Circumgalactic Oxygen Absorption and Feedback
DDraft version October 3, 2018
Preprint typeset using L A TEX style emulateapj v. 12/16/11
CIRCUMGALACTIC OXYGEN ABSORPTION AND FEEDBACK
William G. Mathews and J. Xavier Prochaska Draft version October 3, 2018
ABSTRACTOVI absorption in quasar spectra caused by intervening circumgalactic atmospheres suggests adownturn in the atmospheric column density in sightlines passing beyond about 100 kpc from centralstar-forming galaxies. This turnover supports the hypothesis that the oxygen originates in the centralgalaxies. When converted into oxygen space density using an Abel integral inversion, the OVI columnsrequire > ∼ M (cid:12) of oxygen concentrated near 100 kpc. Circumgalactic gas within this radius coolsin less than 1 Gyr and radiates ∼ . erg s − overall. The feedback power necessary to maintainsuch oxygen-rich atmospheres for many Gyrs cannot be easily supplied by galactic supernovae. How-ever, massive central black holes in star-forming galaxies may generate sufficient accretion power andintermittent shock waves at r ∼
100 kpc to balance circumgalactic radiation losses in late-type L (cid:63) galaxies. The relative absence of OVI absorption observed in early-type, passive L (cid:63) galaxies may arisefrom enhanced AGN feedback from their more massive central black holes. Subject headings: galaxies: abundances — quasars: absorption lines INTRODUCTION
The possibility of high oxygen abundances in hot, viri-alized gas around normal star-forming galaxies has at-tracted much attention. Tumlinson et al. (2011) de-scribe strong UV absorption in quasar spectra due tothe OVI doublet (1031.9, 1037.6˚A) having redshifts sim-ilar to those of ∼ L (cid:63) galaxies observed close to quasarsightlines. Distances between sightlines and the centralgalaxies, i.e. the impact parameters R , are large, imply-ing that the absorption occurs in extended circumgalacticgaseous atmospheres surrounding star-forming galaxies.Impact parameters of 10 < ∼ R < ∼
150 kpc are observed instar-forming galaxies with redshifts 0 . < ∼ z gal < ∼ . . < ∼ log( M (cid:63) /M (cid:12) ) < ∼ .
5. Large oxygencolumn densities N
OVI ≈ . ± . cm − suggest largeoxygen masses.Detailed observations of circumgalactic absorption byO +5 and other ions of C, N, O, Mg, Si, and Feare described in a series of publications: Werk et al.(2012,2013,2014,2016), Johnson, Chen, and Mulchaey(2015 ≡ JCM), Borthakur et al. (2016). Broad OVI ab-sorption is detected in essentially all sightlines near star-forming galaxies, but in only a small fraction of sightlinesnear passive, early-type L (cid:63) galaxies. Of interest here aredecreasing O VI absorption columns that extend beyondthe ∼
150 kpc survey limit of Tumlinson et al. to at least ∼
300 kpc around L (cid:63) galaxies (Prochaska et al. 2011;JCM).Meanwhile, a significant computational effort is under-way by cosmological simulators to understand the originof oxygen in the hot circumgalactic gas: Stinson et al.(2013); Suresh, et al. (2015); Liang, Kravtsov & Agertz(2016); Roca-Fabrega et al. (2016); Oppenheimer, etal. (2016); Sokolowska et al. (2016) etc. In these stud-ies most of the circumgalactic oxygen is provided bysupernova-driven galactic winds from the central galaxy. University of California Observatories/Lick Observatory, De-partment of Astronomy and Astrophysics, University of Califor-nia, Santa Cruz, CA 95064 ([email protected]).
Nevertheless, many of these sophisticated computationalstudies underpredict observed OVI column densities.Recent theoretical treatments of O VI in the circum-galactic medium include the phenomenological model ofStern et al. (2016) which envisions O VI as low-density,externally photoionized gas surrounding denser ionizedgaseous clumps that produce lower ionization absorptionlines of C, N, O, Mg, Si, and Fe. This scenario explicitlylinks the kinematics of the O VI gas with the lower ion-ization material (as suggested by observations of Werk etal. 2016). Most recently, Faerman et al. (2016) introducea model for galaxy coronae to reproduce O VI observa-tions together with X-ray observations of higher ioniza-tion states of oxygen in the Milky Way. In their phe-nomenological model, circumgalactic gas is multi-phasewith a distribution of densities all in pressure equilib-rium.As with lower ionization absorption lines, some OVIabsorption may occur in small, actively cooling regions.However, as seen in Figure 3 of Werk et al. (2013), broadvelocity widths of OVI absorption lines extend to veloc-ities unassociated with low ionization absorption. Thissuggests that most OVI absorption arises not in ther-mally perturbed regions but in extended hydrostatic at-mospheres – we adopt this assumption here.In what follows we pursue the astrophysical impli-cations of the observed OVI column density profile N OVI ( R ), assuming that the oxygen has been expelledfrom the central star-forming galaxies and now residesin circumgalactic atmospheres in collisional ionizationequilibrium. Specifically, we create simple hot gas at-mospheres around a fiducial L (cid:63) galaxy with a stellarmass M (cid:63) = 10 . M (cid:12) near the center of the sampleobserved by Tumlinson et al. (2011), having a dark halomass M h = 10 . M (cid:12) . We begin with a reference at-mosphere devoid of feedback distortion in which the gashas an NFW profile reduced by the cosmic baryon frac-tion. From this we construct several approximate at-mospheres distorted by increasing amounts of feedbackheating which must be maintained by continued feed- a r X i v : . [ a s t r o - ph . GA ] A ug Mathews & Prochaska
Fig. 1.—
Four atmospheres in the L (cid:63) potential. Black: no-feedback atmosphere with reduced NFW profile,
Blue, Red and Green: threeatmospheres with increasing feedback heating and distortion. (a)
Temperature profiles. Radii at 1, 2 and 3 r marked at upper right. (b) gas density profiles. Orange line : L (cid:63) n e ( r ) profile for 50 < r <
240 kpc from Oppenheimer et al. (2016). (c) dark halo ( dashed line ) andcumulative gas mass profiles in M (cid:12) . Black squares are Milky Way density and temperature at r kpc = 10 from Miller & Bregman (2016). back that balances radiation losses. The radial oxygenabundance profile in each atmosphere is adjusted untilthe space density profile of O +5 ions is consistent withthe observed mean column density profile N OVI ( R ). Weshow that the local O abundance can significantly exceedsolar at ∼
100 kpc with large total oxygen masses > ∼ M (cid:12) . Radiative cooling time profiles are almost identicalfor all model atmospheres, but their radiative luminosi-ties require feedback energies in excess of that expectedfrom supernova feedback alone. This provides strong ev-idence that central black holes are the dominant sourceof feedback energy having black hole accretion rates andmasses that match those expected in L (cid:63) galaxies. HYDROSTATIC CIRCUMGALACTIC ATMOSPHERES
Following Oppenheimer et al. (2016), we adopt arepresentative L (cid:63) central galaxy having stellar mass M (cid:63) = 10 . M (cid:12) and dark halo mass M h = 10 . M (cid:12) .The NFW halo that confines the hot gas is assumed tohave concentration c = 4 . M / M (cid:12) ) − . = 7 . r = 240 kpcwhere the density is 1/200 of the critical density. Weexpect that dark halos in L (cid:63) galaxies are no longer ac-creting dark (or baryonic) matter and may have beenquiescent for several Gyrs (Prada et al. 2006; Cuestaet al. 2008; Diemer & Kravtsov 2014). Consequently,we adopt an NFW density profile for the dark matterhalo, and assume that it has not changed substantiallysince redshifts z ∼ . dP/dr = − ρg NF W can be written dT /dr = − ( T /r )( d log ρ/d log r ) − g NFW ( µm p /k ) (1)where P = ( k/µm p ) ρT , m p is the proton mass and µ = 0 .
61 is the molecular weight. In the absence offeedback, baryons also have an NFW density profile re-duced by the cosmic baryon fraction, f b = 0 .
16. Wedisregard the stellar mass of the central galaxy sinceit is only ∼
10 percent of the baryonic mass within thevirial radius. The gas density profile without feedbackis therefore ρ ( r ) = f b ρ /y (1 + y ) where y = c ( r/r ), ρ = (200 ρ c / c /f ( c ), f ( y ) = ln(1 + y ) − y/ (1 + y ),and ρ c = 9 . × − gm cm − is the critical density for Hubble constant H = 70 km s − . The halo massis M h = M = (4 π/ ρ c r and the gravitationalacceleration g NF W = GM h ( r ) /r .In the presence of feedback, atmospheric gas is heated,its central entropy is increased and the entire atmosphereis pushed outward. To mimic these feedback effects, weconsider density profiles described by ρ = f b ρ / ( y + y ) (2)where y is a parameter that increases with the influenceof feedback. The central entropy S = T /n / e has zeroslope similar to galaxy group profiles visible in X-rays(Pratt et al. 2010). Since feedback sources are cen-trally concentrated, these atmospheres are designed toapproach the feedback-free density profile at large radius, ρ ( y ) ≈ f b ρ /y . We therefore disregard any significantdensity change at large radius, either an enhancementas gas is pushed out by feedback or a gas deficiency iffeedback extends to the distant halo. In any case, in-termediate radii, ∼
100 kpc, are most relevant to ourconcerns here. Our adoption of equation (2) does notaffect conclusions discussed below.Solid black contours in Figure 1 show temperature,density, and gas mass profiles for a circumgalactic at-mosphere without feedback. Also in Figure 1 are profilesfor three atmospheres with increasing feedback:blue : red : green :: y = 1 : 2 : 4 (3)Atmospheric structural distortion is stabilized when thetime-averaged feedback power maintains atmosphericprofiles against radiation losses. If feedback drops belowthis maintenance level, large masses of gas can cool to-ward the central galaxy, resulting in an unphysical massaccumulation, similar to the overcooling problem encoun-tered in cosmological simulations. Structural and main-tenance feedback can be provided by the same physicalmechanisms.The green atmosphere in Figure 1 is designed to matcha hot gas density of 10 − cm − at radius ∼ O +5 COLUMN AND SPACE DENSITIES
Blue squares in Figure 2a are COS-Halos detectionsof OVI column densities N OVI in circumgalactic atmo-spheres of L (cid:63) galaxies (Tumlinson et al. 2011) at sight-line offsets R in kpc. While most columns are clusteredabout N OVI ∼ . cm − , there is evidence of decreas-ing columns beyond R ∼
100 kpc. This decrease is con-firmed by additional OVI observations by JCM, manyhaving larger R . In Fig. 2a we plot JCM data for star-forming L (cid:63) galaxies having impact offsets R >
75 kpc,and stellar masses within 10.1 < log( M (cid:63) /M (cid:12) ) < ∼
100 kpc. Such adecrease strongly supports the prevailing interpretationthat very large oxygen masses have been ejected from L (cid:63) galaxies and/or their progenitors. Decreasing N OVI ( R )profiles also constrain the atmospheric O/H abundanceand the feedback mechanism that maintains it.Although data are sparse, we fashion a magenta line inFigure 2 showing a likely column density profile for L (cid:63) galaxies. The magenta line is described by N OVI = N [1 + ( R/R ) p ] − ( R/R ) q (4)where N = 5 . × cm − , R = 40 kpc, q = − / R = 140 kpc and p = 3.The corresponding empirical space density of O +5 ions˜ n O +5 ( r ), can be found from an Abel integral inversion,˜n O +5 ( r ) = − π (cid:90) ∞ r dNdR dR ( R − r ) / which is plotted as a magenta line in Figure 2b. Theinner, dotted part of the ˜n ≡ ˜n O +5 ( r ) profile is uncon-strained by current N OVI data. For comparison, dashedlines in Figure 2b shows O +5 space density profiles n ( r )for each atmosphere in Fig. 1, assuming uniform solarabundance A O (cid:12) = 5 × − and employing ionic fractionsfor collisional ionization equilibrium (Gnat and Sternberg2007).The juxtaposition of n and ˜ n profiles in Figure 2ballows an instant determination of atmospheric O abun-dance profiles Z O ( r ) in solar units required to matchthe mean profile ˜ n ( r ) and, when projected, the adoptedmean column density profile N OVI ( R ) in Figure 2a. Forexample, the oxygen abundance in solar units for thered atmosphere is simply Z O,red ( r ) = ˜n ( r ) /n ,red ( r ) atevery radius.Oxygen abundance profiles derived in this manner foreach atmosphere are illustrated in Figure 3a. Maximumabundances are expected near r ∼
100 kpc where theslope dN OVI /dR changes. Oxygen abundances Z O arelarge, particularly in the high feedback, low density greenatmosphere. At r = 100 kpc, O abundances in the blue,red and green atmospheres are Z O =1.1, 2.6 and 8.5solar respectively. The cumulative oxygen mass shownin Figure 3b also increases with feedback in the blue, redand green atmospheres, totaling 0.84, 1.08 and 1.61 (inunits of 10 M (cid:12) ) respectively. In a closed box modelan M (cid:63) = 10 . M (cid:12) stellar mass can produce an oxygenmass of only ∼ M (cid:12) (Zahid et al. 2012). However,in the simulation of Oppenheimer et al. (2016) the total oxygen mass within r at zero redshift is 3 × M (cid:12) ,but their computed sightline column densities N OVI areabout 3 times lower than those observed and the spacedensity of O ions is not as concentrated near r ∼
100 kpcas we require here.In Figure 4a we show the remarkable similarity of cool-ing time profiles t cool ( r ) for all atmospheres. It is easyto show that this similarity follows from the relation Z O = ˜n /n when oxygen dominates the metallicity, Z ≈ Z O . Therefore, the cooling time profile holds forany L (cid:63) atmosphere, not just those proposed here.Large bolometric X-ray luminosity profiles in Figure4b, log L X ≈ . ± . − , attest to the power-ful cooling effect of ∼ M (cid:12) of oxygen extending to r ≈
100 kpc. To explore the destiny of pure cooling with-out feedback, we computed spherical time-dependentcooling flows allowing blue, red and green (b,r,g) atmo-spheres initially at rest to evolve for the lookback time2.4 Gyrs at redshift 0.2 with abundances from Fig. 3a.These straightforward calculations, not discussed in de-tail here, verify that gas masses of log(
M/M (cid:12) ) ≈ (11 . M ∗ , cool at the origin in(b,r,g) atmospheres during this relatively short time. Alloxygen-rich gas within ≈
210 kpc cools. Such an atmo-spheric collapse is not supported by the recent star for-mation history of the Milky Way (Gonzalez et al. 2017)or by observations of extended circumgalactic OVI ab-sorption in star-forming L (cid:63) galaxies. Clearly, a strongtime-averaged maintenance feedback power comparableto L X must be provided by supernovae or central blackholes. Our concern here is not the energetics of the out-flows that previously carried O-rich gas out to ∼
100 kpc,but how this gas is maintained during more recent times.In the figures we assume that oxygen is completelymixed on atomic scales, but the degree of mixing re-mains uncertain. It is likely that circumgalactic oxygenis inhomogeneous, which would decrease the radiativecooling time below that in Figure 4a, causing small, O-rich regions to cool locally, perhaps even in the pres-ence of maintenance feedback. Such local cooling is sup-ported by low-ionization circumgalactic absorption linesin quasar spectra.Core-collapse supernovae in late type L (cid:63) galaxies occurat a rate of 1-3 per century, but generate only ∼ . × erg s − , most of which is dissipated locally in strongshocks in the galactic disk. Successful L (cid:63) feedback re-quires intermittent shocks propagating through gas at r ∼
100 kpc at some mean interval ∆ t . The rate thatentropy is radiated away by gas at r = 100 kpc (where n ≈ − cm − , T ≈ K and Z ≈ Z (cid:12) ) can be restoredby shocks every log ∆ t = (7, 8, 9) yrs with large Machnumbers (1.25, 1.7, 3.9), roughly similar to feedbackshock strengths in galaxy group atmospheres. Even ifsupernova power were sufficient, neither stochastic vari-ations in the galactic supernova rate nor slowly mov-ing (forward and reverse) shocks expected in sustainedsupernova-driven winds are sufficiently intermittent.Consequently, massive central black holes may be themost promising source of distant, intermittent feedbackshocks. Following King and Pounds (2015), fast-windfeedback from AGNs having luminous accretion disksgenerate mechanical luminosities ∼ . L E where L E =1 . × ( M • /M (cid:12) ) erg s − is the Eddington luminosity. Mathews & Prochaska Fig. 2.— (a) N OVI observations. Bluesquares from COS-Halos, green squares andred upper limits from JCM. (b) Dashed lines: n O +5 ( r ) profiles for the four atmospheres,all with solar abundance. Magenta line: ˜n O +5 ( r ) profile. Fig. 3.— (a)
Oxygen abundance profilesin solar units required for each atmosphere;
Dotted line marks solar abundance. (b)
Cu-mulative oxygen mass profiles.
Fig. 4.— (a)
Short cooling time profiles foreach atmosphere. (b)
Cumulative bolometricluminosity profiles of each atmosphere, emit-ted primarily at energies < . Since central black holes in L (cid:63) galaxies with M (cid:63) = 10 . M (cid:12) have masses M • similar to that in the Milky Way,4 . × M (cid:12) , the expected mechanical power of L (cid:63) windsis ∼ × ∼ L X erg s − . This is sufficient to bal-ance X-ray luminosities in Figure 4b with central AGNsthat are intermittently active only 3 percent of the time.This duty cycle is < ∼
10 percent as required by Oppen-heimer et al. (2017) to enhance OVI-absorbing ions inatmospheres of L (cid:63) galaxies.More collimated feedback modes may also be present.In galaxy groups, having black holes and dark halomasses roughly ten times larger than in L (cid:63) galaxies, theaccretion power of central black holes maintain atmo-spheric temperatures T ∼ K with no evidence ofrecent star formation or luminous accretion disks. In ad-dition, the Fermi bubbles in the Milky Way reveal thatits central black hole is currently delivering an (albeituncertain) energy of ∼ ergs to the circum-Galacticatmosphere (Guo & Mathews 2012, Guo et al. 2012),equivalent to the total Galactic supernova energy gen-erated in 10 yrs. Since outward propagaing shocks inatmospheres with d log ρ/d log r > − L (cid:63) galax-ies (Tumlinson et al. 2011). Since central black holesin passive L (cid:63) galaxies are on average ∼
40 times moremassive than those in star-forming L (cid:63) galaxies (Reines &Volonteri 2015), the accretion feedback in passive galax-ies may drive most of the OVI-absorbing circumgalacticgas entirely out of the galactic potential. Upper limitsof four star-forming L (cid:63) galaxies in Figure 2a also lie farbelow the magenta profile. Perhaps unusually large feed-back events with energy > ∼ ( GM g M h /r ) | kpc ∼ . ergs have removed the entire atmosphere.We acknowledge a very helpful observational communi-cation from Todd Tripp and useful remarks from FabrizioBrighenti. WGM thanks George Poultsides, John Har-ris, Joseph Palascak and Francisco Rhein for sustainingsupport. REFERENCES
Borthakur, S., Heckman, T., Tumlinson, J., et al. 2015, ApJ, 813,46Cuesta, A. J., Prada, F., Klypin, A., Moles, M. 2008 MNRAS 389385Diemer, B., Kravtsov, A. V., 2014 ApJ 789, 1Faerman,Y., Sternberg,A., McKee, C. F. 2016, (arXiv160200689F)Fang, T.; Bullock, J.; Boylan-Kolchin, M. 2013, ApJ 762, 20Gnat, O. & Ferland, G. J. 2012, ApJ 199, 20 Gnat, O. & Sternberg, A. 2007 ApJS 168, 213Gonzalez Delgado, R. M. et al. 2017, arXiv170606119Guo, F., Mathews, W. G., Dobler, G., Oh, S. P. 2012, ApJ 756,182Guo, F., Mathews, W. G. 2012, ApJ, 756, 181Johnson, S. D., Chen, H.-W., & Mulchaey, J. S. 2015, MNRAS,449, 3263King, A. R; Pounds, K. A. 2015, ARA&A, 53, 115. ircumgalactic Oxygen Absorption and Feedback 5ircumgalactic Oxygen Absorption and Feedback 5