You are on page 1of 5

Turbulent viscosity?

Part I of an extensive story


The large Reynolds number argument
As we have seen in the previous lecture, typical laboratory viscosities are of the order 1cm 2s1 . From experience, we know that hydrodynamic flows with similar ratios between the inertia and the viscous forces behave in a similar way1. A measure of this ratio is the so-called Reynolds number which is defined as

inertia forces lflow vflow = viscous forces

(1)

where lflow and vflow are the characteristic length and velocity scales of the macroscopic flow2. If 1 viscosity plays essentially no role for the flow, while for 1 it dominates it. Further laboratory experience tells us that for > crit 102 3 flows become turbulent. While the critical Reynolds number crit is not a natural constant, it is a characteristic number which seems to hold in all flows observed so far. What are the Reynolds numbers of accretion disk, operating under molecular viscosity? The characteristic length scale of the flow is clearly s , the radius of the orbit about the accretor, and the characteristic velocity is the azimuthal velocity v . If we assume for the moment being Keplerian azimuthal motion, then we get for close binary (top), protostellar (middle), and AGN disks (bottom) Reynolds numbers of
1 18 2 M s 1.15 10 1cm 2s1 M 1010 cm 1 2 4.461019 M s disk = 2 1 1cm s M 1AU 1 2 6.41026 M s 1cm 2s1109 M 1pc

crit

(2)

As long as the viscosity does not differ by many orders of magnitudes from unity (in cgs units), this Reynolds number is exceedingly larger than any critical Reynolds number ever encountered anywhere. This is a strong indication of turbulence existing in accretion disks. We will see, however, that the question of whether turbulence exists or nor is not only one of the value of the Reynolds number. As we will see in the next lecture, a linear stability analysis actually argues against self-sustained turbulent viscosity in accretion disk.

We will later show the correspondence of the conservation equations, which we have derived here, with the Navier-Stokes equation. At that point we will see that the Reynolds number plays the role of a scaling quantity between different flows with identical Reynolds numbers. 2 It is important not to confuse this with the microscopic mean free path and the thermal or turbulent velocity.

The Shakura-Sunyaev parameterization of turbulent viscosity


Let us here, however, first pursue the idea of a turbulent viscosity. Such a turbulent viscosity turb would be, with lturb and vturb as the characteristic length and velocity scales of the turbulence (now it is the turbulence and not the macroscopic flow!)

1 turb = lturb vturb 3

(3)

Unfortunately, both lturb and vturb are unknown. A pragmatic solution to this problem was proposed in the early 1970s by Shakura and Sunyaev. Lacking then as today a firstprinciples-based theory of turbulent, Shakura and Sunyaev took resort to a common sense parameterization. They assumed that the viscosity in accretion disks is a turbulent one, indeed, and they made some assumptions about the involved scales: For the length scale lturb , they assumed isotropy and scaled it in terms of the largest macroscopic length scale of the disk, namely its thickness h lturb = l h (4)

where l 1 is a scaling quantity. l needs to be smaller or equal to unity as otherwise the isotropy assumption would be violated, and an additional free parameter had to be introduced. At this point, it is important to note o that by using eq. (4) actually not much was won. The uncertainty of the value of lturb was only transferred into the newly introduced quantity l ; and o that it is somewhat surprising, if not even worrisome that due to the isotropy assumption the radial transport of angular momentum couples to the vertical thickness of the disk. For the velocity scale vturb a somewhat more straightforward scaling can be made. If the turbulent velocity were supersonic, ensuing shock waves would dissipate the energy and reduce the velocity to or below the sound velocity. This leads to vturb = v cs with v 1 . Again, this parameter is unknown and replaces the unknown turbulent velocity. Equations (4) and (5) lead to a turbulent viscosity of 1 turb = l v hcs = hcs 3 (6) (5)

with = l v 3 . This is the famous or Shakura-Sunyaev viscosity. While it turned out to be a extremely successful parameterization of viscosity at least in NSG disks , it largely lacks physical background. Any parameterization makes sense only if the scaling parameter, in this case , is a constant. Given the way how eq. (6) was derived, one has no reason to assume is a constant, or only close to it. The only thing we know is 1 3 , or as often used in a somewhat sloppy way smaller than unity. As we will see later, when we discuss its application to real-world systems, surprisingly is not only either constant or varies only little, as long as self-gravity plays no role, it is also not too different from unity. In most

systems one finds usually derived from time scale observations values somewhere between 103 and 101 . In this respect, one may call the parameterization (6) a stroke of good luck and the stroke of two geniuses at the same time. Moreover, it is also important to realize that on the way to derive eq. (6) we did not specify any driving and/or sustaining mechanism for the turbulence. We will return to this question later. Thirdly, in deriving eq. (6), we made no assumptions about self-gravity. We will see later, however, that it leads to physically reasonable results only for the NSG case (which actually is the domain Shakura and Sunyaev derived it for).

The NSG case


Using eq. (6) we get for the Reynolds number, now assuming a turbulent rather than a molecular viscosity
2 v NSG 1 h 1 s turb = = = hcs h cs s

sv

(7)

where, fort he last transformation, we switched to the NSG case. Typical accretion disks have relative thicknesses h s 101 1.5 leading to Reynolds numbers turb =102 3 which, for an of order unity, is of the same order as crit . If and this only a speculation at this point the turbulent viscosity in a disk is as large as it can be, i.e., the Reynolds number is at its critical value, then the observational fact that differs not too much from unity, could have a natural explanation. The viscous time scale visc = s 2 then is
2 2 s 2 1 s s NSG 1 1 h 1 h 1 visc = = = = dyn h cs s s h s dyn

(8)

For the last three steps, we again used the NSG assumption. In principle, visc is an observable which contains information both about the disks vertical structure (via h s ) and about properties of the material (via ). If one knows for instance through theoretical models the vertical structure, one can use the viscous time scale to determine , and this is the way how this value is being determined. In the NSG case, the turbulent viscosity has another intriguing property. If we use again the vertical hydrostatic equilibrium, we get
cs2 = T 1

(9)

i.e., it is proportional to a quantity which knows about the global structure of the disk ( 2 Ms3 ) and one which is of mainly local type (temperature T ). This viscosity parameterization thus couples in the NSG case local and global disk properties. The global information comes from the hydrostatic equilibrium. This, however, is only true for NSG disks. We will see later that in cases where vertical self-gravity play a role, i.e., in the KSG and FSG cases, this is no longer the case and the viscosity knows only about the local disk structure, but not about the global one. And this will turn out to be a problem.

Accretion disk energetics The effective temperature of stationary Keplerian disks


So far, we have dealt only with the mechanical aspects of an accretion disk. The turbulent viscosity derived above, however, couples the radial disk structure also to its thermal structure. Moreover, in the hydrostatic equilibrium, the pressure term may introduce a coupling to the disks thermal structure, depending on the equation of state. For a first consideration of the energy budget of an accretion disk, we return to the torque exerted by one ring onto a neighbouring ring. Earlier, we derived this torque to be G ( s ) = 2 s s 2 (10)

If we, again, think in terms of a ring of radial extent s s + ds with ds s , we have two torques work on the ring in opposite directions, and the net torque, attempting to accelerate the ring, is G ( s + ds ) G ( s ) = G ( s ) + G G ds G ( s ) = ds s s (11)

As the ring is rotating with a frequency , a rate of working of (G s ) ds is exerted onto the ring. Integrating this in radial direction, one gets two terms of different quality
so so (G ) G so [ ] G ds = G si G ds s ds = s si si si so

(12)

The first term depends only on what is going on at the disks inner and outer boundaries. In this respect it is a global term. The integrand in the second term, however, is of local nature and describes the local loss of mechanical energy into heat. This then is the rate of dissipation of energy due to a viscous torque. For the disk of relevance is the dissipation rate per unit surface area D ( s ) , where we have to take into account that a disk has two surfaces: D ( s) = G ds 1 2 = ( s ) 22sds 2 (13)

Similar to the stellar case, we introduce an effective temperature Teff as a measure of the flux per unit area: 1 4 ( s ) = D ( s ) = ( s)2 Teff 2 (14)

is the radiation constant. In contrast to the (spherical) stellar case, the effective temperature of an accretion disk is not a constant but rather a function of the radius. For the Keplerian case (NSG & KSG) we get, using the stationary version of the radially integrated conservation of angular momentum

4 Teff

=Kepler

9 conservation of 1 M 9 GM 1 si ( ) 2 = 1 = 4 angular momentum 2 4 s3 2 s 3

3 GMM si = 1 8 s3 s

(15)

If one is not too close to the inner boundary, the radial dependence of the effective temperature of a stationary Keplerian accretion disk is
Teff s 4
3

(16)

which can lead to a considerable temperature difference between a disks radially inner and outer regions.

You might also like