Shakura–Sunyaev Alpha Disk Model
Accretion — the slow infall of gas onto a compact object through a rotating disk — is the most efficient sustained energy source in the universe short of matter–antimatter annihilation, releasing up to about 10% of rest-mass energy for a non-rotating black hole and 42% for a maximally spinning one, dwarfing the 0.7% that fusion extracts from hydrogen. Yet a disk cannot simply fall in. Gas on a circular orbit carries angular momentum, and to spiral inward it must hand that momentum to gas further out. Ordinary molecular viscosity is fantastically too weak: it would take longer than the age of the universe to drain a typical disk. Something must transport angular momentum outward far faster than molecular diffusion, and in 1973 nobody knew what. Nikolai Shakura and Rashid Sunyaev made the boldest possible move: rather than solve the transport problem, they hid it. They wrote the vertically integrated shear stress as a single number α times the local pressure, absorbed all the unknown microphysics — now understood to be magnetized turbulence — into that one parameter, and asked what could be said regardless of its value. The answer was: almost everything that matters observationally. The disk's temperature profile, total luminosity, and spectral shape all follow from conservation of mass, angular momentum, and energy alone, completely independent of α; the parameter controls only how fast the gas drains and how much piles up. That clean separation — robust observables on one side, the murky transport physics quarantined on the other — is why the α-disk remains the workhorse model of cataclysmic variables, X-ray binaries, active galactic nuclei, and protoplanetary disks half a century later.
- The Shakura–Sunyaev model describes a geometrically thin (H ≪ R), optically thick accretion disk: gas orbits on nearly Keplerian circles and drifts inward as turbulent stress carries its angular momentum outward.
- The unknown turbulence is parameterized by one number α, with stress τ = αP, equivalently ν ≃ α c_s H, and observationally α ≈ 0.01 to 0.4.
- Steady-state conservation fixes the product νΣ, so the dissipation, the temperature T_eff ∝ R^(−3/4), and the luminosity are all independent of α — the model's triumph.
- Integrating the dissipation gives L = GMṀ/2R_in: exactly half the released gravitational energy is radiated, the other half remaining as orbital kinetic energy at the inner edge.
- The disk radiates as a sum of blackbodies — a multicolor disk with F_ν ∝ ν^(1/3) over an intermediate band — the standard spectrum of X-ray binaries and the AGN big blue bump.
- It fails at high Ṁ (radiation-pressure and thermal/viscous instability) and low Ṁ (collapse into a hot advection-dominated flow), and needs relativistic correction near the ISCO.
- Key idea: replace the unknown turbulent viscosity by one number — the stress equals α times the pressure, so ν ≃ α c_s H — and steady-state conservation then fixes the dissipation, making the temperature and luminosity independent of α while α alone sets the surface density and inflow speed.
Replace the unknown turbulent viscosity by a single dimensionless number: the vertically integrated shear stress equals α times the pressure, so ν ≃ α c_s H. Steady-state conservation of mass and angular momentum then fixes the local dissipation rate, giving a radiative flux and effective temperature that are independent of α, while α sets only the surface density and inflow speed.
Where this lives today
The α-disk is calibrated and stress-tested at the frontier. In X-ray astronomy the "diskbb" and relativistic "kerrbb" models fit the thermal state of black-hole binaries seen by NuSTAR and NICER, and the constancy of the inferred inner radius in the soft state — the L ∝ T⁴ track — measures black-hole spin through the ISCO. In active galactic nuclei, continuum reverberation mapping times UV-to-optical light-travel delays to map the disk temperature profile directly, finding disks the right shape but often puzzlingly larger than T ∝ R^(−3/4) — a live tension. The Event Horizon Telescope images of M87 and Sgr A probe the innermost flow, where the thin disk gives way to a hot, advection-dominated flow. In the dwarf-nova community the thermal–viscous "disk instability model" — the α-disk with an ionization-dependent α — reproduces outburst light curves and demands that α jump from ≈ 0.01 in quiescence to ≈ 0.1–0.4 in outburst. That is the model's sharpest confrontation with theory: magnetorotational-instability simulations tend to give only α ≈ 0.01–0.03, and closing that gap is a standing open problem. Meanwhile ALMA applies the same equations to protoplanetary disks, gauging how fast planet-forming material drains onto young stars.
The physical picture
Picture two adjacent rings of gas orbiting a central mass. The inner ring moves faster — Keplerian rotation means Ω ∝ R^(−3/2), so there is a velocity shear across the gap. Turbulent eddies straddling it carry fast fluid outward and slow fluid inward, a frictional drag: the inner ring is braked, loses angular momentum, and sinks deeper into the well; the outer ring is spun up and drifts out. Angular momentum flows outward, mass inward, and the shear energy is dissipated locally as heat. Because the disk is optically thick, that heat diffuses vertically to the two faces and escapes as a blackbody glow. The first figure shows this engine, together with the edge-on geometry that makes the disk thin.
The second figure is the model's most celebrated prediction: the radial run of surface temperature. Far from the inner edge T_eff ∝ R^(−3/4), so the disk is a nested set of blackbody rings — hot and blue at the center, cool and red at the rim. But at the inner boundary it is forced cold: the zero-torque condition switches off the dissipation, so the temperature climbs from zero at R_in, peaks at exactly R = (49/36)R_in ≈ 1.36 R_in, and only then joins the R^(−3/4) decline. Summing Planck curves of falling temperature yields the flat F_ν ∝ ν^(1/3) plateau that is a disk's fingerprint in accreting-system spectra.
The α prescription itself is a confession of ignorance dressed as a scaling law. Turbulent eddies cannot exceed the disk thickness H — the only available length — nor the sound speed c_s, since supersonic turbulence shocks and thermalizes almost instantly. The viscosity, roughly (eddy speed) × (eddy size), is therefore bounded by ν ≲ c_s H. Writing ν = α c_s H with α ≲ 1 keeps the maximal-turbulence expectation while sweeping every microscopic detail into one knob — which, as shown below, turns out not to touch the observables we most cleanly measure.
The core equations
The closure that defines the model replaces the true stress tensor with a single scaling: the viscosity is set by the largest and fastest eddies the disk can support.
where: ν — kinematic turbulent viscosity (cm² s⁻¹); α — dimensionless viscosity parameter, 0 < α ≲ 1; c_s — midplane sound speed, c_s² = P/ρ; H — vertical scale height. Reading: α measures the turbulence as a fraction of the maximal c_s H; empirically α ≈ 0.01 (quiescent) to ≈ 0.1–0.4 (ionized, outbursting disks). The scale height comes from vertical hydrostatic equilibrium in the central object's tidal field, tying the disk's thickness directly to its temperature.
where: H — pressure scale height; Ω_K — Keplerian angular velocity; M — central mass; G — gravitational constant. Reading: since v_φ = Ω_K R, this says H/R = c_s/v_φ, so a thin disk (H ≪ R) is one whose gas orbits highly supersonically — it must cool efficiently to keep c_s small. In steady state the disk is a conveyor belt: the same mass crosses every radius per unit time, so Ṁ is a constant of the flow, independent of R.
where: Ṁ — mass accretion rate (g s⁻¹); Σ — surface density; v_R — radial drift velocity (negative for inflow); R — cylindrical radius. Reading: a fast low-Σ flow and a slow high-Σ flow carry the same Ṁ — the freedom α later resolves. Conservation of angular momentum, integrated to a zero-torque inner boundary, fixes the crucial combination νΣ — the equation that quarantines α.
where: R_in — inner radius (a stellar surface or the innermost stable circular orbit) where the viscous torque vanishes; νΣ — viscosity–surface-density product. Reading: ν and Σ each depend on α, but their product depends only on Ṁ, M, and geometry — so any observable built from νΣ cannot know α. Equating the viscous dissipation to blackbody emission from each optically thick face gives the effective temperature — the model's signature prediction.
where: T_eff — local effective temperature; σ — Stefan–Boltzmann constant. Reading: far out the bracket → 1 and T_eff ∝ R^(−3/4); the bracket forces T_eff → 0 at R_in and a maximum at R = (49/36)R_in. No α appears anywhere. Integrating that flux over both faces gives the bolometric luminosity — and a famous factor of one-half.
where: L_disk — luminosity from both faces; η — radiative efficiency (≈ 0.06 for a Schwarzschild ISCO, up to 0.42 for extreme Kerr). Reading: the disk radiates exactly half the potential energy liberated down to R_in; the other half survives as orbital kinetic energy at the inner edge. Summing the blackbody rings produces the multicolor-disk spectrum over an intermediate band.
where: F_ν — spectral flux density; ν — photon frequency. Reading: valid for kT_out ≪ hν ≪ kT_in, between the Rayleigh–Jeans tail of the coolest ring and the Wien cutoff of the hottest — a direct consequence of T ∝ R^(−3/4). Finally, α reappears where it must — the rate at which the disk drains.
where: t_visc — radial-drift / viscous-evolution timescale; H/R — aspect ratio. Reading: since H/R ≪ 1, the disk drains over many orbital periods, and a smaller α means a slower, longer-lived disk — this is what α actually controls.
The goal is to derive the surface-temperature profile of a steady thin disk from conservation laws alone, and to show explicitly that the transport parameter α cancels out of it. We write the local dissipation in terms of the shear, insert the Keplerian rotation law, substitute the angular-momentum result that fixes νΣ, and equate to blackbody emission from each face. The pivotal point: dissipation depends on ν and Σ only through their product, which conservation has already pinned to Ṁ — so α, living inside ν and Σ separately, has nowhere to appear in the final temperature.
Step 1. The dissipation per unit area from one face equals half the rate at which the shear stress works against the differential rotation, D(R) = (1/2) νΣ (R dΩ/dR)². Why: stress νΣ times strain rate R dΩ/dR is a power per unit area; the one-half splits emission between the two faces.
Step 2. For Ω = Ω_K ∝ R^(−3/2) the logarithmic shear is exactly R dΩ/dR = −(3/2)Ω_K, so D(R) = (9/8) νΣ Ω_K². Why: the rotation law is fixed by gravity, not the gas, so the shear is a clean geometric factor of 3/2.
Step 3. Substitute the angular-momentum result νΣ = (Ṁ/3π)[1 − (R_in/R)^(1/2)] from equation (4). Why: the crux — ν and Σ enter only as their product, set entirely by Ṁ and the zero-torque boundary, so the α inside each factor cancels identically.
Step 4. Assemble the pieces using Ω_K² = GM/R³ to get D(R) = (3GMṀ/8πR³)[1 − (R_in/R)^(1/2)]. Why: every symbol is fundamental (G), a system parameter (M, Ṁ), or geometry (R, R_in) — the microphysics has vanished.
Step 5. Set D(R) = σ T_eff⁴ (optically thick face) and solve: T_eff = {(3GMṀ/8πσR³)[1 − (R_in/R)^(1/2)]}^(1/4). Far out the bracket → 1, giving the R^(−3/4) law. Why: optical thickness turns a dissipation rate into a surface temperature, letting the disk announce its structure through a spectrum. *Verified: reproduces Frank, King & Raine Eq. (5.44) and Carroll & Ostlie §18; the α-independence and the R^(−3/4) asymptote are standard textbook results.*
Go deeper · Derivation B · 1 minAccretion luminosity and the missing half
The goal is to integrate the disk dissipation over its entire surface and show that the total luminosity is exactly GMṀ/2R_in — precisely half the gravitational potential energy released as gas descends from infinity to the inner edge. We set up the surface integral over both faces, insert the dissipation profile derived above, evaluate the two elementary radial integrals, and read the surviving factor of one-half through the virial theorem. The result is again independent of α and even of the detailed temperature profile — only the total energy budget matters.
Step 1. Write L = ∫ D(R) (2)(2πR) dR from R_in to infinity, the 2 counting both faces and 2πR dR the annulus area. Why: energy conservation lets us total the escaping radiation ring by ring without tracking how it got there.
Step 2. Insert D(R), collapsing the integrand to L = (3GMṀ/2) ∫ [R^(−2) − R_in^(1/2) R^(−5/2)] dR. Why: the boundary bracket splits it into two clean power-law terms.
Step 3. Evaluate: ∫ R^(−2) dR = 1/R_in and ∫ R^(−5/2) dR = (2/3) R_in^(−3/2), both convergent at infinity. Why: convergence confirms nearly all the luminosity comes from the innermost few R_in, where the gas is deepest in the well.
Step 4. Combine: L = (3GMṀ/2)[1/R_in − (2/3)(1/R_in)] = GMṀ/2R_in. The other half of GMṀ/R_in remains as orbital kinetic energy (1/2)v_φ² at R_in. Why: by the virial theorem a circular orbit stores kinetic energy equal to half its potential energy; the disk radiates the dissipated half and carries the orbital half across the boundary — released later in a stellar boundary layer or lost past a black hole's horizon. *Verified: standard result (Frank, King & Raine §4.5; Carroll & Ostlie §18). Efficiencies η range from ~10⁻⁴ for a white dwarf to ~0.06 at the Schwarzschild ISCO to ~0.42 for extreme Kerr.*
Go deeper · Derivation C · 1 minThe multicolor blackbody spectrum
The goal is to show why a nested set of blackbody rings with T ∝ R^(−3/4) emits, over an intermediate band, a spectrum with the distinctive slope F_ν ∝ ν^(1/3). We write the flux as an area-weighted sum of Planck functions, change the integration variable from radius to local temperature using the R^(−3/4) profile, and rescale the frequency out of the integral to expose the power law. The trick: between the coolest and hottest rings the temperature integral becomes a dimensionless constant, leaving all the frequency dependence in the prefactor.
Step 1. Write F_ν ∝ ∫ B_ν(T(R)) 2πR dR, an area-weighted sum of Planck functions B_ν ∝ ν³/(e^(hν/kT) − 1). Why: each optically thick ring is a local blackbody, so the disk spectrum is a superposition, not a single Planck curve.
Step 2. Use T ∝ R^(−3/4) to change variables: R ∝ T^(−4/3) and R dR ∝ T^(−11/3) dT. Why: converting radius to temperature turns a geometric integral into one over the quantity that sets each ring's spectrum.
Step 3. Substitute for F_ν ∝ ν³ ∫ T^(−11/3) / (e^(hν/kT) − 1) dT, then let x = hν/kT so T ∝ ν/x and dT ∝ ν dx/x². Why: pulling the frequency out through x separates the ν-scaling from a pure number.
Step 4. Collecting powers of ν gives F_ν ∝ ν³ · ν^(−8/3) = ν^(1/3), valid for kT_out ≪ hν ≪ kT_in where the x-integral is a constant. Why: below this band lies the Rayleigh–Jeans tail (ν²) of the coolest ring, above it the Wien cutoff of the hottest — the ν^(1/3) plateau is the disk's signature in between. *Verified: the ν^(1/3) intermediate slope (equivalently νL_ν ∝ ν^(4/3)) matches Frank, King & Raine §5.5 and the observed "diskbb" spectra of X-ray binaries and the AGN big blue bump.*
Limiting cases
The single-α closure and thin-disk assumptions carve out a limited but powerful regime; pushing on any assumption reveals a distinct zone.
| Case | Result | Meaning |
|---|---|---|
| Gas-pressure "middle" region | P_gas ≫ P_rad, κ = κ_es | The classic Shakura–Sunyaev region: electron-scattering opacity, gas pressure dominant, thermally and viscously stable. Applies at moderate Ṁ. |
| Radiation-pressure inner region | P_rad ≫ P_gas ⇒ H nearly constant | At high Ṁ near the center, radiation pressure inflates the disk. With stress ∝ αP_rad it becomes viscously and thermally unstable (the Lightman–Eardley instability). |
| Outer, opacity-transition region | P_gas ≫ P_rad, κ = κ_ff ∝ ρT^(−7/2) | Free–free (Kramers) opacity dominates; Σ and T follow different power laws. Hydrogen-ionization opacity here drives the dwarf-nova limit cycle. |
| Thin-disk validity | H/R ≃ c_s/v_φ ≪ 1 | Requires efficient cooling. Fails at low Ṁ (advection-dominated, geometrically thick) and at super-Eddington Ṁ (slim/thick disks). |
| Near the ISCO | Newtonian T_eff(R) breaks down | General relativity modifies the potential, the shear, and the location of the zero-torque boundary; a relativistic correction factor is needed. |
Worked examples
1 · A stellar-mass black-hole X-ray binary (fully worked)
A 10 M_⊙ black hole accretes at Ṁ = 10¹⁸ g s⁻¹ (about 1.6 × 10⁻⁸ M_⊙ yr⁻¹). Take the inner edge at the Schwarzschild innermost stable circular orbit, R_in = 6GM/c². Find R_in, the disk luminosity, the radiative efficiency, the Eddington ratio, and the peak disk temperature.
Step 1. Gravitational radius: r_g = GM/c² = (6.674 × 10⁻⁸)(10 × 1.989 × 10³³)/(2.998 × 10¹⁰)² = 1.48 × 10⁶ cm. The ISCO is R_in = 6 r_g = 8.86 × 10⁶ cm ≈ 89 km.
Step 2. Luminosity from equation (6): L = GMṀ/2R_in = (6.674 × 10⁻⁸)(1.989 × 10³⁴)(10¹⁸)/(2 × 8.86 × 10⁶) = 7.5 × 10³⁷ erg s⁻¹.
Step 3. Efficiency: η = L/(Ṁc²) = 0.083, the Newtonian thin-disk value r_g/2R_in = 1/12; the full Schwarzschild treatment lowers it to ≈ 0.057 because gravity is stronger than Newtonian near the ISCO.
Step 4. Eddington luminosity: L_Edd ≈ 1.26 × 10³⁸ (M/M_⊙) = 1.26 × 10³⁹ erg s⁻¹, so L/L_Edd ≈ 0.06 — comfortably sub-Eddington, so the thin-disk assumption is self-consistent.
Step 5. Peak temperature: evaluate equation (5) at R = (49/36)R_in, where (R_in/R)^(1/2) = 6/7 so the bracket is 1/7 and R³ = (49/36)³R_in³. With 3GMṀ/8πR_in³ = 2.3 × 10²³ erg cm⁻² s⁻¹, this gives σT⁴ = 1.3 × 10²², so T_peak ≈ 3.9 × 10⁶ K, i.e. kT ≈ 0.34 keV. The inner disk radiates in soft X-rays — exactly the thermal "diskbb" component fitted in black-hole binary spectra, whose L ∝ T⁴ scaling at fixed R_in is a direct test of the model.
2 · Why AGN glow in the ultraviolet, not X-rays (scaling worked)
A supermassive black hole of M = 10⁸ M_⊙ accretes at 10% of Eddington. Estimate its peak disk temperature and show it lands in the ultraviolet — the "big blue bump."
Step 1. The ISCO scales with mass: R_in = 6GM/c² = 8.9 × 10¹³ cm, ten million times larger than in Example 1. Setting L = 0.1 L_Edd = 1.26 × 10⁴⁵ erg s⁻¹ and inverting equation (6) gives Ṁ ≈ 1.7 × 10²⁵ g s⁻¹ ≈ 0.27 M_⊙ yr⁻¹.
Step 2. Repeating Example 1's peak-temperature calculation gives T_peak ≈ 7.9 × 10⁴ K, kT ≈ 6.8 eV. That places the bulk of the disk's emission in the far-ultraviolet — the big blue bump seen in quasar spectra, with the predicted F_ν ∝ ν^(1/3) rise on its low-frequency side.
Step 3. The shortcut is the scaling. At fixed Eddington ratio T_in ∝ M^(−1/4)Ṁ^(1/4) ∝ M^(−1/4) (since Ṁ_Edd ∝ M). Going from 10 to 10⁸ M_⊙ is a factor 10⁷ in mass, hence (10⁷)^(1/4) ≈ 56× cooler: 3.9 × 10⁶ K / 56 ≈ 7 × 10⁴ K, matching Step 2. The same physics puts a 10 M_⊙ hole's disk in soft X-rays and a 10⁸ M_⊙ hole's disk in the UV — purely because bigger holes have larger, cooler inner disks.
3 · How long does a protoplanetary disk live? (order of magnitude)
A solar-mass young star has a disk with H/R ≈ 0.05 and α ≈ 0.01; estimate the viscous time at 30 AU. From equation (8), t_visc ∼ α⁻¹(H/R)⁻²Ω_K⁻¹ with Ω_K⁻¹ ≈ 8 × 10⁸ s gives ≈ 3 × 10¹³ s ≈ 1 × 10⁶ yr — matching observed disk lifetimes of a few million years, and showing why the same α-framework, invented for black holes, applies across nine orders of magnitude in central mass.
Physics of the portrait
The α-disk is where a practical confession — "we don't know the viscosity, so name it and press on" — became one of the field's most productive ideas.
Origins · 1 minalpha-disk model and how this was discovered
The alpha-disk model was born from a Soviet blend of boldness and pragmatism. In 1973, at the Institute for Applied Mathematics in Moscow, the graduate student Nikolai Shakura and his advisor Rashid Sunyaev confronted the central embarrassment of accretion theory: everyone agreed the gas must lose angular momentum through some viscous friction, and everyone agreed molecular viscosity was far too feeble. Rather than wait for a theory of the turbulence, they parameterized it — shear stress equals α times pressure, α of order unity or less — and computed everything that one assumption allowed. Their paper, "Black holes in binary systems: observational appearance," derived the temperature profile, spectrum, and luminosity and predicted that accreting black holes would shine as X-ray sources — appearing just as the first galactic X-ray binaries were discovered, and the model fit. Lynden-Bell had already argued in 1969 that quasars were disk accretion onto supermassive black holes, and Pringle's 1981 review turned the α-disk into the standard framework via its time-dependent evolution equation. For eighteen years the origin of α stayed a mystery, until in 1991 Balbus and Hawley showed that a weak magnetic field in a differentially rotating disk is violently unstable — the magnetorotational instability — providing exactly the outward transport the α prescription had assumed. Shakura's placeholder had a name at last.
Going further
Texts. Shakura N. I. & Sunyaev R. A. (1973), Astronomy & Astrophysics 24, 337 — "Black holes in binary systems. Observational appearance" · Frank J., King A. & Raine D., Accretion Power in Astrophysics, 3rd ed., Ch. 5 · Carroll B. W. & Ostlie D. A., An Introduction to Modern Astrophysics, 2nd ed., §18 · Pringle J. E. (1981), Annual Review of Astronomy & Astrophysics 19, 137 · Balbus S. A. & Hawley J. F. (1991), Astrophysical Journal 376, 214 · Binney J. & Tremaine S., Galactic Dynamics, 2nd ed. (viscous disk dynamics).
Provenance. The structure equations follow Shakura & Sunyaev (1973) and Frank, King & Raine Ch. 5, cross-checked against Carroll & Ostlie §18. A verifier should recompute: (1) νΣ = (Ṁ/3π)[1 − (R_in/R)^(1/2)] from angular-momentum conservation with a zero-torque boundary; (2) L = GMṀ/2R_in from integrating D(R) over both faces; (3) the maximum at R = (49/36)R_in via dT_eff/dR = 0; (4) the F_ν ∝ ν^(1/3) slope from summing blackbodies with T ∝ R^(−3/4). Worked-example numbers (X-ray binary L = 7.5 × 10³⁷ erg s⁻¹, T_peak ≈ 3.9 × 10⁶ K; AGN T_peak ≈ 8 × 10⁴ K; protostellar t_visc ≈ 10⁶ yr) follow from the boxed equations. The α values (≈ 0.01 quiescent, ≈ 0.1–0.4 outburst) are standard observational estimates.
Prerequisites. Keplerian Orbits and Angular Momentum · Hydrostatic Equilibrium · Blackbody Radiation and the Stefan–Boltzmann Law · Viscous Fluid Dynamics. Leads to. Magnetorotational Instability · Advection-Dominated Accretion Flows · Eddington Luminosity.
Ask the tutor: Ask me to derive the νΣ angular-momentum result from scratch, or to show how the multicolor spectrum's ν^(1/3) slope emerges from summing blackbody rings — line by line.
Why doesn't the disk temperature depend on α if the whole model is about viscosity?
Because the temperature is set by an energy balance, not a rate balance. In steady state the same Ṁ crosses every radius, and energy conservation says the gravitational power liberated in each annulus is radiated there — a power that depends on Ṁ and the local orbit, not on how the transport happens. Formally, the dissipation depends on ν and Σ only through the product νΣ = (Ṁ/3π)[1 − (R_in/R)^(1/2)]. A larger α speeds the inflow and thins the disk, lowering Σ by the same factor it raises ν, so νΣ — and hence T_eff — is untouched. What α changes is Σ, v_R, and the evolution time, not the light from a steady disk.
What is the difference between "geometrically thin" and "optically thin"?
They are unrelated, and the standard disk is the first without being the second. "Geometrically thin" means the vertical thickness is small compared with the radius, H ≪ R, following from efficient cooling that keeps the sound speed far below the orbital speed. "Optically thin" would mean a photon crosses the disk vertically without being absorbed. The Shakura–Sunyaev disk is geometrically thin but optically thick: a midplane photon is absorbed and re-emitted many times before escaping a face. That optical thickness is essential — it is what lets each annulus radiate as a local blackbody at T_eff, giving the model its clean spectrum. A disk that is both geometrically and optically thin is a different animal, radiatively inefficient and hot.
Why is the disk cold at its inner edge instead of hottest there?
Because of the zero-torque inner boundary condition. At the inner edge there is nothing interior to exert a viscous torque back on the gas, so the torque — and the local dissipation — must vanish. That is the origin of the factor [1 − (R_in/R)^(1/2)], zero at R = R_in. The competition between this rising boundary factor and the falling R^(−3) gravitational dissipation produces a maximum at R = (49/36)R_in ≈ 1.36 R_in. The gas is still hottest near the center, just not exactly at the edge; far out the factor approaches one and the familiar T_eff ∝ R^(−3/4) law takes over.
What actually is α, physically?
For eighteen years, nobody knew — it was an honest placeholder for "whatever transports angular momentum." The breakthrough came in 1991 when Balbus and Hawley resurrected a 1950s result of Velikhov and Chandrasekhar: a differentially rotating, conducting fluid threaded by even a weak magnetic field is linearly unstable. The field lines act like springs between fluid elements at different radii; the inner (faster) element is braked and the outer flung out, amplifying the field and driving turbulence. This magnetorotational instability produces magnetic and velocity fluctuations whose averaged stresses behave just like an α-viscosity, with simulations giving α ≈ 0.01–0.03. So α is the vertically averaged, pressure-normalized turbulent stress — mostly magnetic in origin.
Why does the disk radiate only half the gravitational energy?
Integrating the dissipation gives L = GMṀ/2R_in, exactly half of the potential energy GMṀ/R_in released in dropping the gas from infinity to R_in. The missing half is not lost — it is the orbital kinetic energy the gas still carries as it crosses the inner edge. The virial theorem guarantees the split: a circular orbit stores kinetic energy equal to half the magnitude of its potential energy. What happens to that half depends on the accretor. On a white dwarf or neutron star the gas is decelerated in a boundary layer that can radiate it (sometimes outshining the disk). For a black hole, the gas carries its kinetic energy across the event horizon and it is lost — one reason black-hole efficiencies are quoted for the disk alone.
When does the thin-disk model break down, and what replaces it?
Whenever H/R stops being small. At high rates near Eddington, radiation pressure inflates the inner disk; the stress-∝-radiation-pressure branch is thermally and viscously unstable (Lightman–Eardley), and the disk puffs into a "slim," super-Eddington thick disk. At low rates the gas is too tenuous to cool within an inflow time, so the heat is advected inward rather than radiated — a hot, thick, radiatively inefficient accretion flow (ADAF), the model for quiescent black holes like Sgr A*. And near a black hole's ISCO, Newtonian gravity fails and the profile, shear, and boundary location need general-relativistic corrections (the Novikov–Thorne disk). The thin α-disk is the well-behaved middle of a richer family.