Simulations of Disk Accretion onto Black Hole Binaries Brian D. Farris 1,2 Zoltan Haiman 2 Paul Duffell 1 Andrew MacFadyen 1 Center for Cosmology and Particle Physics, New York University 2 Department of Astrophysics, Columbia University May 19, 2013 Brian D Farris LISA Symposium 2014 1 Introduction Physical Picture All bulge galaxies have SMBH at center with M ∼ 105 − 109 M Galaxy mergers → formation of massive BH binary in merged remnant. Separation decreases by: 1 2 3 dynamical friction gravitational slingshot interactions gravitational radiation The Final Parsec Problem is Not a Problem! Circumbinary disk forms Brian D Farris LISA Symposium 2014 Motivation GWs from SMBH binaries detectable by eLISA during inspiral. May be detectable by Pulsar Timing Arrays for massive (∼ 108 − 109 M ) binaries at z ≈ 1. Gaseous accretion flow around binary may be a source of detectable EM radiation Help with source localization Standard Sirens (distance from GWs, redshift from EM) Learn about SMBH merger rates. Brian D Farris LISA Symposium 2014 General Picture Viscous stresses in the disk transport angular momentum outward and allow gas to migrate inward Tidal torques from the binary add angular momentum and drive gas outward Viscous stresses balance balance tidal torques at inner edge of disk (redge ≈ 2a) Binary carves out a low-density cavity surrounded by a circumbinary disk. Quasi-equilibrium state can be maintained provided tvis tgw (pre-decoupling epoch) When tgw . tvis (i.e. GW dominated regime), binary inspiral must be included in simulations. We focus on pre-decoupling epoch. Questions How hollow is the cavity? Is the accretion rate suppressed by the presence of a binary? Does the binary leave an imprint in the accretion rate? How is the accreted mass divided between the primary and the secondary? How are continuum spectra modified by presence of a binary? Brian D Farris LISA Symposium 2014 Previous Work 2D Newtonian MacFadyen & Milosavljević 1980 Cuadra et al. 2009 Roedig & Sesana, 2012 D’Orazio et al. 2012 Farris et al. 2013 1D Examples: Goldreich & Tremaine 1980 Artymowicz & Lubow 1994 Milosavljević and Phinney 2005 Haiman et al. 2009 Tanaka & Menou 2010 Liu & Shapiro 2010 Kocsis et al. 2012 Tanaka 2013 Useful for predecoupling, widely separated binaries. Can accomodate high res., many orbits. 3D Use approximate angle-averaged tidal torque formulae. Newtonian: Useful for probing qualitative features of accretion. Relativistic: Fails to capture important, nonaxisymmetric features such as accretion streams. Shi et al. 2012 Roedig et al. 2012 Bode et al 2011 Farris et al. 2012 Noble et al. 2012 Giacomazzo et al. 2012 Computationally expensive. Often require excised inner regions, thick disks, short simulations, etc. Brian D Farris LISA Symposium 2014 DISCO - Duffell & MacFadyen 2012, 2013 Solves (Magneto)Hydrodynamics equations Uses conservative, shock-capturing finite-volume methods Effectively “Lagrangian”, as cells are able to move with fluid Minimizes advection errors, allows for longer timesteps Brian D Farris LISA Symposium 2014 normalized total accretion rate For all mass ratios, accretion rate is enhanced relative to that of single BH. 1.8 1.7 Consistent with some previous studies (e.g. Shi et al. 2012) Ṁ /Ṁ 1.5 1.4 Consensus emerging that accretion is not significantly reduced by presence of binary. 0 1.6 1.3 1.2 Simulations needed to verify that this holds at smaller h/r . hṀi/hṀ0 i approaches 1 for small q, as expected. Brian D Farris 1.1 1.0 0.0 0.2 0.4 LISA Symposium 2014 q 0.6 0.8 1.0 Growth of disk eccentricity Eccentric inner cavity for large mass ratio binaries seen in many calculations, e.g. MacFadyen and Milosavljević 2008 D’Orazio et al. 2012 Shi et al. 2012 Noble et al. 2012 Farris et al. 2012 A fraction of gas in each stream does not accrete onto BH, but rather is flung outward. This fraction impacts cavity wall on the opposite side from which it entered. If one stream is slightly larger it will push the opposity wall more, weakening the opposite stream. ⇒ the imbalance grows. Brian D Farris LISA Symposium 2014 Comparison with analytic mini-disk size estimates Artymowicz & Lubow 1994 (cyan circles) Brian D Farris LISA Symposium 2014 Minidisk timescale Accretion timescale tvis,md = = 2 ri2 3 νi α −1 h/r −2 r 3/2 q −1/2 md 42 tbin 0.1 0.1 0.25a 0.1 tvis,md > tbin ⇒ minidisks are persistent. caveats We have assumed h/r ∼ 0.1 everywhere, including minidisks. If they are actually much hotter the accretion timescale is shortened. We have assumed α = 0.1 everywhere. MHD simulations have indicated it may be larger near inner disk edge, and possibly inside minidisks as well. Binary eccentricity may reduce sizes of minidisks, leading to shorter accretion timescale. Brian D Farris LISA Symposium 2014 q 2.5 = 0.053 = 2.0 1.0 0 Ṁ/Ṁ 0 2.0 1.5 1.0 0.5 0.5 0.0 400 0.0 400 405 410 415 t/tbin 420 425 405 410 415 t/tbin 420 100 80 80 20 0.5 1.0 ω/ωbin q 1.6 = 1.5 2.0 0.43 1.4 0 Ṁ/Ṁ Ṁ/Ṁ 0 1.2 1.0 0.8 0.6 0.4 400 405 410 t/tbin 415 420 425 0.5 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 400 1.0 ω/ωbin q = 1.5 2.0 0 2.5 2.0 2.5 410 t/tbin 415 420 425 60 40 0 = 1.5 2.0 2.5 420 425 2.0 2.5 1.0 0.6 405 400 Power Power Power 1.5 ω/ωbin q 1.2 20 ω/ωbin 425 1.0 80 1.0 420 0.8 80 0.5 1.0 1.4 80 0 0.5 0.67 100 20 415 60 100 60 t/tbin 40 100 40 410 20 0 2.5 405 0 0 60 40 Ṁ/Ṁ 20 Power Power Power 100 80 60 0.25 1.0 0.5 400 425 = 1.5 100 40 q 0.11 2.5 1.5 Ṁ/Ṁ 0 Ṁ/Ṁ q 3.0 2.0 405 410 t/tbin 415 60 40 20 0.5 1.0 Brian D Farris ω/ωbin 1.5 2.0 2.5 LISA Symposium 2014 0 0.5 1.0 ω/ωbin 1.5 Mass Ratio Evolution dq d = dt dt M2 M1 = M2 M1 Ṁ2 Ṁ1 − M2 M1 ! 10 3 Ṁ2 /M2 > Ṁ1 /M1 ⇒ q increasing 2 1 ) ( 1 2 10 Consistent with previous studies (Hayasaki et al. 2007, Cuadra et al. 2009, Roedig et al. 2011,2012, Hayasaki et al. 2012) ( 2 Ratio > 1 for all cases. Binary driven toward equal mass. Ṁ /M / Ṁ /M 1 ) 10 10 10 Possible that ratio < 1 for q less than some q0 , leading to bimodal mass-ratio distribution. Brian D Farris 0 -1 0.0 0.2 LISA Symposium 2014 0.4 q 0.6 0.8 1.0 Radiative Cooling Dynamics are sensitive to tvis of minidisks Isothermal prescription locks h/r to match that of circumbinary disk Need to self-consistently balance viscous heating and shock heating with radiative cooling Optically thick disk ⇒ qcool = 4σ 4 T 3τ Assume electron scattering opacity Neglect radiation pressure ⇒ P = (Σ/mB )kT Include viscous heating source term in energy evolution equation Test that scheme can reproduce Shakura-Sunyaev solution Brian D Farris LISA Symposium 2014 show movie Brian D Farris LISA Symposium 2014 Density Brian D Farris LISA Symposium 2014 Emission - 2D Brian D Farris LISA Symposium 2014 Emission 1D Brian D Farris LISA Symposium 2014 spectrum 103 minidisks non-minidisk ∝ν1/3 total 102 Lν 101 100 100 101 102 hν [a −13/16 M −1/8 M−305/4 eV] 100 Brian D Farris 8 LISA Symposium 2014 103 Summary & Future Work Summary First simulations of circumbinary disk accretion using moving-mesh, finite-volume code. Ṁ onto binary not reduced. Gas efficiently enters cavity along streams. For each mass-ratio, persistent mini-disks are formed. Accretion timescale of mini-disks exceed binary orbital timescale. Mini-disk sizes in rough agreement with analytic predictions (Artymowicz & Lubow 1994). Significant periodicity in Ṁ for q & 0.1. Binary torques can excite eccentricity in inner disk and create overdense lump. Orbital frequency of lump can dominate Ṁ periodograms. For each mass-ratio considered, accretion rate onto secondary is large enough to cause q to increase. Emission can be enhanced in cavity region relative to single BH case Continuum spectra steepen in X-rays due to hot emission from minidisks and streams Brian D Farris LISA Symposium 2014 Extra Slides Brian D Farris LISA Symposium 2014 Setup Keep aspect ratio of cells ≈ 1, concentrate resolution near BHs. α-law viscosity ν = αcs h, with α = 0.1. Run simulation for longer than a viscous time at the cavity edge, so that quasi-equilibrium state is reached. tsim & tvis (redge ) ∼ 300 r edge 2a 3/2 10 δ/a include cavity in computational domain, treat accretion by adding sink term to continuity equation: dΣ Σ =− dt sink,i tvis,i 10 10 δr rδφ 0 -1 -2 10 -1 tbin Vary mass ratio in range 0.026 ≤ q ≤ 1.0. Brian D Farris LISA Symposium 2014 10 0 r/a 10 1 10 2 Ṁ/Ṁ 0 shorten tacc in sink prescription by factor of 100 q 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 400 405 410 1.0 t/tbin 415 420 425 2.0 2.5 Power 140 120 100 80 60 40 20 0 = 0.5 1.0 ω/ωbin 1.5 Conclusion The minidisk accretion timescale is extremely important in determining the periodicity of Ṁ. Total time-averaged accretion rate mostly unchanged. Brian D Farris LISA Symposium 2014 Mini-disks ”dampen” accretion variability q 10 = 1.0 0 8 ( = ) Ṁ r a /Ṁ Compare actual accretion rate with rest mass flux through surface at r = a. Flux through r = a much more variable. 6 4 2 0 −2 400 405 410 t/tbin 415 420 425 2.0 2.5 200 Exaggerates Ω = 2Ωbin component. 150 Power Time averaged Ṁ is unchanged (expected for quasi-steady state). 100 50 0 Brian D Farris 0.5 1.0 LISA Symposium 2014 ω/ωbin 1.5 Viscosity Code Test “Cartesian” test performed on cylindrical grid Test balance of all viscous forces and hydrodynamic fluxes and source terms in all directions. t =0 1.0 , 1D high-res , 2D DISCO t =8 t =8 0.8 initial state 0.6 = 1.0 P = 0.1 vx = 0 = (x − x0 )2 exp − σ2 vy v y ρ 0.4 0.2 0.0 0 1 2 viscosity ν = ν0 ΓP 2 x ρ Brian D Farris LISA Symposium 2014 3 x 4 5 6 q=0.053 q=1.0 Brian D Farris LISA Symposium 2014 Brian D Farris LISA Symposium 2014 Surface Density initial =0.026 =0.053 =0.11 q =0.25 q =0.43 q =0.67 q =0.82 q =1.0 q 0.6 q q 0.5 Σ/Σ0 0.4 0.3 cavity size and max eccentricity vs q 3.5 0.2 3.0 0.1 2.5 rc 0.0 4 2 6 8 10 r/a 14 12 2.0 1.5 1.0 0.16 Eccentricity 0.14 ẽmax 0.12 0.20 q =0.026 q =0.053 q =0.11 0.08 0.06 q =0.25 0.15 0.10 q =0.43 0.04 q =0.67 ẽ1 q =0.82 0.02 0.0 q =1.0 0.10 0.2 0.05 0.00 1 2 3 4 5 r/a 6 7 8 9 10 Brian D Farris LISA Symposium 2014 0.4 q 0.6 0.8 1.0