WIXLIB

Ripple effect in broad FeKα linesfrom a point in the disk to the observed energy is given by(Fabian et al. 1989) 1/2 6Mcosβ(1 z) 1 (1)1 r[r(1 tan2 ξ)/M 1]1/2where (r, φ) is the location of the emitting point in the disk(with r in units of rg GM/c2 , the gravitational radius). Ineqn. (1) M is the black hole mass, β is the angle between thedisk plane and the plane of the photon trajectory, ξ π/2is the angle between the emission of the photon and theline connecting the emitting point to the black hole. Thequantity cosβ may be written as (Fabian et al. 1989)cosβ cosφsinθ[cos2 θ cos2 φsin2 θ]1/2(2)with θ the angle between the observer and the disk. Thespecific intensity in the frame of the emitting plasma is approximated as a delta function Iem ǫδ(E E0 ) where E0is the line rest energy. The structure of both the hot corona(source of the X-ray continuum) and the inner disk (sourceof the fluorescing Fe) is unknown, so ǫ the disk emissivityis simply assumed to be (r/M ) k . Therefore the specificflux measured by the observer from portion rdrdφ of thedisk is then given by (Fabian et al. 1989)dFobs 1 ΩIemrdrdφ(1 z)3 (r, φ)(3)where the solid angle dΩ D 2 bdbdφ′ where D is the distance to the disk from the observer, b is the photon impactparameter and φ′cosθ. φcos2 φ sin2 φcos2 θ(4)In the weak field limit, the photon geodesics can be approximated with straight lines, which means that not allrelativistic effects are fully calculated at 20rg in this approach (Beckwith & Done 2004). By integrating eqn. (3) asZ 2π Z r1Z 2π Z routerFobs dFobs(5)dFobs 0rinner0r2where the disk extends from rinner to router , with a gap excavated from radii r1 to r2 , we obtain the observed broad lineflux from a disk containing a gap. In the limit as r1 r2 ,and if the disk extends to the innermost stable circular orbits (ISCO), rinner rISCO , we recover the observed broadline flux from the total disk. In the limit as rinner r1 werecover the observed broad line flux due to the inner diskminus a cavity spanning rISCO to r2 .2.2Gap-opening in disks and merging black holesThe disk emissivity which produces the broad Fe Kα linecomponent depends on both the structure of the X-ray continuum emitting corona plus the gas surface density distribution in the circumbinary disk. The geometry of both thecorona and inner disk are poorly understood, particularlyat the late stages of a merger with the central supermassiveblack hole. The main purpose of this paper is to illustratethat different possible gas configurations produce very different broad FeKα line profiles. Although our illustrationsare based on toy models, the main features should be robust,and potentially distinguishable in observations with Astro-Hc 2013 RAS, MNRAS 000, 1–183and IXO/Athena. In this section, we outline possible models for the circumbinary gas, including those with a narrowannular gap, a fully empty central cavity, or a central cavitywith possible minidisks inside the cavity. Throughout thisdiscussion, we emphasize the very large existing theoreticaluncertainties.BH binaries are commonly believed to be surrounded by a thin (H/R . 1) circumbinary disk.In models where the binary results from a galaxygalaxy merger, the merger is expected to deliver thenuclear BHs (e.g. Springel, Di Matteo & Hernquist 2005;Robertson et al. 2006), along with copious amounts of gas(Barnes & Hernquist 1992), to the central regions of the newgalaxy. The gas cools efficiently, and loses angular momentum, creating a thin sub-parsec disk in which the pair of BHsare embedded (e.g. Dotti et al. 2009; Hopkins & Quataert2010; Chapon, Mayer & Teyssier 2011). IMBHs with lowermasses can also grow quickly in situ in the disk around asingle supermassive black hole, again producing a pair (ormore) of BHs embedded in a thin disk (McKernan et al.2012).A BH binary can exchange angular momentum withthe disk, distorting the disk’s density profile, causing thesecondary BH to migrate inward. In the limit of a very lowmass secondary (q M2 /M1 1), the distortions, in theform of spiral density waves, remain in the linear regime(and the corresponding Type I migration is very rapid; e.g.Goldreich & Tremaine 1980). In this limit, the FeKα linewould be hardly affected by the presence of the low-masssecondary 2 . As the mass of the secondary increases, gas isgravitationally torqued away from the orbit, lowering theco-rotating gas density and a particularly rapid form of migration (Type III) may occur (Masset & Papaloizou 2003;McKernan et al. 2011b). Once the secondary becomes sufficiently massive it will open an empty annular gap in the disk,analogous to gaps known to exist in protoplanetary disks.This happens above the critical mass ratio (e.g. Armitage2010) of 1/2 5/227πHq α1/2(6)8rwhere H/r is the geometric thickness (aspect ratio)of the disk and α is the disk viscosity parameter(Shakura & Sunyaev 1973). A black hole with q 10 4should be able to open a gap in a fiducial AGN disk. Thismass range spans massive stars or black holes of 102 M inaccretion disks around 106 M supermassive black holes(McKernan et al. 2012), all the way up to supermassiveblack hole binaries that form long after a major galacticmerger has occurred. The radial extent of the gap would beof order a few Hill radii ( RH ). In this case, the secondarycan be regarded as a particle in the disk, migrating inwardon the viscous time-scale (so-called Type II migration).When the secondary arrives at the radius where thelocal disk mass is too small to absorb the secondary’s angular momentum, its migration must stall (or at least slow2However, if the local gas mass is low enough, or for low enoughvalues of H/r, α in the innermost disk, even low-mass migratorsexpected in the AGN disk (McKernan et al. 2011a, 2012) couldopen a gap or cavity and affect the Fe Kα line as described here.

4B. McKernan, K.E.S. Ford, B. Kocsis & Z.Haimandown). An important point is that “standard” thin diskshave relatively low mass, and this stalling occurs at fairlylarge binary separations. With the exception of very unequalmasses q . 0.01, the transition takes place well in the outerregion of the disk, at r several 103 rg . Even for q 1binaries, the transition is located atrstall 260(M1 /107 M ) 2/7 (q/10 4 )2/7(7)Schwarzschild radii, which may lie outside the region of interest for producing broad FeKα lines depending on router(the above equation assumes standard disk parameters;see eqn. 25 in Haiman et al. 2009 for the full expression).Once inside this stalling radius, the secondary is no longerable to maintain steady Type II migration on the viscous time-scale, and the evolution of the binary disk system from this point onward becomes very poorly known.The inner disk - whose viscous time is short - is believed to drain onto the primary, creating a central cavity. Continued accretion from larger radii causes a pileup of gas just outside the secondary’s orbit, analogous toa dam in a river (Syer & Clarke 1995; Ivanov et al. 1999;Milosavljevic & Phinney 2005; Chang et al. 2010; Rafikov2012). If the dam is 100% efficient in continuing to blockthe gas arriving from large radii, then the binary, at smallseparations, will be surrounded by an empty inner cavity, devoid of any gas. Such inner cavities have been seen in manynumerical simulations (starting with the seminal work byArtymowicz & Lubow (1994)). However, simulations do notfollow the system on long enough timescales to determinewhether such a cavity could persist for many viscous times.In the face of ongoing accretion, an initial cavity may begradually refilled (if the “dam” is porous), or else the gasaccumulated outside the orbit may eventually cause the damto “burst” (Kocsis, Haiman & Loeb 2012a,b).An empty cavity would of course strongly modify theshape of the FeKα line. However, this picture is oversimplified. Many numerical simulations imply that gascan enter such a cavity in narrow collimated streams(Artymowicz & Lubow 1996; Hayasaki, Mineshige & Sudou2007; MacFadyen & Milosavljevic 2008; Cuadra et al. 2009;Roedig et al. 2011, 2012; Shi et al. 2011). Such streams mayfeed a “minidisk” around the secondary and perhaps alsothe primary BH (Hayasaki et al. 2008; Sesana et al. 2012;Farris et al. 2012). A minidisk around the secondary may befavored, but the details depend on the angular momentumand shocks in the material in the streams. The streamers andminidisks could create some additional FeKα emission. Forexample, if both disks emit, there would be a double FeKαline (Sesana et al. (2012)). On the other hand, it is unclearwhether such streams and minidisks may persist down tosmall binary separations. At separations of . 200rg q 2 (seeeqn. 30 in Haiman et al. 2009 for the full expression, including dependencies on disk parameters), the binary BH emitsGWs efficiently and is driven to merger by this GW emission. It effectively decouples from the disk (Liu, Wu & Cao(2003); Milosavljevic & Phinney (2005)) and “runs away”.The merger timescale is given by (Peters 1964) 3 6 2 410 M 10 M aτGW 1012 yr(1 e2 )7/2 (8)M2M10.001pcwhere M2 is the mass of the secondary, M1 is the mass ofthe primary, a is the binary separation and e is the sec-ondary orbital eccentricity. For a secondary on a circularized orbit with q 3 10 3 located at a (10)100rg inan AGN disk, τGW (30yr)0.3Myr around a 108 M SMBHand τGW (1yr)0.01Myr around a 106 M SMBH respectively. Simulations in the relativistic regime found that gasstreams can follow the binary to small radii ( few Rs ; e.g.Noble et al. 2012; Farris et al. 2011, 2012), but these simulations start with small binary separations, with gas assumedto have followed the BH to these initial separations.As emphasized recently by Kocsis, Haiman & Loeb(2012a,b), even apart from possible streamers and minidisks inside the central cavity, the formation of the cavity itself, at large radii, is still uncertain. The coupledtime-dependent processes of the secondary’s migration, inner cavity formation, and damming-up of gas outside thesecondary’s orbit, has not been modeled self-consistently,even in one-dimensional calculations. However, using selfconsistent steady-state solutions, Kocsis, Haiman & Loeb(2012b) have argued that in many cases, pile-up can causea dam overflow at large binary separations. In this case,when the binary arrives at small radii relevant to the FeKαline, the gas configuration would presumably resemble afilled disk, with an annular gap (of order the Hill radius).Although the above is yet to be demonstrated by a selfconsistent calculation, it is at least suggested by recent simulations of Baruteau et al. (2012). This is the only existingsimulation that looks at small binary separations, in theGW-driven regime, and does not manually insert an emptycavity in the beginning. They show that the secondary movesinward, “ice-skating” across the inner disk (with the innerdisk material continually crossing the secondary’s orbit outward, on horse-shoe orbits). This simulation is in 2D, andassumes a constant kinematic viscosity physics, which maynot be accurate. Nevertheless, it suggests that if the cavityfilled in at large radius, it will remain filled until the end,except for a narrow annular gap around the secondary.We are motivated by the above discussion to considerthe impact on the Fe Kα line of four different circumbinarygas disk geometries, all variations on a standard single-BHaccretion disk: (i) a narrow annular gap around the secondary, (ii) an empty central cavity, (iii) a central cavitysurrounded by an overdense ring (’dam’), and (iv) a centralcavity with a small, circumsecondary accretion disk (’minidisk’) inside it. We regard (i) as our fiducial case, addressingit first in §2.3 and §3. We then examine the alternative scenarios (ii)-(iv) in §4.2.3A ripple effect in the broad FeKα linecomponentAs a gap-opening black hole migrates inward in an AGNdisk, the empty gap spans parts of the disk that produceever bluer and redder parts of the broad Fe Kα line. Therefore we expect flux ’notches’ to be removed from the broadline profile. As the annulus moves inward on the disk viscous timescale, we expect the notches to ripple red-wardand blue-ward respectively from the line centroid. Fig. 1(from McKernan et al. (2013)) demonstrates this ripple effect for a gap-opening black hole located at circularizedorbit radius R centered in an empty gap of width 2RHwhere RH (q/3)1/3 R. The black hole migrates inwardwith the gap from 100rg to 6rg . We assumed q 3 10 3c 2013 RAS, MNRAS 000, 1–18

00.2Flux0.4Ripple effect in broad FeKα lines45678EnergyFigure 1. From McKernan et al. (2013), showing Iron line fluxin arbitrary units versus Energy in units of keV,binned at approximately the energy resolution ( 7eV) expected for Astro-H.The curves demonstrate the progression of the ’ripple effect’ inthe Fe Kα line profile given the in-migration of a gap-openingblack hole (mass ratio q 3 10 3 ) located at R in the innerdisk. The empty gap has width 2RH where RH (q/3)1/3 R. Thedisk has an assumed X-ray emissivity radial profile of r 2.5 andthe angle to the observer’s sightline is θ 60o (where θ 0o isface-on). The solid black curve indicates the unperturbed Fe Kαline profile (i.e. no gap). The turquoise solid curve corresponds tothe same Fe Kα line but with an annulus at 90 9rg in the innerdisk. Dark blue solid curve denotes an annulus at 50 5rg , Greensolid curve corresponds to an annulus at 20 2rg and the dashedred curve corresponds to an annulus at 10 1rg .for ease of calculation, the disk is oriented at 60o to theobserver’s line of sight, the X-ray emissivity profile goesas r 2.5 and the broad line flux is assumed to come from 100rg in the disk (i.e. rinner 6rg , router 100rg ). If wechange router to larger radii (e.g. Niedźwiecki & Zycki 2008;Niedźwiecki & Miyakawa 2010), the broad line profiles depicted throughout this paper would appear more centrallypeaked with ’horns’ closer together, resulting in a slight decrease in the prominence of the notches and oscillationsdescribed here. If router is larger, we become sensitive tostalling radii, gaps and cavities further out in the accretion disk. Exactly how sensitive to detection of the effectsdiscussed here depends on the line emissivity ( r k ) profile. If the emissivity profile is steep (k 3) then most lineflux originates close to the BH and we have low sensitivityto features in the disk much further than 100rg . If theemissivity profile is shallower (k 2), for large router , theline ’horns’ correspond to material around 100 150rg ,so gaps at these radii could be detected. In general, oursimple assumption that the broad line flux comes from. 102 rg ), is reasonably consistent with fits to many observedbroad line components (Reynolds & Nowak 2003). Likewise,in the following discussion we do not correctly treat generalrelativistic effects such as light-bending, blurring and BHc 2013 RAS, MNRAS 000, 1–185spin (e.g. Dovčiak et al. 2004; Brenneman & Reynolds 2009;Dauser et al. 2010 2010). These effects, if properly treated,will tend to smear out sharper notches in the red wing ofthe broad FeKα line and modify line flux originating from. 20rg . Even though we ignore the many possible confounding effects (e.g. continuum absorption, the location and variation of the continuum source, broken axi-symmetry), weexpect that the kinds of spectral features here are broadlyrobust and allow us to take a first step in characterizing different scenarios of merging massive black holes in the inneraccretion disk.From Fig. 1, the progression of the ripple effect is quiteclear. As the satellite black hole and gap migrate into theinner 100rg of the disk, a flux deficit appears (turquoisesolid curve) compared to the unperturbed profile (black solidcurve). As migration continues, matching notches are removed from the broad red wing and the blue wing of theprofile, with the deepest notches in the blue wing. The twinnotches move apart, one red-ward, the other blue-ward further from the line centroid (6.40keV) in a distinctive monotonic progression. As the gap moves inward in the disk from50rg (dark blue curve) to 20rg (green curve) to 10rg (redcurve), the blue notch moves from 7.1keV to 7.6keV in theblue wing, and the red notch moves from 5.6keV to 4.2keVin the broad red wing. For a 3000M gap-opening IMBHaround a 106 M supermassive black hole, the progressionfrom green to red curve in Fig. 1 via GW emission takes(eqn. 8) 16yrs and from red curve to final merger takes 1yr. Because of the shorter timescales, low mass AGNexhibiting broad Fe Kα line profiles (e.g. MCG-6-30-15) arethe prime targets for observations searching for this rippleeffect, and the massive mergers that cause it.AGN that display this ripple effect in the broad Fe Kαline will be some of the most luminous sources of gravitational radiation in the Universe. By searching for the latterstages of a progression as in Fig. 1, we can follow the final stages of massive mergers electromagnetically. The finalstages of a ripple effect in the broad FeKα line will provideadvance warning of an outburst of gravitational waves fromthis AGN. Conversely, once gravitational waves consistentwith an IMBH-supermassive black hole merger are actuallydetected, an archival search for the FeKα ripple effect canbe used to localize the source.We expect departures from the predicted progression inFig. 1 at very small radii either as the disk profile changesor as the gravitational radiation luminosity increases dramatically and perturbs the disk (α and H/r may changeconsiderably). So the behaviour of the broad component ofthe Fe Kα line may be complicated in the final stages ofthe merger (in the limit of small disk radii). Variations fromthe predicted profile at small radii will include the effectsof gravitational radiation on the accretion disk. So tracking the ripple effect may allow us to directly test modelsof massive mergers and gravitational radiation, without actually detecting gravitational radiation. The Fe Kα bandcan also be spectrally complicated by emission lines. Bymonitoring the monotonic change in energy of the blue andred dips, we can break the degeneracy associated with highenergy Fe emission lines superimposed on the FeKα complex (e.g. McKernan & Yaqoob 2004). Curious ’notch’-likefeatures have been observed in broad Fe Kα line components (e.g. Yaqoob & Serlemitsos 2005), but the collecting

6B. McKernan, K.E.S. Ford, B. Kocsis & Z.Haimanarea and the energy resolution of the high energy transmission gratings on board Chandra are insufficient to resolve these features and follow them over time. The broadcomponent of the Fe Kα line can vary quite significantlyover time (e.g. Iwasawa et al. 1996; deMarco et al. 2009;Svoboda et al. 2012), but it has been difficult to determinethe nature of the variability given detector limitations.The energy resolution of the micro calorimeter plannedfor the Astro-H mission will be 7eV in the FeKα band(Takahashi et al. 2010). This exceptional resolution combined with moderate collecting area should be sufficient todistinguish between absorption features and narrow, monotonically migrating notches of the kind depicted in Fig. 1in long, repeated observations of nearby, X-ray bright AGN(e.g. Nandra et al. 1997). However, it is important to notethat Fig. 1 displays only the predicted broad line flux. Tosystematically obtain broad line spectra that can distinguishbetween the different cases in Fig. 1 requires a minimumtelescope effective area of that planned for LOFT ( 10m2at 6.4keV), together with the planned spectral resolutionof ( 5 7eV at 6.4keV) for Astro-H and IXO/A

disk, which shows up in the broad Fe Kα line component. If the secondary stalls at 102r g in its in-migration, due to low co-rotating gas mass, a detectable ripple effect occurs in the broad line component on the disk viscous timescale as the inner disk drains and the outer disk is dammed. If the secondary maintains an accretion disk