.g.,Tremaine et al. 2002), and their significantly low ac-cretion rates ( L/L Edd . . L [OIII] /σ ∗ = − L/L Edd = 0 . 05, according to Kewley et al. 2006); appar-ently, for a canonical value for the accretion efficiency,i.e., 10%, it takes approximately a Hubble time to e-foldin BH mass. The key is in the fact that the low lu-minosity AGN accrete inefficiently, as very little energygenerated by accretion is radiated away (the opticallythin cooling time of the gas is longer than the inflow2 Constantin et al.time). Studies show that, in these cases, the radiativeefficiency can be as low as 10 − (e.g., Rees et al. 1982;Narayan & Yi 1994; Quataert 2003). With a (not anecessarily extreme) efficiency value of 10 − then, the e-folding time in the BH mass is ≈ . × /l years, andthus, as little as few Myrs for, e.g., l = L/L Edd = 0 . ii → S/T → LINER evolutionaryscenario are clearly needed. These tests require largersamples that allow separation into different morpholog-ical subsamples, and observables that parametrize thegalaxy morphology better than the concentration index.Moreover, we need better constraints on the BH massesand consequently the Eddington rates for the H ii s inparticular, or the late type galaxies in general. Whenavailable, analysis of such parameters would shed light onthe assumed coevality of star-formation and Seyfert-likeBH accretion in centers of galaxies, removing ambigui-ties regarding the initial stages of the H ii → S/T → REFERENCESBaldwin, J. A. Phillips, M. M, & Terlevich, R., 1981, PASP, 93, 5Balogh, M., et al., 2004, MNRAS, 348, 1355Barth, A., & Shields, C. J., 2000, PASP, 112, 753Becker, R. H., White, R. L., & Helfand, David J., 1995, ApJ, 450,559Begelman, M. C., & Nath, B.B., 2005, MNRAS, 361, 1387Benson, A. J., Hoyle, F., Torres, F & Vogeley, M. S., 2003,MNRAS, 340, 160Best, P. N., 2004, MNRAS, 351, 70Blanton, M. R., Lupton, R. H., Maley, F. M., Young, N., Zehavi,I., & Loveday, J. 2003, AJ, 125, 2276Blanton, M. R., et al. 2005, AJ, 129, 2562Brinchmann, J., Charlot, S., Heckman, T. M., Kauffmann, G.,Tremonti, C., & White, S.D.M., 2004, astro-ph/0406220Byrd, G. G., Sundelius, B., & Valtonen, M., 1986, A&A, 171, 16Christlein, D., & Zabludoff, A. I., 2005, ApJ, 621, 201Constantin, A., and Vogeley, M.S., 2006, ApJ, 650, 727 Constantin, A., et al., 2007, ApJ, submitted.Croton et al., 2006, MNRAS, 365, 11Cruzen et al., 2002, ApJ, 123, 142Dahari, O., 1984, AJ, 89, 966de Robertis, M. M., Hayhoe, K., & Yee, H. K. C., 1998, ApJS,115, 163de Robertis, M. M., Yee, H. K. C., & Hayhoe, K., 1998, ApJ, 496,93DiMatteo et al., 2005, Nature, 433, 604Dopita, M. A. & Sutherland, R. S. 1995, ApJ, 455, 468Doyle, M. T., & Drinkwater, M. J., 2006, MNRAS, 372, 977Einasto, J., Joeveer, M., Saar, E., 1980, MNRAS, 193, 353Eisenstein, D. J., et al. 2001, AJ, 122, 2267Ferrarese, L. & Merritt, D., 2000, ApJ, 539, 9Filippenko, A. V., & Terlevich, R., 1992, ApJ, 397, 79LFuentes-Williams, T., & Stocke, J. T., 1988, AJ, 96, 1235 GN in Voids 13 TABLE 1Object Sample Statistics. in voids in wallsSample N obj Fraction a N obj Fraction b Emission-line 424 42.6% 4183 33.4%...Seyfert 13 1.5% 190 1.5%...LINER 20 2.0% 510 4.1%...Transition 53 5.3% 796 6.4%...H II 327 32.8% 2610 20.8% a Fraction of all 996 void galaxies. a Fraction of all 12153 wall galaxies.Fukugita, M., Ichikawa, T., Gunn, J. E., Doi, M., Shimasaku, K.,& Schneider, D. P., 1996, AJ, 111, 1748G´omez, P. L., et al. 2003, ApJ, 584, 210Gebhardt, K., et al., 2000, ApJ, 539, 13Goldberg D. M., Jones, Timothy D., Hoyle, F., Rojas, R. R.,Vogeley, M. S., & Blanton, M. R., 2005, ApJ, 621, 643Greene, J.E., Lim, J., & Ho, P. T. P., 2004, ApJS, 153, 93Gunn, J. E., et al. 1998, ApJ, 116, 3040Heckman T. M., 1980, A&A, 87, 152Heckman, T. M., Kauffmann, G., Brinchmann, J., Charlot, S.,Tremonti, C., White, S. D. M., 2004, ApJ, 613, 109Hernquist, L., & Mihos, J. C., 1995, ApJ, 448, 41Ho, L. C., Filippenko, A. V., & Sargent, W. L. W., 1997, ApJS,112, 315Ho, L. C., Filippenko, A. V., & Sargent, W. L. W., 1997, ApJS,487, 568Hogg, D. W., Finkbeiner, D. P., Schlegel, D. J., & Gunn, J. E.2001, AJ, 122, 2129Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T., Martini,P., Robertson, B., & Springel, V., 2005, ApJ, 630, 705Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T.,Robertson, B., & Springel, V., 2006, ApJ, 163, 1Hoyle, F. & Vogeley, M. S., 2002, ApJ, 566, 641Hoyle, F. & Vogeley, M. S., 2004, ApJ, 607, 751Hoyle, F. Rojas, R. R., Vogeley, M. S. & Brinkmann, J., 2005,ApJ, 620, 618Kauffmann, G., & Haehnelt, M., 2000, 311, 576Kauffmann, G., et al., 2003, MNRAS, 341, 33Kauffmann, G., et al., 2003, MNRAS, 341, 54Kauffmann, G., et al., 2003, MNRAS, 346, 1055Kauffmann, G., et al., 2004, MNRAS, 353, 713Kewley, L., Dopita, M. A., Sutherland, R. S., Heisler, C. A., &Trevena, J., 2001, ApJ, 556, 121Kewley, L.J., Groves, B., Kauffmann, G., & Heckman, T., 2006,MNRAS, 372, 961Kirshner, R. P., Oemler, A., Jr., Schechter, P. L., & Shectman, S.A., 1981, ApJ, 248, 57Laine, S., Shlosman, I., Knapen, J. H., Peletier, R. F., 2002, ApJ,567, 97Laurikainen, E., & Salo, H., 1995, A&A, 293, 683Lewis, I. et al., 2002, MNRAS, 334, 673Lupton, R.H., 1993, Statistics in Theory and Practice , Princeton:Princeton University PressLupton, R.H., Ivezic, Z., Gunn, J. E., Knapp, G., Strauss, M. A.,Yasuda, N., 2002, Proc. SPIE, 4836, 350Magorrian, J., et al., 1998, AJ, 115, 2285Malkan, M. A., Gorjian, V., & Tam, R., 1998, ApJS, 117, 25 Marconi, A., & Hunt, L. K., 2003, ApJ, 589, 21Miller, C. J., Nichol, R. C., G´omez, P. L., Hopkins, A. M., &Bernardi, M., 2003, ApJ, 597, 142Mulchaey, John S., & Regan, Michael W., 1997, ApJ, 482, 135Narayan, R., Yi, I., 1994, ApJ, 428, 13Osterbrock, D. E., 1989, Astrophysics of Gaseous Nebulae andActive Galactic Nuclei , University Science BooksPark, C., Choi, Y., Vogeley, M. S., Gott, J. Richard, III, &Blanton, Michael R., 2007, ApJ, 658, 898Pier, J. R., Munn, J. A., Hindsley, R. B., Hennessy, G. S., Kent, S. M., Lupton, R. H., & Ivezi´c, ˇZ. 2003, AJ, 125, 1559Quataert, E., 2003, Astron. Nachr., Supp. Issue 1, in Proceedingsof the Galactic Center Workshop 2002 - The central 300parsecs of the Milky Way , p.435-443Rafanelli, P., Violato, M., & Baruffolo, A., 1995, AJ, 109, 1546Rees, M. J., Begelman, M. C., Blandford, R. D., Phinney, E. S.,1982, Nature, 295, 17Rojas, R. R., Vogeley, M. S., Hoyle, F. & Brinkmann, J., 2004,ApJ, 617, 50Rojas, R. R., Vogeley, M. S., Hoyle, F. & Brinkmann, J., 2005,ApJ, 624, 571Schawinski, K., et al., 2006, Nature, 442, 888Schawinski, K., et al., 2007, MNRAS, in press(astro-ph/0709.3015)Schmitt, H. R., 2001, AJ, 122, 2243Schmitt, H. R., 2004, in The Interplay among Black Holes, Starsand ISM in Galactic Nuclei , Cambridge Univ. Press,Proceedings of IAU Symposium, No. 222, eds. T.Storchi-Bergmann, L.C. Ho, and Henrique R. Schmitt, p. 395.Silk, J., & Rees, M.J., 1998, A&A, 331, 1Springel, V., Di Matteo, T., & Hernquist, L., 2005, MNRAS, 361,776Shields, J. C., 1992, ApJ, 399, 27LShields, J. C., Rix, H.-W., Sarzi, M., Barth, A. J., Filippenko, A.V., Ho, L. C., McIntosh, D. H., Rudnick, G., Sargent, W. L.W., 2007, ApJ, 654, 125Smith, J. A., et al. 2002, AJ, 123, 2121Strauss, M., et al., 2002, AJ, 124, 1810Taniguchi, Y., Shioya, Y., & Murayama, T., 2000, AJ, 120, 1265Tremonti, C.A., et al., 2004, ApJ, 613, 898Veilleux, S., & Osterbrock, D. E., 1987, ApJS, 63, 295Wild, V., Kauffmann, G., Heckman, T., Charlot, S., Lemson, G.,Brinchmann, J., Reichard, T., & Pasquali, A., 2007, MNRAS,submitted (astro-ph/0706.3113)York, D. G., et al., 2000, AJ, 120,1579 TABLE 2Host Properties M r C a Sample void wall void wallSeyfert -19.8 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note . — Central values are medians of parameters listed in theheader. Errors are the standard deviation of the median (= 1 . × σ mean , assuming Gaussian distributions; Lupton 1993). a The Concentration index ( C ) is defined by the ratio C = r /r r 90 and r 50 correspond to the radii at which the integratedfluxes (in r -band) are equal to 90% and 50% of the Petrosian fluxrespectively. TABLE 3Radio Properties N obj (Fraction) a νL ν (1.4GHz) b Sample void wall void wallSeyfert 2 (15%) 38 (20%) 0.9 ± ± · · · ± ± ± ± ± a counts of sources with FIRST detections at the flux den-sity threshold of 1mJy at 1.4 GHz. b in units of 10 erg s − . Central values are medians anderrors are the standard deviation of the median. TABLE 4Nearest Neighbor (nn) Statistics d d d , whole sampleSample void wall void wall void wallSeyfert 8.4 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note . — Central values are medians of distances to the 3rd and 1st nearest neighbor, in unitsof h − Mpc. Errors are the standard deviation of the median. The first two columns indicatemeasurements for objects in the M r < − . . < m r < . 7, and thus adds intrinsically faint sources. GN in Voids 15 Fig. 1.— Diagnostic diagrams for emission-line galaxies in voids and walls, for objects with relatively high ( > 2) signal-to-noiseline flux measurements. With the exception of void and wall Seyferts, and void LINERs and Transition objects, which are plottedindividually, object types are shown in density contours corresponding to factors of n of the total number of objects in each class, where n = 0 . , . , . , . , . , . Fig. 2.— Left : Fractions of void and wall galaxies that are spectroscopically classified as Seyferts, LINERs, Transition objects, and H ii s,as a function of their r -band absolute magnitude ( M r ) and their concentration indices ( C ). Right : Median values for C and (dust corrected)stellar mass (log M ∗ /M ⊙ ) in bins of 0.5 mag of M r , separately in voids (black) and walls (gray). Error bars are shown only for bins thatinclude at least two objects. Fig. 3.— Left : Fractions of galaxies containing strong and weak AGN, with log L [OIII] /(erg s − ) higher and lower than 39 respectively,plotted as a function of stellar mass, for voids (black) and walls (gray). Right : Comparison of void and wall distributions in [O III ] lineluminosities of objects with potential contribution from accretion (Ss + Ls + Ts), shown separately for two ranges in the host stellar mass.The [O III ] luminosities have been corrected for internal dust attenuation as described in the text, and are expressed in erg s − . Fig. 4.— Mean values of σ ∗ , log L [OI] /σ and log L [OIII] /σ (two proxies for the accretion rate), the Balmer decrement H α /H β , and the[S II ] λ λ M r , for galaxies in voids (black) and walls (gray). The error bars represent thestandard deviation of the mean. Data are shown only for bins that include more than 2 objects. Each row of plots represents a differentsubclass of objects, as indicated. The line luminosities have been corrected for internal dust attenuation, and are expressed in erg s − ; thestellar velocity dispersions σ ∗ are in km s − . GN in Voids 17 Fig. 5.— Mean values of the 4000˚A break ( D δ absorption line (H δ A ), two independent measuresof the mean stellar population ages. Symbols and errors are as in Figure 4. Note that with the exception of LINERs hosted by bright voidgalaxies which have younger stellar populations than their wall counterparts (smaller D δ A values), the star-formationhistories in void and wall galaxies are very similar. Fig. 6.— The 3rd and 1st nearest neighbor statistics averaged in M r bins. Symbols and colors are as in Figure 4. The tendencies shownby the median values are preserved for all brightness ranges: LINERs are furthest from other bright galaxies in voids, but have the nearestneighbors when they live in walls. GN in Voids 19 Fig. 7.— Trends of galaxy properties in voids (black) and walls (gray) along the H ii → S/T → L sequence described in the text: accretionrate, as expressed by L/L Edd , estimated based on both log L [OI] /σ ∗ and log L [OIII] /σ ∗ , BH mass, as given by σ ∗ , stellar mass, starformation rate (or youngness of the associated stellar population), and level of obscuration. Data points represent individual measurementswhile square boxes indicate median values of subsets of objects separated in four bins in M r ; the size of each square scales with therespective median M rr