Jonelle Walsh (Texas A&M) TMT Science Forum Dec 10, 2018 Introduction: Black Holes Are Everywhere

Exploring the Supermassive Black Hole - Galaxy Connection

Jonelle Walsh (Texas A&M) TMT Science Forum Dec 10, 2018 Introduction: Black Holes are Everywhere

Types of Black Holes ✦ Supermassive BHs reside in essentially every massive galaxy. Supermassive 1 billion

1 billion 1 ✦ The strongest evidence we have for a BH Black Holes comes from the Milky Way (e.g., Genzel et al. 2010, Boehle et al. 2016).

Supermassive Black Holes Black Supermassive 1 million (This image 1 million 1 was created by Prof. Andrea Ghez and her research team at UCLA and are from

Types of Black Holes Black of Types data sets obtained

1000 with the W. 1000 black hole mass (relative to Sun) to (relative mass hole black M. Keck Telescopes.)

100100

1010 Stellar-mass Black Holes Black Stellar-mass

Stellar-mass 1 1 ✦ Beyond the Milky Way, BHs have been Black Holes black hole mass dynamically detected in ~100 galaxies (e.g., (M⊙) Saglia et al. 2016). Dynamical Searches for Black Holes

✦ Precise MBH measurements require high angular (HST) (credit: ESA/Hubble) resolution observations.

✦ Observations need to probe region over which the BH potential dominates — the BH sphere of inﬂuence (rsphere).

✦ Typical values for rsphere are small, so we are limited to studying nearby (~100 Mpc) objects.

✦ HST has played a fundamental role in detecting BHs over the past two decades. Dynamical Searches for Black Holes

✦ Signiﬁcant progress has recently been made using large ground-based telescopes + AO (e.g., Mazzalay et (Keck) al. 2016, Erwin et al. 2018, Krajnović et al. 2018).

✦ ALMA also provides superb sensitivity and angular resolution high enough to directly detect molecular gas within rsphere (e.g., Barth et al. 2016, Onishi et al. 2017, Davis et al. 2017).

(ALMA) (credit: Stephane Guisard/ESO)

(credit: Laurie Hatch Photography http://www.lauriehatch.com/) Black Hole Mass Measurement Methods

STELLAR DYNAMICS IONIZED GAS DISKS MASER DISKS

✦ Widely applicable, BUT●●● ✦ Conceptually simple, ✦ Produce some of the most BUT●●● precise MBH’s, BUT●●● ✦ Models are complex. ✦ Assumption of circular ✦ Disks are rare — they ✦ Models can be biased due rotation must be veriﬁed. require special physical to a number of systematic conditions and nearly edge- effects. ✦ Often the observed on inclinations. velocity dispersion is larger ‣ MBH - M/L - DM than that predicted from ✦ There can be signiﬁcant degeneracies models. gravitational contributions ‣ intrinsic shape/ from disks themselves that orientation effects ‣ what is the physical lead to departures from origin and what does Keplerian rotation. that mean for MBH?

(e.g., Gebhardt & Thomas 2009, van den (e.g., Barth et al. 2001, Verdoes Kleijn et (e.g., Miyoshi et al. 1995, Lodato & Bertin Bosch & de Zeeuw 2010, Rusli et al. 2013, al. 2006, Walsh et al. 2010) 2003, Pesce et al. 2015) McConnell et al. 2013) Black Hole Mass Measurement Methods

STELLAR DYNAMICS IONIZED GAS DISKS MASER DISKS

6 6

A Stellar-Dynamical Black Hole Mass Measurement Example

✦ Stellar-dynamical measurements require high-angular resolution spectroscopy, high-

angular resolutionFIG imaging,. 2.— Results and of stellar-dynamical large-scale models spectroscopy. fit to the NIFS kinematics. The χ2 is shown with black hole mass (left panel) and V-band mass-to-light ratio (middle panel). Each black dot corresponds to a single model, with the best-fit model highlighted as the red square. The red solid line indicates where ∆χ2 =9.0, corresponding to the 3σ confidence level for one parameter. χ2 contours are shown for a grid of black hole masses and mass-to-light ratios (right panel). Each 9 2 gray point represents a model, and the best-fit model (red square) corresponds to MBH =4.9 × 10 M⊙ and ΥV =9.3 Υ⊙. Relative to the minimum, the χ 2 hasFIG.F increased 2.—IG. 2.— Results by Results 2.3 of (thick stellar-dynamical of✦ stellar-dynamical black contour), models models 6.2 fit (thin to fit the black to NIFSthe contour), NIFS kinemat kinemat andics. 11.8ics. The (red Theχ2 contour),isχ shownis shown with corresponding with black black hole tohole mass the mass 1 (leftσ,2 (left panel)σ,and3 panel) andσ andconfidenceV-bandV-band mass-to-light intervals mass-to-light for ratio two ratio NGC 1277 Obtained Gemini/NIFS observations assisted by LGS ∆ 2 (middleparameters.(middle panel). panel). Each Each black black dot corresponds dot corresponds to a to single a single model, model, with w theith best-fit the best-fit model model highlighted highlighted as the as red the square. red square. The The red so redlid so linelid line indicates indicates where where∆χ2 =9χ .0,=9.0, 2 correspondingcorresponding to the to 3 theσ confidence 3σ confidenceAO. level The level for one forIFU one parameter. parameter.dataχ2 probescontoursχ contours are shown arethe shown for central a for grid a grid of black of ~1” black hole hole masses(340 masses and pc) and mass-to-l mass-to-l regionightight ratios ratios (right (right panel). panel). Each Each ×9 9 Υ Υ 2 2 graygray point point represents represents a model, a model, and and the best-fitthe best-fit model model (red (red squar square) correspondse) corresponds to M toBHM=4BH .=49 ×.910 10M⊙Mand⊙ andΥV =9V .=93 Υ.3⊙. Relative⊙. Relative to the to minimum,the minimum, the χ the χ hasFhas increasedIG. increased 2.— Results by 2.3 by (thick2.3 of stellar-dynamical (thick blackwith black contour), contour), a modelsPSF 6.2 (thin6.2 fit~0.15” (thin to black the black c NIFSontour), c.ontour), kinemat and and 11.8ics. 11.8 The (red (redχ contour),2 is contour), shown corresponding with corresponding black hole to the to mass 1theσ,2 (left 1σσ,2,and3 panel)σ,and3σ andconfidenceσVconfidence-band mass-to-light intervals intervals for two for ratio two parameters.(middleparameters. panel). Each black dot corresponds to a single model, with the best-fit model highlighted as the red square. The red solid line indicates where ∆χ2 =9.0, corresponding to the 3σ confidence level for one parameter. χ2 contours are shown for a grid of black hole masses and mass-to-light ratios (right panel). Each ✦ 9 2 gray point represents a model,Measured and the best-fit modelthe (redline-of-sight square) corresponds velocity to MBH =4 .9distribution× 10 M⊙ and ΥV =9(LOSVD).3 Υ⊙. Relative of to the minimum, the χ has increased by 2.3 (thick black contour), 6.2 (thin black contour), and 11.8 (red contour), corresponding to the 1σ,2σ,and3σ confidence intervals for two parameters. stars as a function of spatial location. LOSVD parameterized with V, σ, h 3, and h4.

(Walsh et al. 2016)

FIG. 3.— The observed NIFS kinematics (top) show that NGC 1277 is rotating, with the west side of the galaxy being blue-shifted,thatthereisasharprise in the velocity dispersion and generally positive h4 values at the nucleus, and that h3 and V are anti-correlated. The best-fit stellar dynamical model (bottom) is shown on the same scale given by the color bar to the right and the minimum/maximum values are provided at the top of the maps. The best-fit model nicely reproduces the NIFS observations and has a reduced χ2 of 0.7. FIG.F 3.—IG. 3.— The The observed observed NIFS NIFS kinematics kinematics (top) (top) show show that that NGC NGC 1277 1277 is rotating, is rotating, with with the west the west side side of the of galaxy the galaxy being being blue-shifted blue-shifted,thatthereisasharprise,thatthereisasharprise h h V in thein velocity the velocity dispersion dispersion and and generally generally positive positiveh4 values4 values at the at nucleus, the nucleus, and and that thath3 and3 andV are anti-correlated.are anti-correlated. The The best-fit best-fit stellar stellar dynamical dynamical model model (bottom) (bottom) is is shownitedshown mass on the on degeneracies samethe same scale scale given given with by the by the colorthe galaxy’s color bar barto the dark to rightthe halo right and (e.g.,and the minimum/maximumthe Ge minimum/maximumb- Shapiro values values et are al. providedare 2006; provided at Krajnovi the at topthe topofcetal.2009).TheNIFSkine-´ the of maps.the maps. The The best-fit best-fit model model nicely nicely 2 ′′ ∼ reproduceshardtreproduces & the Thomas NIFS the NIFS observations 2009; observations Rusli and and has et hasa al. reduced a 2013) reducedχ2 of andχ 0of.7. uncertainties 0.7. matics extend out to a radius of 1. 3(442pc),or 0.4 Re. associatedFIG. 3.— The with observed the assumption NIFS kinematics of a (top) constant show that stellar NGC mass-t 1277 iso- rotating, withTherefore, the west side the of NIFS the galaxy data being not blue-shifted only resolve,thatthereisasharprise the black hole inlight the velocity ratio (e.g., dispersion McConnell and generally et al. positive 2013).h values Moreover, at the nucleus, owing and to that hsphereand V are of anti-correlated. influence, but The also best-fit probe stellar the dynamical region where model (bo thettom) stars is itedited mass mass degeneracies degeneracies with with the the galaxy’s galaxy’s dark4 dark halo halo (e.g., (e.g., Geb- Geb- Shapiro3 Shapiro et al. et al. 2006; 2006; Krajnovi Krajnovicetal.2009).TheNIFSkine-´ cetal.2009).TheNIFSkine-´ shownthe presence on the same of the scale nuclear given by dust the color disk, bar uncertainty to the right and in the the minimum/maximum view- begin values to dominate are provided the at thepotential. top of the It maps.′′ is encouraging The best-fit model that nicely the hardt & Thomas 2009; Rusli et al. 2013)2 and uncertainties matics extend out to a radius of′′ 1. 3(442pc),or∼ ∼0.4 R . hardtreproducesing orientation & Thomas the NIFS is 2009;observations minimal, Rusli unlike and et has al. some a reduced 2013) pastχ and stellar-dynamof 0 uncertainties.7. ical maticsbest-fit extend model out constrained to a radius by of only 1. 3(442pc),or the NIFS kinematics0.4 R pre-e. e associated with the assumption of a constant stellar mass-to- Therefore, the NIFS data not only resolve the black hole associatedstudies (e.g., with Shapiro the assumption et al. 2006; of Onken a constant et al. stellar 2007). mass-t The duo-st Therefore,sented at the the beginning NIFS data of not Section only 8 resolve largely the reproduces black hole the lightlight ratio ratio (e.g., (e.g., McConnell McConnell et al. et al.2013). 2013). Moreover, Moreover, owing owing to to spheresphere of influence, of influence, but but also also probe probe the the region region where where the the stars stars iteddisk, mass however, degeneracies may be problematic with the galaxy’s for construction dark halo (e.g., of the Ge lub-- Shapiroobserved et HET al. 2006; kinematics Krajnovi overcetal.2009).TheNIFSkine-´ the radial range of the NIFS thethe presence presence of the of the nuclear nuclear dust dust disk, disk, uncertainty uncertainty in the in the view- view- beginbegin to dominate to dominate the the potential. potential. It is It encouraging is encouraging that that the the hardtminous & mass Thomas model 2009; in Section Rusli et 2. al. We 2013) conservatively and uncertainties masked maticsdata, as extend can be out seen to in a radius Figure of 4. 1 Nevertheless,′′3(442pc),or we∼ also0.4 ranR . inging orientation orientation is minimal, is minimal, unlike unlike some some past past stellar-dynam stellar-dynamicalical best-fitbest-fit model model constrained constrained by by only only the. the NIFS NIFS kinematics kinematics pre- pre-e associatedthe dust when with generating the assumption the MGE of a model, constant but stellar acknowledge mass-to- Therefore,dynamical models the NIFS that data fit to not the onlycombination resolve of the the black NIFS hole and studiesstudies (e.g., (e.g., Shapiro Shapiro et al. et 2006; al. 2006; Onken Onken et al. et 2007). al. 2007). The The dust dust sentedsented at the at the beginning beginning of Section of Section 8 largely 8 largely reproduces reproduces the the lightthat if ratio the (e.g., MGE McConnell under-predicts et al. the 2013). nuclear Moreover, stellar mass owing then to sphereHET kinematics, of influence, continuing but also to probe use the the same region sampling where the of stars dark disk,disk, however, however, may may be problematic be problematic for for construction construction of the of the lu- lu- observedobserved HET HET kinematics kinematics over over the the radial radial range range of the of the NIFS NIFS thethe presence black hole of mass the nuclear would be dust overestimated. disk, uncertainty in the view- beginhalos described to dominate in Section the potential. 7. We recovered It is encouraging best-fit values that the of minousminous mass mass model model in Section in Section 2. We 2. We conservatively conservatively masked masked data,data, as can as can× be be seen9 seen in Figure in FigureΥ 4. 4. Nevertheless, Nevertheless,Υ we we also also ran ran ing orientation is minimal, unlike some past stellar-dynamical best-fitMBH =4 model.8 10 constrainedM⊙ and byV only=10. the3 NIFS⊙ when kinematics fitting all pre- 31 thethe dust dust when8.1. when generatingModels generating with the Large-scale the MGE MGE model, model, Kinematics but but acknowledge acknowledge dynamicaldynamical models models that that fit to fit the to the combination combination9 of the of the NIFS NIFS and and studies (e.g., Shapiro et al. 2006; Onken et al. 2007). The dust sentedHET spatial at the bins, beginning and M of=4 Section.6 × 10 8 largelyM⊙ and reproducesΥ =9.8 Υ the⊙ thatthat if the if the MGE MGE under-predicts under-predicts the the nuclear nuclear stellar stellar mass mass then then HETHET kinematics, kinematics, continuing continuingBH to use to use the the same same sampling samplingV of dark of dark disk,Large-scale however, maydata be is problematiccommonly used for construction to complement of the high lu- observedwhen excluding HET kinematics three outer over bins the because radial their range dispersions of the NIFS are thethe black black hole hole mass mass would would be overestimated. be overestimated. haloshalos described described in Section in Section 7. We 7. We recovered recovered best-fit best-fit values values of of minousangular mass resolution model spectroscopy in Section 2. in We order conservatively to better constra maskedin data,below as the can HET/LRS be×9 seen9 in instrumental FigureΥ 4. Nevertheless, resolution.Υ Thus, we also we find ran MBHMBH=4=4.8 ×.810 10M⊙Mand⊙ andΥV =10V =10.3 Υ.3⊙ when⊙ when fitting fitting all all 31 31 thethe dust stellar when mass-to-light8.1. generatingModels with ratio the Large-scale MGE and orbital model, Kinematics distribution but acknowledge (e.g., dynamical models that fit to the combination9 of the NIFS and 8.1. Models with Large-scale Kinematics HET spatial bins, and M =4×.6 ×910 M⊙ andΥ Υ =9Υ.8 Υ⊙ that if the MGE under-predicts the nuclear stellar mass then HETHET spatial kinematics, bins, and continuingMBH =4BH to.6 use10 theM same⊙ and samplingV =9V . of8 dark⊙ Large-scaleLarge-scale data data is commonly is commonly used used to complementto complement high high whenwhen excluding excluding three three outer outer bins bins because because their their dispersions dispersions are are theangular black hole resolution mass would spectroscopy be overestimated. in order to better constrain halos described in Section 7. We recovered best-fit values of angular resolution spectroscopy in order to better constrain belowbelow the the HET/LRS× HET/LRS9 instrumental instrumentalΥ resolution. resolution.Υ Thus, Thus, we we find find the stellar mass-to-light ratio and orbital distribution (e.g., MBH =4.8 10 M⊙ and V =10.3 ⊙ when fitting all 31 the stellar8.1. mass-to-lightModels with ratio Large-scale and orbital Kinematics distribution (e.g., 9 HET spatial bins, and MBH =4.6 × 10 M⊙ and ΥV =9.8 Υ⊙ Large-scale data is commonly used to complement high when excluding three outer bins because their dispersions are angular resolution spectroscopy in order to better constrain below the HET/LRS instrumental resolution. Thus, we find the stellar mass-to-light ratio and orbital distribution (e.g., A Stellar-Dynamical Black Hole Mass Measurement Example The Astrophysical Journal, 817:2 (12pp), 2016 January 20 Walsh et al. 6

✦ Construct orbit-based models using supercomputers. ✦ Potential consists of contributions from the BH, stars,

and dark matter. Figure 2. Results of stellar-dynamical models fit to the NIFS kinematics. The χ2 is shown with black hole mass (left panel) and V-band mass-to-light ratio (middle panel). Each black dot corresponds to a single model, with the best-fit model highlighted as the red square. The red solid line indicates where Δχ2=9.0, fi 2 corresponding to the 3σ con dence level for one parameter. χ contours are shown for2 a grid of black hole masses and mass-to-light ratios (right panel). Each gray FpointIG. 2.— represents Results a model, of stellar-dynamical and the best-fit models model (red fit to square the NIFS) corresponds kinemat toics.M The=χ4.9is× shown109 M withand blackϒ = hole9.3 ϒ mass. Relative (left panel) to the and minimum,V-band the mass-to-lightχ2 has increased ratio NGCBH 1277e V e ∆ 2 ✦ Integrate orbits in the potential. (middleby 2.3 panel).(thick black Each contour black dot), 6.2 corresponds(thin black to contour a single), model, and 11.8 with(red the contour best-fit), corresponding model highlighted to the as 1 theσ,2 redσ, and square. 3σ con Thefidence red so intervalslid line indicates for two parameters. where χ =9.0, corresponding to the 3σ confidence level for one parameter. χ2 contours are shown for a grid of black hole masses and mass-to-light ratios (right panel). Each 9 2 gray point represents a model, and the best-fit model (red square) corresponds to MBH =4.9 × 10 M⊙ and ΥV =9.3 Υ⊙. Relative to the minimum, the χ Assign weights to each orbit has increased by 2.3 (thick black contour), 6.2 (thin black contour), and 11.8 (red contour), corresponding to the 1σ,2σ,and3σ confidence intervals for two such that the superposition parameters. MBH = (4.9 ± 1.6)x109 M⊙ matches the observed kinematics and surface brightness. ✦ Repeat for different combination of parameters until lowest χ2 is found.

(e.g., Gebhardt et al. 2003, Valluri et al. 2004, Thomas et al. 2004, Cappellari et al. 2006,

van den Bosch et al. 2008) Figure 3. The observed NIFS kinematics (top) show that NGC 1277 is rotating, with the west side of the galaxy being blueshifted, that there is a sharp rise in the velocity dispersion and generally positive h4 values at the nucleus, and that h3 and V are anti-correlated. The best-fit stellar dynamical(Walsh model et(bottom al. )2016)is shown on the same scale given by the color bar to the right and the minimum/maximum values are provided at the top of the maps. The best-fit model nicely reproduces the NIFS FobservationsIG. 3.— The and observed has a reduced NIFS kinematicsχ2 of 0.7. (top) show that NGC 1277 is rotating, with the west side of the galaxy being blue-shifted,thatthereisasharprise in the velocity dispersion and generally positive h4 values at the nucleus, and that h3 and V are anti-correlated. The best-fit stellar dynamical model (bottom) is shownFinally, on the same the NIFSscale given kinematics by the color appear bar to to the be right sensitive and the minimum/maximum to the Adopting values are higher-order provided at the polynomials top of the maps. resulted The bes int-fit slightly model better nicely 2 reproducesadopted the pPXF NIFS continuum-correction observations and has a reduced model.χ Weof 0.7. measured the fits to the galaxy spectra, but generally lower stellar velocity stellar kinematics by comparing spectra of K and M giant and dispersions, especially at the nucleus. Notably, extracting the itedsupergiant mass degeneracies stars to the withNGC the 1277 galaxy’s spectra. dark Since halo the (e.g., stars Ge wereb- ShapiroNIFS kinematics et al. 2006; using Krajnovi a multiplicativecetal.2009).TheNIFSkine-´ degree 2 polynomial hardtalso &observed Thomas with 2009; NIFS, Rusli their et spectra al. 2013) have and a similar uncertainties shape as maticsresulted extend in velocity out to dispersions a radius that of 1 were′′3(442pc),or lower by as∼ much0.4 R as. −1 . e associatedthe galaxy with spectra; the assumption however, of minor a constant adjustments stellar mass-t to theo- Therefore,36 km s compared the NIFS to data when not an only additive resolve degree the black 0 with hole a lighttemplates ratio (e.g., are needed McConnell to account et al. 2013). for differences Moreover, in owing the CO to spheremultiplicative of influence, degree but 1 also polynomial probe the regionwere utilized. where the Other stars – thebandhead presence equivalent of the nuclear widths dust and disk, for reddening uncertainty/imperfect in the vieflw-ux begincombinations to dominate of multiplicative the potential. degree It is 2 encouraging4 polynomials that with the – ingcalibration. orientation In is order minimal, to provide unlike a some better past match stellar-dynam to the observedical best-fitno additive model contribution, constrained by or only with the additive NIFS kinematics degree 0 pre-1 polynomials, produced consistent kinematics as the multi- studiesgalaxy (e.g., spectra, Shapiro we et used al. 2006; additive Onken/multiplicative et al. 2007). TheLegendre dust sented at the beginning of Section 8 largely reproduces the disk, however, may be problematic for construction of the lu- observedplicative degree HET kinematics 2 polynomial over case. the radial Therefore, range we of tested the NIFS the polynomials with pPXF. As described in Section 4, we selected fi minous mass model in Section 2. We conservatively masked data,effect as on canM beBH seenwhen in Figuretting 4. to Nevertheless, point-symmetrized we also NIFS ran an additive degree 0 and multiplicative degree 1 polynomial, kinematics extracted with an order 2 multiplicative polynomial. the dust when generating the MGE model, butfi acknowledgefi dynamical models that fit to the combination of the NIFS and and allowed pPXF to solve for the best- t coef cients We found that the black hole mass decreased to 4.2×109 M , thatsimultaneously if the MGE with under-predicts the Gauss–Hermite the nuclear moments. stellar An mass additive then HET kinematics, continuing to use the same sampling of darke the black hole mass would be overestimated. halosor by described 14%, and the in SectionV-band stellar7. We mass-to-light recovered best-fit ratio increased values of degree 0 and multiplicative degree 1 polynomial produced a to 10.0 ϒ , or by9 8%. MBH =4.8e× 10 M⊙ and ΥV =10.3 Υ⊙ when fitting all 31 good fit8.1. to theModels NGC with1277 Large-scale observations, Kinematics and the use of other fi HETThe spatialnal bins, uncertainties and M are× determined9 M byΥ addingΥ in low-order polynomials yield consistent kinematics. BH =4.6 10 ⊙ and V =9.8 ⊙ Large-scale data is commonly used to complement high whenquadrature excluding the change three outer in the binsMBH becauseand ϒ theirV from dispersions each of the are angular resolution spectroscopy in order to better constrain below the HET/LRS instrumental resolution. Thus, we find the stellar mass-to-light ratio and orbital distribution (e.g., 6

the mass of the host bulge and, as we have already begun to see, to qualitatively new conclusions.

in Hubble types. These developments lead to a signiﬁcant recalibration of the ratio of BH mass to

in ellipticals. Finally, the sample of galaxies with dynamical BH detections is larger and broader

reasons to omit BH monsters, mergers in progress, and (Section 6.7) the two largest BHs known

broad line widths are taken into account (Section 6.3). We omit these masses. Fourth, we have

• unless partly because we are now conﬁdent that emission-line rotation curves underestimate M

because of improvements in modeling (e. g., three-integral models that include dark matter), and

partly because of improvements in data (ground-based AO and integral-ﬁeld spectroscopy), partly

BH galaxies (Kormendy & Bender 2013b). Third (Section 3), we have more accurate BH masses,

ellipticals. Therefore we omit them here. Second, we now have bulge and pseudobulge data for all

• –host-galaxy correlations as classical bulges and that pseudobulges do not satisfy the same tight M

pseudobulges are grown secularly by the internal evolution of galaxy disks. We show in Section 6.8

is now a strong case that classical bulges are made in major mergers, like ellipticals, whereas

like ellipticals from pseudobulges that are structurally more disk-like than classical bulges. There

systematic eﬀects in their scatter. First, we distinguish classical bulges that are structurally

Recent advances allow us to derive more robust correlations and to better understand the

• Figure 16 shows the updated correlations of M with bulge luminosity and velocity dispersion.

classical bulges and elliptical galaxies

• • • bulge bulge e

6.6 The M –L ,M –M ,andM – correlations for σ

(points in light colors), NGC 3842, and NGC 4889. Figure 17 shows this ﬁt with 1-σ error bars.

elliptical galaxies. The lines are symmetric least-squares ﬁts to all the points except the monsters

K, e bulge

L and (right) velocity dispersion σ for (red ) classical bulges and (black)

and luminosityThe Black Hole - Galaxy Relations , • K bulge

Correlation of dynamically measured BH mass M with (left) K-band absolute magnitude M Figure 16 53 (Kormendy & Ho 2013) )] ⊙ (M BH [M )] ⨀ (M NGC & Ho 2013) (Kormendy 1277 BH [M Black Hole Mass Black Hole Mass

FigureBulge 16Bulge LuminosityLuminosity 53 StellarStellar Velocity Velocity DispersionDispersion Correlation of dynamically measured BH mass M• with (left) K-band absolute magnitude MK,bulge [L (L )] -1-1 and luminosity[LK,bulgeK,bulgeLK,bulge (LandK,K,⊙ (right⨀)]) velocity dispersion σe for (red )[ classicalσ[σe (km(km bulges s ands)] ()]black) elliptical galaxies. The lines are symmetric least-squares ﬁts to all the points except the monsters NGC 1275 (points in light colors), NGC 3842, and NGC 4889. Figure 17 shows this ﬁt with 1-σ error bars. σ ✦ The relations indicate6.6 The that M• –L BHs,bulge,M which• –M bulgeoccupy,andM scales• – e correlationsof 10−4 forpc in galaxy nuclei, classical bulges and elliptical galaxies ≲

correlate by the large-scaleFigure 16 shows the (∼ updated1 kpc) correlations bulge of M properties• with bulge luminosity of and their velocity host dispersion. galaxy, implying that Recent advances allow us to derive more robust correlations and to better understand the BHs and galaxiessystematic somehow effects in grow their scatter. and First, evolve we distinguish together. classical bulges that are structurally like ellipticals from pseudobulges that are structurally more disk-like than classical bulges. There is now a strong case that classical bulges are made in major mergers, like ellipticals, whereas pseudobulges are grown secularly by the internal evolution of galaxy disks. We show in Section 6.8 that pseudobulges do not satisfy the same tight M•–host-galaxy correlations as classical bulges and ellipticals. Therefore we omit them here. Second, we now have bulge and pseudobulge data for all BH galaxies (Kormendy & Bender 2013b). Third (Section 3), we have more accurate BH masses, partly because of improvements in data (ground-based AO and integral-field spectroscopy), partly because of improvements in modeling (e. g., three-integral models that include dark matter), and partly because we are now confident that emission-line rotation curves underestimate M• unless broad line widths are taken into account (Section 6.3). We omit these masses. Fourth, we have reasons to omit BH monsters, mergers in progress, and (Section 6.7) the two largest BHs known inFigure ellipticals. 16 Finally, the sample of galaxies with dynamical BH detections is larger and broader inCorrelation Hubble types. of dynamically These developments measured BH lead mass to aM significant• with (left recalibratio) K-band absoluten of the ratio magnitude of BHM massK,bulge to theand mass luminosity of the hostLK,bulge bulgeand and, (right as we) velocity have already dispersion begunσ toe for see, ( toredqualitatively) classical bulges new conclusions. and (black) elliptical galaxies. The lines are symmetric least-squares fits to all the points except the monsters (points in light colors), NGC 3842, and NGC 4889. Figure 17 shows this fit with 1-σ error bars.

6.6 The M• –Lbulge,M• –Mbulge,andM• – σe correlations for classical bulges and elliptical galaxies

Figure 16 shows the updated correlations of M• with bulge luminosity and velocity dispersion. Recent advances allow us to derive more robust correlations and to better understand the systematic effects in their scatter. First, we distinguish classical bulges that are structurally like ellipticals from pseudobulges that are structurally more disk-like than classical bulges. There is now a strong case that classical bulges are made in major mergers, like ellipticals, whereas pseudobulges are grown secularly by the internal evolution of galaxy disks. We show in Section 6.8 that pseudobulges do not satisfy the same tight M•–host-galaxy correlations as classical bulges and ellipticals. Therefore we omit them here. Second, we now have bulge and pseudobulge data for all BH galaxies (Kormendy & Bender 2013b). Third (Section 3), we have more accurate BH masses, partly because of improvements in data (ground-based AO and integral-field spectroscopy), partly because of improvements in modeling (e. g., three-integral models that include dark matter), and partly because we are now confident that emission-line rotation curves underestimate M• unless broad line widths are taken into account (Section 6.3). We omit these masses. Fourth, we have reasons to omit BH monsters, mergers in progress, and (Section 6.7) the two largest BHs known in ellipticals. Finally, the sample of galaxies with dynamical BH detections is larger and broader in Hubble types. These developments lead to a significant recalibration of the ratio of BH mass to the mass of the host bulge and, as we have already begun to see, to qualitatively new conclusions. The Black Hole - Galaxy Relations

✦ The local BH mass census is highly incomplete, particularly for low- and high-mass BHs. ✦ Recent work detecting high-mass BHs in BCGs and cored galaxies suggest that these objects lie above MBH - σ (e.g., McConnell et al. 2011, 2012, Thomas et al. 2016). There are hints that spiral galaxies with low-mass BHs exhibit large scatter below the MBH - σ and MBH - Lbul relations (e.g., Greene et al. 2010, Läsker et al. 2014, 2016). The Astrophysical Journal,764:184(14pp),2013February20 The Astrophysical Journal Letters,McConnell832:L26 & Ma (3pp), 2016 December 1 Greene et al.

1010

109 ) ☉ (M 8

BH 10 M

107

80 100 200 300 400 Figure 1. M –σ relation for our full sample of 72 galaxies listed in Table 3 and at http://blackhole.berkeley.edu.Brightestclustergalaxies(BCGs)thatarealsothe (McConnellcentral galaxies• of their clusters& Ma are plotted 2013) in green, other elliptical and S0 galaxiesσ are(km/s) plotted in red, and late-type spiral galaxies are plottedinblue.NGC1316is (Greene et al. 2016) the most luminous galaxy in the Fornax cluster, but it lies at the cluster outskirts; the green symbol here labels the central galaxy NGC 1399. M87 lies near the center of the Virgo cluster, whereas NGC 4472 (M49) lies 1 Mpc to the south. The black hole masses are measured using the dynamics of masers (triangles), stars (stars), or fi gas (circles). Error bars indicate 68% confidence intervals.∼ For most of the maser galaxies, the errorFigure bars in M 3.areRelationship smaller than the plotted between symbol.σ* Theand blackM dottedBH. line We t the entire sample (gray dashed line) and the early-type galaxies alone (red solid). Note the systematic offset to 1 • fi shows the best-fitting power law for the entire sample: log10(M /M ) 8.32+5.64 log10(σ/200lower km s− ).M WhenBH early-typeat a xed and late-typeσ for galaxies the megamaser are fit separately, disk the galaxies. We show elliptical (red circles), S0 (green triangles), spiral (blue squares), and megamaser disk (blue • ⊙ = 1 * 1 resulting power laws are log10(M /M ) 8.39+5.20 log10(σ/200 km s− ) for the early type (red dashed line), and log10(M /M ) 8.07+5.06 log10(σ/200 km s− ) • ⊙ = circles); double circles• ⊙ indicate= our new measurements. for the late type (blue dot-dashed line). The plotted values of σ are derived using kinematic data over the radii rinf

(L), and stellar bulge mass (Mbulge). As reported below, our new In Section 2 we summarize our updated compilation of 72 compilation results in a significantly steeper power law for the black hole mass measurements and 35 bulge masses from dy- M –σ relation than in G09 and the recent investigation by B12, namical studies. In Section 3 we present fits to the M –σ, M –L, • • • who combined the previous sample of 49 black holes from G09 and M –Mbulge relations and highlight subsamples that yield in- with a larger sample of upper limits on M from Beifiori et al. teresting• variations in the best-fit power laws. In particular, we • (2009). We still find a steeper power law than G09 or B12 when examine different cuts in σ, L,andMbulge,aswellascutsbased we include these upper limits in our fit to the M –σ relation. on galaxies’ morphologies and surface brightness profiles. In We have performed a quadratic fit to M (σ)andfindamarginal• Section 4 we discuss the scatter in M and its dependence on • • amount of upward curvature, similar to previous investigations σ, L,andMbulge.InSection5 we discuss how our analysis of (Wyithe 2006a, 2006b; G09). galaxy subsamples may be beneficial for various applications of Another important measurable quantity is the intrinsic or the black hole scaling relations. cosmic scatter in M for fixed galaxy properties. Quantifying the Our full sample of black hole masses and galaxy properties is scatter in M is useful• for identifying the tightest correlations available online at http://blackhole.berkeley.edu.Thisdatabase from which to• predict M and for testing different scenarios of will be updated as new results are published. Investigators are galaxy and black hole growth.• In particular, models of stochastic encouraged to use this online database and inform us of updates. black hole and galaxy growth via hierarchical merging predict decreasing scatter in M as galaxy mass increases (e.g., Peng 2. AN UPDATED BLACK HOLE AND GALAXY SAMPLE • 2007;Jahnke&Maccio` 2011). Previous empirical studies of Our full sample of 72 black hole masses and their host the black hole scaling relations have estimated the intrinsic galaxy properties are listed in Table 3,whichappearsatthe scatter in M as a single value for the entire sample. Herein, • end of this paper. The corresponding M versus σ, L,andMbulge we take advantage of our larger sample to estimate the scatter are plotted in Figures 1–3.Thissampleisanupdateofour• as a function of σ , L,andMbulge. previous compilation of 67 dynamical black hole measurements,

fi fi Figure 4. Distribution of MBH at xed σ*. Megamaser galaxies (open) are offset to lower MBH than the full spiral sample (blue lled) or the early-type galaxies (red filled).

Table 1 has been fixed to reflect correct “method” flags for all galaxies. The number of maser galaxies drops from 21 to 20 (note that NGC 4945 is not included in the Saglia et al. compilation) and the number of non-maser, non-S0 spiral galaxies is 17. Figures 3 and 4 from the published article were also marginally impacted and are reproduced here.

2 Current Limitations RESEARCH LETTER

1066 ✦ 10 NGC 1600 Detecting high-mass BHs is challenging because: NGC 1600 NGC 4472 1 NGC 4486 NGC 3842 ✦ 5 They are rare, and require searches over large 10105 NGC 4889 distances.

) 4 ) 104 -2 ✦ The physical extent of rsphere is large, but the –2 10 pc (kpc) pc SO I r ( L angular size is small for distant objects. ☉ 101033 ✦ They tend to be found in galaxies with central μ (L 0.1

surface brightness cores. 2 10102 (Thomas et al. 2016)

✦ Detecting low-mass BHs is challenging because: 110 100 1,000 0.1 1 1 10 r (pc) 100 1000 rb (kpc) Figure 2 | Central stellar light profiles forr (pc)NGC 1600 and for a sample Figure 3 | Black-hole sphere-of-influence radii and galaxy core radii. ✦ The physical extent of rsphere is small. We are of other core and coreless elliptical galaxies. Surface brightness profiles The radius of the black-hole sphere-of-influence, rSOI, of NGC 1600 and 15 (in the V-band) are shown for a sample of galaxies, on the basis of HST 20 core galaxies with dynamical MBH measurements is plotted against 13 limited to studying only the closest objects. observations up to a distance of 100 Mpc from Earth. NGC 4889, the core radius, rb, of each host galaxy. We calculate rSOI directly from the 2 at 102 Mpc, is included because its black-hole mass has been measured . measured MBH and the de-projected stellar density profile of each galaxy. µ is the surface brightness and r is the galactic radius at which the We calculate the core radius from fits to a core-Sérsic function 30. brightness was measured. Lower-luminosity elliptical galaxies typically Our best-fit linear correlation is log10(rSOI/kpc) = (− 0.01 ± 0.29) + ✦ High S/N and moderately high spectral resolution have rising light profiles towards the galactic centres (steep grey curves), (0.95 ± 0.08)log10(rb/kpc) (straight red line), which is statistically whereas NGC 1600 and other very massive elliptical galaxies often exhibit consistent with rb = rSOI (dotted black line). The intrinsic scatter of this is needed to measure the LOSVD. a marked deficit of stars in the central region (red curves). Highlighted are relation is ε = 0.17 ± 0.04. the brightest galaxies in the Leo Cluster (NGC 3842) and the Coma Cluster (NGC 4889), and the brightest (NGC 4472) and central (NGC 4486 or hole. Gravitational core scouring by black-hole binaries can consistently M87) galaxies of the Virgo Cluster. The stellar core in NGC 1600 (dark red produce the observed homology in the light profiles of galaxy cores16,17 curve) is the faintest known among all galaxies for which dynamical MBH and the velocity anisotropy in stellar orbits18 (see Methods). measurements are available. TMT's exquisite angular resolution and sensitivity will enable Figure 4 shows the correlation between the core radius, rb, and the ‣ measured MBH for NGC 1600 and the same sample of galaxies as is 13 observations , and a lower central surface brightness than any other shown in Fig. 3. We find an intrinsic scatter of 0.3 dex in the MBH–rb the robust measurement of both high and low-mass BHs. −2 galaxy within 100 Mpc of Earth in this sample (3,100L( pc ; Fig. 2). relation for these 21 core galaxies, in comparison to a root-mean- Such a diffuse light distribution indicates a substantial deficit of stars in squared scatter of 0.41–0.44 dex in the MBH–σ relation (where σ is the the central region of NGC 1600 in comparison with lower-luminosity stellar velocity dispersion) for the same sample (Extended Data Fig. 1). elliptical galaxies, which typically show rising light profiles towards the The MBH values in core galaxies generally do not correlate well with the galactic centres down to the smallest observable radius. Like NGC 1600, stellar velocity dispersion10,11,19. Simulated mergers of elliptical galaxies other very massive elliptical galaxies often exhibit a cored light profile, also suggest that the MBH–σ correlation may steepen or disappear where the steep light distribution characteristic of the outer part of the altogether at the high-mass end20. The most massive black holes discovered galaxy flattens to a nearly constant surface brightness at small radius thus far predominantly reside at the centres of massive galaxies con- (Fig. 2). A plausible mechanism for creating depleted stellar cores is via taining stellar cores. For this high-mass regime, our findings indicate three-body gravitational slingshots that scatter stars passing close to a that the core radius of the host galaxy is more robust than its velocity supermassive black-hole binary to larger radii. While the stars are being dispersion as a proxy for MBH. scattered, the orbit of the black-hole binary shrinks and the binary will The Schwarzschild radius (or ‘event horizon’) of the black hole in emit gravitational waves if it coalesces14. NGC 1600 is 5 × 1010 km, or 335 au, subtending an angle of 5.3 µas on 21,22 In Fig. 3 we present a new correlation between the radius, rb, of the the sky (at a distance of 64 Mpc). Recent mass measurements of the 9 9 galaxy core in the observed light profile and the radius of the black black hole in the galaxy M87 range from 3.3 × 10 M( to 6.2 × 10 M(, hole’s sphere of influence, for NGC 1600 and for a sample of 20 other corresponding to a Schwarzschild radius of 3.8–7.3 µas (at 16.7 Mpc). 15 core galaxies with reliable MBH measurements . We find a tight con- For comparison, the Schwarzschild radius of Sagittarius A* is 10 µas. nection between the two radii, and the best-fit relation is consistent Thus, after the Milky Way and possibly M87, NGC 1600 contains the with rb = rSOI. The intrinsic scatter of 0.17 dex in the rb–rSOI relation is next most easily resolvable black hole, and is a candidate for observations a factor of two smaller than that in the known scaling relations between with the Event Horizon Telescope23. 10 black-hole mass and galaxy properties. This small scatter holds over a Black holes with masses of about 10 M( are observed as quasars wide range of galaxy environments sampled in the 21 core galaxies. Our in the young Universe1. Finding the dormant-black-hole descendants finding that the two radii, rb and rSOI, are statistically indistinguishable of these luminous quasars and understanding their ancestral lineages with a small scatter strongly indicates that the core-formation mech- have been strong motivations for the search for very massive black anism is homogeneous and is closely connected to the central black holes nearby. At redshifts of 2 to 3 (about 10 billion years ago), when the

2 | N A T URE | VO L 000 | 00 M ONTH 2016 The Current Sample is Not Representative

✦ Dynamical MBH measurements have been made in a biased set of galaxies that are not representative of the local galaxy population (e.g., van den Bosch et al. 2015, Shankar et al. 2016). ✦ Proper sampling of the luminosity-size space is crucial for covering a wide variety of galaxies that have experienced diverse growth pathways (e.g., Cappellari 2016, Krajnovic et al. 2018).

Figure 2: Numerous galaxy properties(adapted such as from velocity Cappellari dispersions, 2016) mass-to-light ratios, bulge (adapted from van den fractions, molecular gas content, stellar populations, and morphology vary from the top left to the Bosch et al. 2015) bottom right of the galaxy size – mass plot. The prevailing growth process that moves galaxies in this direction are those associated with cold gas accretion, minor gas-rich mergers, and secular evolution. Conversely, dry merging increases galaxy size and mass, but not dispersion, and is the main growth channel for slow rotators found to the top right of the plot. Measuring MBH for objects that populate diﬀerent regions of the galaxy size – mass (or similarly luminosity) plot is vital for properly characterizing the distribution of BH masses with galaxy properties and for gaining insight into the role BHs play in galaxy evolution. Figure adapted from Cappellari 2016.

4.2 Analysis Methods After reducing the NIFS data with our custom pipeline, we will use velocity template stars to measure the kinematics from the K-band CO bandheads as a function of IFU location for each of the galaxies. The line-of-sight velocity distribution (LOSVD) of the stars will be parameterized in terms of a Gauss-Hermite expansion, and we will measure the first two terms, V and σ, as well as the higher order moments, h3 and h4, which quantify the distribution’s asymmetric and symmetric deviations from a Gaussian. Spatial pixels will be binned together as needed to yield sufficient S/N in the outer regions of the IFU for accurate kinematic measurements. We will then construct state-of-the-art, orbit-based, stellar-dynamical models. The main idea behind this method is that the galaxy’s gravitational potential consists of contributions from the BH, stars, and dark matter. The BH potential is given by MBH, and we will adopt a functional form for the dark halo, such as a Navarro-Frenk-White (Navarro et al. 1996) profile. The stellar potential is determined from imaging observations, which we will parametrize as the sum of concentric, two- dimensional Gaussians (e.g., Emsellem et al. 1994, Cappellari et al. 2002). This multi-Gaussian expansion (MGE) will be deprojected assuming a viewing orientation, and converted to an intrinsic stellar mass density using the M/L. Next, we will set up a representative orbital library and integrate orbits in the potential. We will assign weights to the orbits such that the superposition reproduces the observed kinematics and stellar density (e.g., Gebhardt et al. 2003, Valluri et al. 2004, van den Bosch et al. 2008). Finally, we will calculate many models, varying the parameters of interest, such as MBH, M/L, the dark halo parameters, and the viewing orientation angles. The best model will be taken to be the one that most closely matches the data in a χ2-sense. Moreover,

6 A Gemini Large Program to Measure Black Holes

✦ Awarded 253 hours with Gemini North to measure MBH using stellar-dynamical modeling methods in 31 galaxies using AO-assisted NIFS.

USA Germany Canada Netherlands Jonelle Walsh (PI) Markus Kissler-Patig John Blakeslee Mariya Lyubenova Karl Gebhardt Davor Krajnović Laura Ferrarese Aaron Barth Sabine Thater Benjamin Boizelle Glenn van de Ven Jenny Greene Jakob Walcher Brazil Kayhan Gültekin Akin Yilidirm Chile Inger Jørgensen Roderik Overzier Tod Lauer Thaisa Storchi-Bergmann Bryan Miller Nora Lützgendorf Daniel Grün Chung-Pei Ma Australia Renuka Pechetti Holger Baumgardt Anil Seth Richard McDermid Argentina Finland Monica Valluri Carlos Donzelli Ronald Läsker Gemini Proposal Section 2 Page 5 A Gemini Large Program to Measure Black Holes A Gemini Large Program to Measure Black Holes + !' !% 0$((%! !, )#$&&*& 9 10 11 12 13 14 15 16 17

/#$%'-&*,' - "#$-.% & )#$!'*&& /#$!*-'+(( !! "#$*(*( !( "#$(..

7 8 12 18 19

* "#$%!++ 5. 6 "#$(&*' 20 !. "#$*%,+21 !* "#$,&,

3 4 22 23 % "#$,(** , "#$%&(- !- "#$(&., !+ "#$%*'+ ( )#$*+,*

1 2 24

! "#$%&&' (' "#$,(,* !& "#$,(.'

Figure 2:✦ WeWe selectedwill produce a samplethe largest that uniform improves survey of coverage black hole of masses galaxy carried properties out so far, by and examining a near face-on projectionwill begin to of address the galaxy a bias in fundamental the galaxies for plane. which M MostBH’s haveM beenBH measurements measured. have been made for galaxies that fall within the oval. Our sample of 28 galaxies, suitable for AO observations and ‣ Remaining upsampled region requires observational facilities with better angular dynamicalresolution modeling and methods, sensitivity. significantly improves coverage of parameter space without popu- lating the region already well covered by previous work in thefield.SDSSr-band images for 20 galaxies in the sample are shown along with their (⟨µe⟩, Re) locations. Each box is 20 kpc on a side. We will significantly increase the number of MBH measurements for small galaxies, large galaxies, high-luminosity galaxies, low-luminosity galaxies, low-dispersion galaxies, and spiral galaxies.

Figure 3: The Kormendy & Ho (2013) relations (black dot-dashed line) are shown along with the galaxies used to derive the relations (gray circles). Objects with BH masses that occupy sparsely sampled galaxy property regimes (i.e., the green triangles outside of the oval in Figure 2) are shown by the green triangles here as well. A number of these measurements are oﬀset from the MBH−σ⋆ and MBH−Lbul relations. We show where our proposed sample could fall on the relations (red crosses). The sample is placed on MBH−σ⋆ using MBH from the MBH−Lbul prediction and on MBH−Lbul using MBH from the MBH−σ⋆ expectation. We have adopted the HET σc measurements and bulge luminosities measured from decompositions of 2MASS images. Our sample is ideal for gaining a better understanding of the distribution of MBH with galaxy properties, thereby honing the only practical tool for exploring the role of BHs in galaxy evolution throughout the Universe. Fundamental Questions

✦ Given the uncertainties in the scaling relations, the exact role BHs play in galaxy evolution and the primary physical mechanisms that drive the empirical correlations are far from understood. ✦ Other basic questions remain unanswered:

✦ Do BHs and galaxies grow in lockstep with one another over time, or do the growth of BHs precede that of host galaxies, or vice-versa?

✦ Are there BHs in low-mass galaxies and in globular clusters?

✦ How do supermassive BHs form, what are the initial seed masses, and how can they acquire enough mass so quickly after the Big Bang?

‣ TMT will allow us to explore these open questions, and gain insight into the interplay between BHs and galaxies. Redshift Evolution of the Black Hole - Galaxy Relations

The Astrophysical Journal,742:107(6pp),2011December1 Bennert et al. (Bennert et al. 2011) ✦ Currently, we cannot search for redshift evolution in the BH relations using dynamical mass measurements. We need to estimate MBH using properties of broad emission lines in AGN.

✦ Some studies report a positive evolution, such that BH growth precedes bulge growth (e.g., Woo et al. 2006, 2008, Merloni et al. 2010, Matuoka et al. 2014), but others ﬁnd no change in the relations (e.g., Salviander & Shields 2013, Shen et al. 2015).

✦ Selection biases can lead to false identiﬁcation of evolution in the relations (e.g., Later 2007, Shen & Kelly 2010, Schulze & Wisotzki 2014).

Figure 2. Left panel: MBH–Msph,⋆ relation for our sample (red pentagons; green circles if fitted by spheroid plus disk), local BLAGNs (black circles; Bennert et al. 2011), and local inactive galaxies (black triangles; Bennert et al. 2011), with the z 0relation(seeSection2.5). The errors for the local samples are omitted for = clarity (0.4 dex in MBH and 0.25 dex in Msph,⋆). Right panel: the same as in the left panel for total host-galaxy stellar mass. Here, we overplot the 89 BLAGNs from Merloni et al. (2010)(bluefilledsquares;10withupperlimitsindicatedbyarrows). (A color version of this figure is available in the online journal.) where α is the slope of the relations at z 0andisassumednot Table 3 to evolve, γ is the intercept of the relations= at z 0, and σ is Evolving MBH–M Scaling Relations = ∗ the intrinsic scatter which is also assumed to be non-evolving. Model αβ γ σ Here β describes the evolution of the scaling relation (with a Msph, 1.09 1.96 0.55 0.48 0.36 0.1 β 0implyingnoevolution).Weimposeδ-function priors of ∗ a ± − ± = Mhost, 1.12 1.68 0.53 0.68 0.35 0.1 α 1.09 (Bennert et al. 2011)andγ 0.48 for the Msph, ∗b ± − ± = = − ∗ Mhost, 1.12 1.15 0.15 0.68 0.16 0.06 relation and α 1.12 (Haring¨ & Rix 2004)andγ 0.68 for ∗c ± − ± Mhost, 1.12 1.11 0.16 0.68 0.17 0.07 = = − ∗ the MHost, relation; the priors on γ were determined by fitting ± − ± ∗ to the local AGNs from Bennert et al. (2011)whilekeepingthe Notes. slope fixed to the noted values. A normal distribution prior is a Fitted only using the 11 objects presented here. used for the intrinsic scatter with mean 0.4 and variance 0.01 b Fitted using the objects presented here and the objects from Merloni et al. and we employ a broad uniform prior for β. (Note that, strictly (2010). speaking, the variable σ accounts for both the intrinsic scatter in c Fitted only using data from Merloni et al. (2010). the relationship and the (much smaller) uncertainty on γ .) We use the z 1.0–1.2 “elliptical” stellar mass function from Ilbert = et al. (2010)toplacepriorsonthestellarmasses.Furthermore, Keeping in mind the small number statistics, the fraction of we include a prior on the black hole masses that models our disk-dominated host galaxies (36% 17%) is lower than what selection effects by using a hard cutoff at the low-mass end. This has been found by Schawinski et al. (±2011)(80% 10%) for 20 cutoff is determined from the data and models the lower limit X-ray-selected AGNs at a comparable redshift (1±.5 Star-forming galaxies at combined sample in the following. aredshiftofz 2, on the other hand, show a 33% 6% ∼ ¨ ± 3. RESULTS AND DISCUSSION fraction of interacting or merging systems (Forster-Schreiber et al. 2009). Schawinski et al. (2011)interprettheirhighfrac- Out of 11, four AGNs are clearly hosted by late-type spi- tion of spiral galaxies as a sign that secular evolution may play ral galaxies, while the rest seem to be spheroid dominated. anon-negligibleroleingrowingspheroidsandblackholes.Our

4 Dynamical Black Hole Mass Measurements in Local, Compact Galaxies

✦ NGC 1277 is part of a sample of nearby, early-type galaxies that have small sizes and luminosities for their large stellar velocity dispersions: NGC 1277 ‣ re ~ 1-3 kpc ‣ LK ~ 5x1010 L⊙ - 2.5x1011 L⊙ ‣ σc > 250 km s-1

✦ Objects are interesting because they: ‣ host some of the most massive BHs known NGC 1275

‣ are different from the massive elliptical galaxies expected to host the heaviest BHs (e.g., Dalla Bontá et al. 2009, Gebhardt et al. 2013)

‣ appear similar to z~2 galaxies (“red nuggets”) (e.g., Szomoru et al. 2012)

(SDSS) 1

Dynamical Black Hole Mass Measurements in Local, Compact Galaxies

✦ NGC 1271, NGC 1277, and Mrk 1216 host some of the most massive BHs dynamically detected to date, with MBH ~(3-5)x109 M⊙ (Walsh et al. 2015, 2016, 2017). They are surprising positive outliers from MBH - Lbul. 1

✦ Given the similarities to the z~2 galaxies (e.g., Ferré-Mateu et al. 2015, 2017, Yildirim et al 2017, Beasley et al. 2018), perhaps the local compact galaxies are relics, and reﬂect the relationship between BHs and galaxies at earlier times. Perhaps BH growth precedes that of its host galaxy. Direct Searches for Redshift Evolution with TMT

✦ TMT will facilitate the first direct search for evolution in the BH relations. We can dynamically measure MBH in various redshift bins and compare to the local relations. ✦ This is best accomplished with high-mass BHs because angular resolution is no longer a concern. The Astrophysical Journal, 803:61 (24pp), 2015 April 20 Medling et al. spectroscopy to measure black hole masses. With kinematic (Medling et al. 2015) maps that resolve inside the sphere of influence of a black hole, the complex and unrelaxed large-scale dynamics are less 1 billion important, and black hole masses can be measured to within a solar factor of a few. We note that this technique measures the masses unresolved central mass, which includes both the black hole and its accretion disk, and in some cases may also include a BH Grows Lockstep reservoir of gas feeding the accretion disk. In U et al. (2013), First Growth the black hole massTMT inK Mrk 273 N measured with this technique BH mass was consistent with the measurement made by OH maser 1 million kinematics (KlöcknerTMT Z and Baan 2004). solar Using a robust technique such as this to measure black hole masses masses in these gas-rich mergers is important because more Galaxy Grows First traditional methods of black hole mass measurements rely on assumptions that are not valid in the case of galaxy mergers. Three-integral orbital superposition models (as in Gültekin bulge mass et al. 2009; Siopis et al. 2009) are able to use large-field kinematics to separate different components to the mass profile of the galaxy; however, this approach requires a dynamically Figure 1. MBH-s* relation for isolated galaxies from McConnell & Ma relaxed system and is used therefore in isolated galaxies. (2013)(black) with three possible evolutionary tracks for merging galaxies (KSP overlaid.team: G If theültekin, black hole Gebhardt, grows first Walsh, or more quickly Greene, than theU, galaxyMalkan, bulge, Barth, AnotherJuneau) successful black hole mass measurement technique is mergers would lie above the relation (as shown by the gold arrows). If the reverberation mapping (e.g., Denney et al. 2009, and references black hole growth lags the bulge growth and is responsible for curtailing therein), which measures the time lag between flux variations evolution (e.g., quenching through AGN feedback), mergers would lie below of the continuum and the lines in the broad line region. Since the relation (as shown by the purple arrows). If instead the black hole and the bulge grow in lockstep, the mergers would remain on the relation (as shown by the cores of (U)LIRGs are so dusty, the broad line regions are the blue arrow). too obscured to view. Obtaining data at spatial resolutions sufficiently high to resolve inside the sphere of influence of a supermassive black occurs later in a merger relative to peak star formation hole at the typical redshifts of local (U)LIRGs (z 0.1) (Hopkins 2012). This scenario would predict that gas-rich requires adaptive optics systems, which are becoming available mergers would fall below black hole scaling relations. at an increasing number of ground-based observatories. Merger-driven galaxy evolution is not a complete explana- Though the Hubble Space Telescope (HST) has excellent tion for all black hole growth, however. Though some AGN resolution in the visible bands, its relatively small mirror size fi studies nd a correlation with major mergers (Koss et al. 2010; limits the resolution at the longer wavelengths (>2mμ ) Ellison et al. 2011), AGNs found with other selection necessary to look through the dust in these galactic nuclei; techniques and at different redshift ranges do not show a even at the longest wavelength available to the Wide Field fi higher rate of merging than eld galaxies (Cisternas Camera 3 (H-band), these nuclei are still sometimes obscured. et al. 2011b; Kocevski et al. 2012). Instead of mergers In order to achieve high spatial resolution in K-band, large bringing in gas, some systems probably are undergoing secular ground-based telescopes have employed adaptive optics evolution, accreting their gas directly from the cold inter- systems which measure turbulence in the Earth’s atmosphere galactic medium. Bar and spiral disk instabilities are capable of and use a deformable mirror to correct for the resulting dissipating sufficient angular momentum from this gas to fuel distortions. These distortion measurements require references, black hole growth (Hopkins & Quataert 2010a, 2011). The either a natural guide star (NGS AO) or a laser guide star plus a dividing line between these two processes is not yet well fainter natural “tip-tilt” star (LGS AO). The addition of laser understood; here we focus on understanding the black hole guide star adaptive optics has increased the area of the sky growth due to major mergers. observable with this technique. Nearby (Ultra-)Luminous InfraRed Galaxies ((U)LIRGs; Throughout this paper we have adopted a cosmology of −1 −1 Sanders & Mirabel 1996) are an excellent sample with which H0 = 70 km s Mpc , Ωm = 0.28, and ΩL= 0.72 (Hinshaw to study such gas-rich mergers. These galaxies have infrared et al. 2009). In Section 2 we present our data and reduction 11 12 fl fi luminosities upwards of 10 L☉ (10 L☉ for ULIRGs), techniques. In Section 3 we brie y describe the kinematic tting generally caused by a starburst and/or an AGN heating up techniques demonstrated in Medling et al. (2011) and U et al. dust. Their infrared luminosities correlate with merger rate, star (2013). In Section 4 we present the black hole masses measured formation rate, and AGN fraction (Ellison et al. 2013). In fact, from several tracers and in Section 5 compare them to black in the local universe, such strong infrared activity is almost hole scaling relations. Section 6 contains our conclusions. exclusively triggered by major mergers: Veilleux et al. (2002) showed that in a complete sample of IRAS ULIRGs, 117 out of 118 galaxies are in the midst of strong tidal interactions. 2. OBSERVATIONS In a companion paper (Medling et al. 2014), we have studied 2.1. The Sample the kinematics of both gas and stars in the inner kiloparsecs of a sample of (U)LIRGs, finding that nuclear disks on scales of a We have selected nine gas-rich merging galaxies in which to few tens to hundreds of parsecs are common. In two other measure the black hole masses using gas and stellar kinematics. papers (Medling et al. 2011; U et al. 2013), we demonstrated a These galaxies represent the subset of merging galaxies technique that uses high spatial resolution integral field presented in Medling et al. (2014) for which high quality

2 Direct Searches for Redshift Evolution with TMT

✦ We will use the IRIS imager to determine the luminous mass distribution, and the integral-ﬁeld spectrograph to map out the stellar kinematics in the central regions of galaxies.

✦ For z ≲ 1.5, we will measure kinematics from the (TMT) Courtesy TMT International Observatory 2.22 μm CO bandheads, 1.6 μm CO bandheads, and 0.85 μm Ca II triplet. Beyond z > 1.5, we will (IRIS) use emission lines (H�), provided that the gas is in regular rotation.

✦ R~4000 is sufﬁcient spectral resolution for the high-mass BHs. We will make use of the 9 mas pixel scale.

Courtesy TMT International Observatory (KSP team: Gültekin, Gebhardt, Walsh, Greene, U, Malkan, Barth, Juneau) Direct Searches for Redshift Evolution with TMT

The Astronomical Journal,147:93(17pp),2014April Do et al. ✦ We estimate exposure times using the (Do et al. 2014) central surface brightness of M87 [V=16.2 1000 2 2.5 (5 hours on-source) mag/arcsec (Kormendy et al. 2009); assumed K=13.2 mag/arcsec2], accounting for surface 2.0 brightness dimming and luminosity evolution 100 (Gebhardt et al. 2003). 1.5 ✦ Aim for S/N = 40 in an unbinned lenslet. 1.0 SNR Included time for offset sky exposures (1 sky Flux (photons/s/m^2/micron) 2.0 2.1 2.2 2.3 2.4 10 for every 1.5 hours on-source). Wavelength (micron) IRIS 9 mas, R = 4000 1.4 IRIS 9 mas, R = 8000 10 IRIS 50 mas, R = 4000 ✦ 10 BHs with MBH > 10 M☉ locally 1.2 OSIRIS 50 mas, R = 4000 (40 hours) 1 1.0 10 11 12 13 14 15 16 17 ✦ 0.8 K−band Surface Brightness (mag/sq. arcsec) 10 BHs in each of four redshift bins / 0.6 Figure 5. Average S N per spaxel per average spectral channel for extended (z~0.1, 0.5, 1.0, 1.5); 40 BHs total sources with uniform surface brightness with 5 hr of total integration time (20 (510 hours) 0.4 observations of 900 s) for an M0III stellar template source for different IRIS Flux (photons/s/channel) spaxel scales and spectral resolutions: 9 mas, R 4000 (solid); 9 mas, R 2.0 2.1 2.2 2.3 2.4 8000 (dotted); 50 mas, R 4000 (dot-dashed). For= comparison, we also plot= Wavelength (micron) the(KSP extended team: sourceGültekin, sensitivity Gebhardt,= of the OSIRISWalsh, IFSGreene, on Keck U, (dashed).Malkan, Barth, Juneau)

6 100 SNR 4

2 2.0 2.1 2.2 2.3 2.4 Wavelength (micron) SNR

Figure 3. Top: synthetic spectrum of a Teff 4500 K ( K3III) giant used for 10 the simulations. The ﬂuxes are given for observations= of∼ a location with surface brightness with K 13 mag arcsec 2,ataspectralresolutionR 8000. Middle: = − = the spectrum of the background and noise sources (see Section 3.2)simulated Z for an observation with a plate scale of 9 mas with an integration time of 900 s. Y Bottom: the resulting S/Nateachspectralchannelinasinglespaxel.Thesharp J dips in the S/Nareduetothehigherskybackgroundatthewavelengthsofthe H OH lines. K 1 10 11 12 13 14 15 16 17 Surface Brightness (mag/sq. arcsec) Figure 6. Average S/N per spaxel per average spectral channel for extended 150 sources with uniform surface brightness using the Z (solid), Y (dotted), 9 mas plate scale J (dashed), H (dot-dashed), and K (dot-dot-dot-dashed) ﬁlters. These curves 4 mas plate scale are calculated for the 4 mas plate-scale at a spectral resolution of 4000 and 20 900 s integrations (5 hr). ×

100 the gravitational influence of the central black hole the extended mass distribution of stars, gas, and dark matter. The LOSVD also holds information about the velocity anisotropy of the system, which can be degenerate with the black hole mass when using velocity dispersions alone to constrain the potential. 50 Typically, the LOSVD at each spatial location is either derived Min. Exposure Time (s) non-parametrically by fitting to orbit libraries (Gebhardt et al. 2000b)oritismeasuredfromthespectrabydescribingthe LOSVD as Gauss–Hermite moments (van der Marel & Franx 0 1993). Both methods lead to comparable precision in the 2.0 2.1 2.2 2.3 2.4 measurements of the black hole mass (e.g., Rusli et al. 2011; Wavelength (microns) McConnell et al. 2011a). In order to investigate how the S/Nand Figure 4. Minimum integration time, tint, necessary to be background limited, spectral resolution of IRIS affect the determination of LOSVD, with background signal greater than the combination of dark current and the 2 we use the latter method to simulate different LOSVD with the read noise (B>D+ R /tint) for the 4 mas (dashed) and 9 mas (solid) plate scale at R 8000 in the K band. The sharp dips are from the OH sky lines, first four Gauss–Hermite polynomials: the mean velocity v,the = which increases the amount of sky background at those wavelengths. velocity dispersion σ , h3,andh4.TheLOSVDisdefinedas

6 U.S. ELT Key Science ProgramIntermediate-Mass BHs and BH SeedsPage 4 Draft (G1) (courtesy of J. Greene) (Omega Cen) (Gebhardt et al. 2005) (Noyola et al. 2010) SgrA*

6 10 15 Cluster BHs = ! More Common

) 105 Heavy BHs !

Current Nuclei! (3 detections) (credit: NASA/Hubble) = rare (credit: ESO) M 10 Nuc ( 4 Current GC ! (NGC 4395) (NGC 5206) 10 (6 possible ! BH detections) (Peterson et al. 2005) (Nguyen et al. 2018) ELTs(25 - proposed) int. light!

M MWGC

Black Holes?? ELT PM ! 3 (43 proposed) 5 10 ?? Intermediate-mass Nuc

TOTAL NUMBER MWGC 102 (credit: Carnegie- LIGO ! (credit: Galex/NASA) Irvine Galaxy Survey) 10 3 10 2 10 1 100 0 Binaries 103 104 105 D (Henize 2-10) (M33) (Mpc) (Reines et al. 2011) log(Gebhardt (M et/M al. 2001)) ✦ There are some dynamical hints for IMBHs in Figureglobular 1: Left clusters: The and landscape low-mass galaxies, of dynamical but the IMBH detections within 5 Mpc of the sun. We schematicallyresults are show controversial the regions due to where insufﬁcient existing limits are found. ELTs may provide detections nearly spatial resolution and sensitivity. (credit: NASA, NRAO, an order of magnitude better, depending on the unknownSTScI) relationship between(credit: ESO) galaxy structure and MBH. Right: The cumulative distribution of MBH from our experiments under two example 4 5 seeding scenarios. In the “heavy” seeding case, all the black holes are have MBH= 10 10 M , the occupation fraction is only 50%, and very few clusters harbor seed BHs. In contrast, if runaway processes in clusters make 103 M seeds, then we expect a much larger number of detections in clusters, and a larger fraction of detections overall.

a very small fraction of dwarf nuclei (Reines, etc). There are a class of dynamical detections in low-mass stellar systems that we believe started out more massive and then were stripped; the BH masses in these systems range from 104 107 (G1, UCD, etc). Finally, there is an additional class of accretion-powered Ultra-Luminous X-ray sources (ULXs) that have luminosities above the Eddington limit for stellar-mass BHs (Soria). Many of these show pulsations and thus have been shown to be neutron stars accreting at super-Eddington rates. The best candidate for an IMBH is HLX1, with an X-ray luminosity of XXX, which is found XX kpc from a nearby S0 galaxy. There is a tentative detection of a stellar component to this object, so HLX1 is likely to be a BH in the nucleus of a stripped galaxy. Thus, we have no evidence (dynamical or accretion) for 103 104 M BHs, and a handful of dynamical and accretion-based detections of 104 105 M BHs. The ELTs will enable the first robust dynamical detection of 103 105 M BHs should they exist. This KSP: We aim to definitely determine the dominant seeding mechanisms that form supermassive BHs with two complementary experiments. First, we will use proper motions to determine the mass distribution of BHs in galactic globular clusters. Second, we will use integrated light observations to determine the mass distribution of BHs in dwarf galaxy nuclei out to 5Mpc. To- gether, these two components will determine, for the first time, the distribution of BH mass in the “intermediate-mass” range from 103 105 M . Intermediate-Mass BHs and BH Seeds

3 ✦ TMT will be sensitive to 10 M☉ BHs at the distance of Andromeda, and 4 10 M☉ BHs at 3-5 Mpc. ✦ The systems that might host IMBHs have dispersions ~15-30 km/s. We need both high spectral and high spatial resolution. ✦ R~8000 is needed, but R~10,000 preferred.

✦ See Tuan Do’s talk, Wednesday 9:20 am in the “Local Universe, Nearby Galaxies and Black Holes” breakout session! (Greene 2012) (KSP team: Bellini, Do, Libralato, Greene, Gultekin, Gebhardt, Walsh, Barth, Baldassare) Summary

✦ The local BH mass function is incomplete, and we do not know exactly what roles BHs play in galaxy evolution.

✦ We need more MBH measurements at the ends of the BH mass scale and in a wider variety of galaxies.

✦ TMT’s angular resolution and sensitivity will open up new regimes:

✦ We can dynamically measure the most massive BHs in the Universe, exploring the extreme end of BH evolution.

✦ We can dynamically measure the smallest BHs in the Universe, exploring the mass function of initial BH seeds.

✦ We can directly trace redshift evolution in the BH scaling relations over a meaningful baseline, exploring when/how BHs and galaxies grow.