The Golden Mass of Galaxies and Black Holes with Deep Learning Avishai Dekel the Hebrew University of Jerusalem & UCSC
Total Page:16
File Type:pdf, Size:1020Kb
The Golden Mass of Galaxies and Black Holes with Deep Learning Avishai Dekel The Hebrew University of Jerusalem & UCSC Ofer Lahav’s 60 th Birthday, April 2019 A Characteristic Mass for Galaxy Formation Efficiency Black Holes of galaxy formation low-mass high-mass quenching quenching Behroozi+ 2013 A Characteristic TimeMass for Galaxy Formation Star-formation density high-mass low-mass quenching quenching typical halos at z~2 11-12 are ~10 M AGN Activation at the Critical Mass Forster-Schreiber+18 KMOS 3D z=0.6-2.7 599 galaxies Why do galaxies like to form at the golden mass (and time)? Why do black holes grow rapidly once their host galaxies are above the golden mass? A Characteristic Mass for Galaxy Formation 10.5 12 Mstar ~10 M Mvir ~10 M Vvir ~100 km/s - Hot halo (virial shock heating) at M>M crit Rees & Ostriker 77, Silk 77, Binney 77, Dekel & Birnboim 06 - Supernova feedback efficient at M<M crit (V crit ) Larson 74 , Dekel & Silk 86 -> Compaction to Blue Nuggets + quenching at ~M crit (any z) Zolotov+15, Tacchella+16, Dekel+17 -> Black Holes suppressed by supernovae at M<M crit , Black hole growth at M>M crit -> Quenching of star formation at M>M crit triggered by compaction, maintained by hot halo & black hole 2. Virial Shock Heating in Massive Halos Dekel & Birnboim 06, Rees & Ostriker 77, Silk 77, Binney 77 Hot Halo Scale: shutdown of cold gas supply Dekel & Birnboim 06, Rees & Ostriker 77, Silk 77, Binney 77 11.8 critical halo mass ~~10 Mʘ −1 < −1 tcool tcompress slow cooling -> shock log T Kravtsov+ 11.5-12 Mcrit by Shock Heating: M vir ~10 M 10 13 Simulations: Ocvirk, Pichon, Teyssier 08; Dekel+ 09 hot CGM hot CGM 12 cold streams 10 Dekel & shock heating Birnboim 06 Mvir [M ʘ] cold CGM 10 11 typical halos Press & Schechter 74 0 1 2 3 4 5 redshift z Empirical Model From Observations Behroozi+ Fraction of star-forming galaxies 3. Supernova Feedback in Low-mass Halos Larson 74, Dekel & Silk 86 Supernova Fdbk Scale V vir ~100 km/s Dekel & Silk 86 Energy fed to the ISM by supernova bubble before fading: ≈ 2 ≈ ESN M ∗VSN , VSN 120 km/s Gas binding energy: ≈ 2 2 = Egas M gas Vvir , Vvir GM / R ESN ~ E gas & star formation at peak efficiency M *~M gas --> an upper limit for efficient supernova feedback: ≈ → ≈ × 11 = Vvir, SN 120 km/s M vir, SN 7 10 M ⊗ at( z )0 11.5-12 Mcrit by SN Feedback: M vir ~10 M 10 13 hot CGM hot CGM 12 cold streams 10 Dekel & shock heating Birnboim 06 Mvir [M ʘ] cold CGM 10 11 Dekel & Silk 86 typical halos Press & Schechter 74 0 1 2 3 4 5 redshift z 4. Wet Compaction to Blue Nuggets Dekel & Burkert 14, Zolotov+15, Tacchella+16a,b, Dekel, Lapiner+18 Red Nugget Blue Nugget gas + young stars wet compaction max density core: blue nugget vdi disk core depletion: a hole and a ring vela v 2 07 stars wet compaction max density core: blue nugget green nugget red nugget + envelope vela v 2 07 Compaction -> quenching in the inner 1 kpc stars 10 1 kpc dark matter M gas 8 diffuse compaction quenching 2 star-formation rate M 0 time Compaction and Quenching in Simulations diffuse star forming blue nugget L-shape track quenched red nugget Zolotov+15 Tacchella+16 Dekel+17 diffuse compact Observed L-Shape Track Barro+17 BN compaction star forming quenching quenched diffuse compact Blue Nugget --> Red Nugget gas stars gas stars blue nugget red nugget dense gas core -> dense stellar core gas depletion from core, gas ring may form, -> inside-out quenching stellar core remains dense from BN to RN What is the Trigger of wet Compaction? Drastic loss of angular momentum - Mergers (major, minor) (Barnes, Hernquist 91; Hopkins+ 06) - VDI-driven inflow (Dekel, Burkert 15) - Counter-rotating streams (Danovich+ 15) - Tidal compression (Dekel+ 03; Renaud+ 14; Mandelker+ 16) - Triaxial halo core (Ceverino+ 15; Tomassetti+ 16) - Return of recycled low-AM gas (Elmegreen+ 14; DeGraf+ 16) Counter-rotating Streams Danovich+15 Rv 1/3 streams are counter-rotating A Critical Mass for Blue Nuggets (at all z) 10 11.5 Mstar ~10 M Mvir ~10 M Hot CGM SN fdbk Origin of BN scale? SN fdbk + hot CGM time Deep Learning: BNs inTheory vs Observations Huertas-Company+ 18 Pre BN Blue Nugget Post BN Identify BNs in 3D simulations (gas, stars) Train CNN on mock images of simulations Classify BNs in HST images Blue Nuggets at a Golden Mass in CANDELS Deep Learning: Huertas-Company+ 18 Luminosity (mass) eliminated Convoluted Neural Network Transition in Galaxy Properties at the BN - Diffuse -> compact core + extended disk - Star forming -> quenched - Oscillations across the Main Sequence -> quenching - Dark-matter -> baryon dominated core - Prolate -> oblate stellar system - V/ σ~1 -> V/ σ~4, dispersion -> rotation dominated -… 4. Compaction and Black-Hole Growth Dekel, Lapiner, Dubois+ 2017 Interplay between SNe and BHs RAMSES Simulation SN+BH by Dubois+ 15 z=3.60 z=3.44 no SN merger with SN merger compaction time - M<M crit : strong SN fdbk suppresses BH growth - Compaction overcomes SNe -> rapid BH growth - M>M crit : AGN fdbk helps quenching Compaction -> BH Growth stars 10 1 kpc dark matter M gas 8 black hole 2 star-formation rate M 0 diffuse compaction quenching time Compaction driving BH Growth New Horizon simulations SN+BH Dubois+ Lapiner+ stars stars 1kpc 4 compaction BH x10 DM 1kpc gas 1kpc BH growth gravity rBH Compaction driving BH Growth New Horizon simulations SN+BH Dubois+ Lapiner+ stars stars stars 1kpc BH x10 4 BH x10 4 stars 1kpc compaction compaction gas 1kpc BH growth BH growth gas 1kpc 5 BH Below the Linear Relation for M BH ~10 M New Horizon Simulation Lapiner+ Blue Nugget Magic Mass for BH is Robust in Simulations EAGLE Illustris TNG Bower+ 16 Habouzit+18 no SN w SN FIRE Angeles-Alcazar+17 11.5-12 BH Growth by Compaction at M vir ~10 M - Mvir < M crit , pre-compaction SN phase : Vesc < 100 km/s -> SN winds of 100 km/s evacuate the core -> BH growth is suppressed , Blue-Nugget formation is suppressed - Mvir ~ M crit , compaction overcoming SN fdbk : The compressed gas activates rapid BH growth - Mvir > M crit , post-compaction hot CGM phase : Vesc > 100 km/s -> SN winds are bound (by halo potential and hot gas) -> gas falls back in -> BH growth continues -> AGN self-regulates with the accretion, AGN fdbk keeps the CGM hot and suppresses SFR long term Observed High Fraction of AGN in BN Phase Kocevski+17 SFR RN diffuse BN comactness s -> Compaction triggers BH growth and AGN -> quenching AGN Activation at the Critical Mass Forster-Schreiber+18 KMOS 3D z=0.6-2.7 599 galaxies Fraction of AGN dM/dt L 2 L = 0.1 dM bh /dt c simulations BH accretion rate L=10 42.5 KMOS 3D New Horizon sims AGN Correlate with Compactness Forster-Schreuer+18 KMOS 3D z=0.6-2.7 599 galaxies M*=10.75-11.3 M*=10-10.75 incidence incidence A Characteristic Mass for Galaxy Formation 10.5 12 Mstar ~10 M Mvir ~10 M Vvir ~100 km/s - Hot halo (virial shock heating) at M>M crit Rees & Ostriker 77, Silk 77, Binney 77, Dekel & Birnboim 06 - Supernova feedback efficient at M<M crit (V crit ) Larson 74 , Dekel & Silk 86 -> Compaction to Blue Nuggets + quenching at ~M crit (any z) Zolotov+15, Tacchella+16, Dekel+17 -> Black Holes suppressed by supernovae at M<M crit , Black hole growth at M>M crit -> Quenching of star formation at M>M crit triggered by compaction, maintained by hot halo & black hole Conclusions 10 The golden mass M*~10 M (and z~2) is due to suppression of SFR by SN fdbk at low masses and CGM heating at high masses (+AGN) Wet compaction to a Blue Nugget is a dramatic event in the history of many galaxies (~40% by mergers, 30% counter-rotating streams) The major compaction is typically at the golden mass, when SN fdbk becomes ineffective and cold streams penetrate the hot CGM The Blue Nugget marks transitions in most galaxy properties: mass, size, SFR, gas fraction, metallicity, dust, dark-matter dominance, shape, kinematics, … BH growth is suppressed by SN feedback in M<M crit halos, and is triggered by Wet compaction in M>M crit halos, where SNe are ineffective & the CGM is shock heated. The resulting AGN helps maintaining the quenching Deep Learning is a tool for confronting simulations with observations 11.5-12 The Golden Mass M vir ~10 M 10 13 hot CGM SF quenching hot CGM 12 BH growth cold streams 10 Dekel & shock heating Birnboim 06 Mvir [M ʘ] 10 11 Dekel & Silk 86 SN-suppressed SF SN-suppressed BH typical halos Press & Schechter 74 0 1 2 3 4 5 redshift z Cosmic Velocity Fields, 9th IAP Colloquium, 1993 .