Astrophysical Jets E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

Astrophysical Jets
There are many circumstances in astronomy when gas
is found to circulate around a central gravitating mass.
This happens in BINARY STARS, where one component is a
compact object—white dwarf, neutron star or black hole—
and mass is transferred from the normal stellar companion.
It happens in the nuclei of galaxies, where the gravitating
object is a massive black hole and it also happens during
the earliest phases of evolution of a protostar. These gas
flows are called ACCRETION DISKS and they are frequently
accompanied by rapid outflows, launched in antiparallel
directions, roughly perpendicular to the disks. The
outflows are usually called jets and this article describes
some of what we know about them.
The first example of an astrophysical jet was
discovered by HEBER CURTIS in 1918 using the Lick
Observatory in California. He observed the galaxy known
as M87 in the Virgo cluster of galaxies and, when he
inspected the nucleus, he found a ‘curious straight ray. . .
apparently connected with the nucleus by a thin line
of matter’. This was a remarkable observation (by a
remarkable astronomer), and the jet in M87 is still one of
the most carefully studied of this class of objects (figure 1).
What we observe directly in M87 is a fairly straight
feature, some 2 kpc in length. It can be seen from long radio
wavelengths to high-energy x-rays and is found to exhibit
about eight regions of high intensity, known as ‘knots’,
along its length. The jet can be traced down to a scale size
of order 0.01 pc which is smaller than 10−5 times its total
length and less than 100 times the radius of the black hole
that has been found to lie in the nucleus of M87, and whose Figure 2. HH 111. Protostellar jets emerging from a young
mass is measured to be about 3 billion solar masses (see star—one member of a triple system observed using Hubble
M87: THE NUCLEUS AND JET. Space Telescope. The jets are over 10 light years in length. The
We know that the gas is flowing away from the bright knots are probably associated with shock waves formed
black hole in M87 because we can use very long baseline when faster moving gas runs into slower gas.
interferometry to measure its speed. What is found is
that some parts of the jet appear to be moving outward
with a speed that appears to be as much as six times faster away from us, it will appear to be faint. This explains
than the speed of light. This is actually quite a common why we are only able to detect one jet in M87 and why a
phenomenon called superluminal expansion. It does not high proportion of the brightest sources that we see exhibit
SUPERLUMINAL MOTION.
violate the special theory of relativity (which stipulates
that all material motion occurs slower than the speed Let us now turn to the knots. These are believed
of light). Instead, superluminal expansion is an illusion to be shock waves formed in the outflowing jets when,
which only requires that the gas in the jet moves towards for example, the jet speed at the source increases by an
us with a speed close to that of light. For this reason, jets amount greater than the speed of sound in the jet and
like that in M87 are often styled relativistic jets. It is quite therefore faster than the speed with which information
natural that these jets move with speeds close to that of can be transmitted by the moving jet fluid. A strong
light; after all this is the escape velocity from the central shock discontinuity must eventually form in the flow and
black hole. this turns out to be an excellent site for accelerating high-
The jet in M87 exhibits another common property of energy, relativistic electrons with energies up to 100 TeV.
these objects in that its antiparallel counter jet, which is These electrons radiate by spiralling in the magnetic field
expected to be present, has not yet been detected. This is an that permeates the jet—a process known as SYNCHROTRON
illustration of an effect called ABERRATION. When a source of RADIATION—and this is thought to be responsible for the
radiation moves at near light speed, it will appear to beam entire radio through x-ray emission.
its emission along its direction of motion. Consequently, M87’s jet is relatively weak. We know so much about
if the source happens to be approaching us, then it will it because it is so close. In the past, it was almost surely
appear to be very bright, whereas, if the source moves supplied with gaseous fuel at a far higher rate than at

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 1
Astrophysical Jets E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

Figure 1. Radio and optical maps of the jet observed emerging from the nucleus of the nearby elliptical galaxy M87. The jet can be
traced down to a distance smaller than 100 times the size of the central black hole.

present. It would then have been classified as a QUASAR. few local examples, it has been possible to detect very-
Quasars are the hyperactive nuclei of very distant galaxies high-energy γ -rays with TeV energies that are probably
that can outshine their stellar hosts at optical wavelengths. produced close to the central black hole.) The surprisingly
They emit prodigiously throughout the electromagnetic large intensity of these γ -ray jets tells us that, in many
spectrum. Quasars often exhibit giant, double, radio instances, they are transporting away a large fraction of the
sources—two ‘lobes’ of intense radio emission located on total power generated by the central black hole engines.
opposite sides of the galaxy. Historically, this was the way Not all jets are associated with distant galaxies. There
in which quasars were originally identified. For a long are many stellar-sized black holes that have been found in
while there was a mystery as to how these radio lobes could our Galaxy and in nearby galaxies. When these have stellar
be formed and sustained. With the discovery of relativistic companions that can lose gas, the gas will be attracted by
jets, the mystery was solved. The source of their power is the black hole and can also form an accretion disk. An
the central black hole and the accretion disk that surrounds early example was the source SS433 where two precessing
it. This creates relativistic particles and magnetic field antiparallel jets moving with speed just over a quarter the
which flow outward, along two antiparallel directions to speed of light have been found. In some instances, these
replenish the radio lobes and to propel them away from sources also produce relativistic jets, small-scale versions
the galaxy, rather like the exhausts that are associated with of the extragalactic radio sources. They exhibit giant
automobiles and jet aircraft. In some sense, the jets carry outbursts, superluminal expansion and beaming and, as
away the waste products of accretion onto a black hole. We a result, are sometimes called ‘microquasars’. These
know of thousands of these double radio sources and have are particularly good to study because astronomers can
documented their properties so that we are beginning to observe many cycles of variation in the time it takes a
understand how they fit into an evolutionary sequence. quasar to change.
The most powerful relativistic jets, that are beamed Jets have also been found associated with newly
almost directly at us, are sometimes called BLAZARS and formed stars. Here, again there is an orbiting disk
these are commonly observed as γ -ray sources. Photons, made of the gas that has not yet settled onto the star
with energies as high as several GeV, are formed when (or condensed into planets), These ‘YOUNG STELLAR OBJECTS’
an x-ray photon is scattered by a relativistic electron— are observed primarily in optical and infrared emission
a process known as inverse Compton scattering. (In a lines and the jet moves with speeds of several hundred

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 2
Fluid Dynamics E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

introduced, the linear size of the fluid element being small

Fluid Dynamics compared with L but large compared with λ. This implies
Fluid dynamics is a subfield of continuum mechanics that the number of particles in the fluid element is large,
which deals with the theory of motion of a fluid and hence mean physical quantities, like, for example, the
(hydrodynamics) or gas (gas dynamics). fluid density ρ or the fluid velocity v can be defined for
Fluid dynamics is studied theoretically, experimen- the fluid element. The velocity of the individual particles
tally and by means of numerical simulations. The is then given by u = v + w, i.e. it possesses a random
phenomena of fluid motion are governed by well known component w above the mean. Because λ
L, however,
continuum laws of physics: the conservation of mass, mo- the particles only perform a random walk about the mean
mentum and energy. Whether these continuum laws ac- motion v , and consequently the fluid element remains
curately describe a given physical system depends both well defined during the evolution except for some small
on the mechanical and thermodynamic state of the sys- loss of particles near its surface, which, if necessary, can
tem (i.e. its density, temperature, etc) and on the kind and be well described by an additional diffusion term in the
strength of the interaction of its constituents (e.g. the scat- hydrodynamic equations.
tering rate of gas particles). The matter of astrophysical Secondly, the forces between particles must be
objects (e.g. of a star or galaxy) can often be approximated saturating or ‘short range’, because otherwise collective
as a gas or fluid, i.e. the equations of fluid dynamics are effects must be taken into account. Formally this can
adequate to describe the astrophysical phenomena. Here- be expressed by the requirement that the energy per
after, for simplification, the word fluid will be used as a particle E/N must approach some constant value when the
synonym for both fluid and gas. number of particles tends to infinity. For a fluid element
The conservation laws describing the motion of a fluid with volume V one can then define an energy density
can be formulated as set of nonlinear partial differential ε ≡ E/V = n(E/N) where n ≡ N/V is the number density,
equations. Analytic solutions of this set of equations only and a pressure p = n(∂ε/∂n) − ε exerted on the ‘walls’
exist for simplified highly symmetric flow problems, and confining the fluid element. Examples of non-saturating
the mathematical theory is often difficult. For this reason, forces are gravity and electromagnetic forces which both
observations of fluid motion in nature, in the laboratory scale as (∼ r −2 ). Consequently, gravity must be included
and with computers are also essential for understanding in the hydrodynamic equations as a macroscopic external
the motion and behavior of fluids. force. Although electromagnetic forces are long range
In astrophysics the numerical tool is often the only too, they saturate in electrically neutral systems due to
one available to the researcher. Moreover, as most screening.
astrophysical conditions are inaccessible in the laboratory,
and as astrophysical fluid motion may occur on time Equations of fluid dynamics
scales long compared with the lifespan of humans or The equations of fluid dynamics can be derived by a
deep inside astrophysical objects, numerical simulation phenomenological approach or within the framework of
is the only means to study such fluid motion. The the KINETIC THEORY. In the phenomenological approach one
study of astrophysical fluid flows is further complicated postulates certain relations between stress and rate of
by the effects of self-gravity, which must be considered strain and between heat flux and temperature gradient and
in many astrophysical flow problems, by the enormous develops the fluid dynamic equations from the universal
range of length scales and time scales to be covered laws of conservation of mass, momentum and energy.
in the simulations, and by a variety of other physical The required constants of proportionality between stress
effects which must be taken into account frequently. and rate of strain and between heat flux and temperature
The latter include radiation transport (of photons and gradient are called transport coefficients and must be
neutrinos), heat conduction, radiative cooling, ionization determined experimentally.
and recombination of atoms, magnetic fields, energy The equation that results from requiring the conserva-
generation by thermonuclear reactions, flow velocities tion of mass in a fluid flow is called the continuity equation.
near the speed of light and in strong gravitational fields, The conservation of momentum law or Newton’s second
and the properties of dense matter. law applied to a fluid yields a vector equation known as the
Navier–Stokes equation. The First Law of Thermodynamics
Applicability of fluid dynamics or the conservation of energy law imply a fluid dynamic
Two requirements must be fulfilled in order for the equation which is called the energy equation.
continuum approximation of fluid dynamics to hold. In the kinetic theory approach the fluid dynamic
Firstly, one has to assume that the microscopic equations are derived from the Liouville equation by
behavior of single particles can be neglected. This requires defining the transport coefficients in term of certain
that λ
L, where λ is the collisional mean free path integral relations which involve the dynamics of colliding
of the particles and L is a characteristic macroscopic particles. The drawback to this approach is that the
linear dimension of the system, or a scale over which interparticle forces must be specified in order to evaluate
the distribution function varies significantly. If the above the collision integrals. Both the phenomenological
equation holds, the concept of a fluid element can be approach and the kinetic theory approach yield the

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 1
Fluid Dynamics E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

same fluid dynamic equations provided equivalent where

assumptions are made during their derivations. P ≡ 13 ρ |w|2

Mathematically speaking the five fluid dynamic

equations form a set of coupled nonlinear partial is the isotropic fluid (or gas) pressure
differential equations which are of first order in time and
π ≡ ρ 13 |w|2 I − w ⊗ w

of second order in space. They describe the evolution of

mass density ρ, momentum density ρ v and (total) energy
is the viscous stress tensor and I is the unit tensor.
density e in time and space. In addition to the evolution
equations, it is necessary to supply a relationship between
fluid properties in order to close the system of equations. ε ≡ 12 |w|2

The equation of state is an example of such a relationship.

is the specific internal energy (erg g−1 ), and
It relates the thermodynamic variables pressure p, density
ρ and temperature T or internal energy density ε.
h ≡ ρ w 12 |w|2

In many astrophysical applications of fluid dynamics,

additional equations and/or additional source terms have is the conduction heat flux. The operator
denotes
to be considered, which for example describe composition an average over the one-particle Boltzmann distribution
changes and energy generation due to nuclear reactions, function.
the diffusive transport of matter or radiation (diffusion
In the absence of viscosity the Navier-Stokes equation
equation), magnetic fields and their coupling to matter
reduces to the Euler equation
(magnetohydrodynamics), the evolution of radiation
fields and their interaction with matter (radiation ∂
hydrodynamics), and self-gravity. Moreover, in certain (ρ v ) + div(ρ v ⊗ v ) + grad P = −ρgrad.
∂t
applications the conservation equations have to be
formulated consistently with SPECIAL or GENERAL RELATIVITY. For the special case of an adiabatic flow the total energy
equation reduces to the simpler form
Form of the equations
Here the equations of fluid dynamics are given without ∂e
+ div[(e + P )v ] = −ρ v grad
derivation. The various quantities appearing in the ∂t
equations below are defined as follows: u = v + w is
the particle velocity, v the average fluid velocity, w the where e = 12 ρ|v |2 +ρε is the total energy density (erg cm−3 ).
statistical fluctuation around the mean flow, ρ is the fluid
density, ρ v is the fluid momentum, and ρε (erg cm−3 ) is Flow regimes
the kinetic energy density about the mean motion (i.e. the Fluid flows can be uniquely classified into different flow
internal energy), respectively. regimes depending on the value of certain characteristic
The continuity equation dimensionless flow parameters. Different flows having
the same dimensionless flow parameters are called similar.
∂ρ Depending on the degree of compressibility of a fluid
+ div(ρ v ) = 0 under forces acting on it, it can be treated as either being
∂t
incompressible or compressible. In the former case the fluid
describes the conservation of mass, the Navier–Stokes dynamic equations can be simplified considerably. If the
equation of motion state of the fluid is independent of time at every point in
space occupied by the fluid, the flow is called stationary.
∂
(ρ v ) + div(ρ v ⊗ v ) + div = −ρgrad (1) If the opposite is true, one has an unsteady flow. For
∂t stationary flows Bernoulli’s equation holds, which says that
describes the transport of (relative) momentum in the flow. the sum of the specific kinetic energy (|v |2 /2) and of the
The total energy equation (assuming that the particles have enthalpy of the fluid is constant along a streamline. A
no internal degrees of freedom) streamline is defined as a curve whose tangent vector at
each point of the curve has the same direction as the

 
  velocity field of the flow at that point.
∂ ρ 2 ρ 2
|v | + ρε + div |v | + ρε v If the fluid moves with a velocity smaller or larger
∂t 2 2
than the sound speed, the flow is called subsonic and
+ divh + div(v ) = −ρ v grad supersonic respectively. The Mach number Ma ≡ |v |/cs ,
which is the ratio of the absolute value of the flow velocity
finally describes the conservation of energy.  ≡ ρ w ⊗ w
v and the sound speed cs , is the parameter characterizing
is the pressure tensor, from which one usually separates this aspect of a flow. Transonic flows have Ma ≈ 1.
out the trace of the symmetric dyadic w ⊗ w, i.e. one writes Supersonic flows accessible in the laboratory have Mach
numbers Ma
10, while astrophysical flows can be
 = PI − π extremely supersonic (Ma > 100).

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 2
Fluid Dynamics E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

When a flow accelerates (decelerates) from subsonic If the turbulence has quantitatively the same structure in
(supersonic) to supersonic (subsonic) speeds shock waves all parts of the flow field, the turbulence is said to be
occur in the flow. A shock wave is a front across which homogeneous. If the statistical features have no preference
there is a nearly discontinuous, finite jump in pressure, for any direction, the turbulence is called isotropic. At a
with corresponding jumps in temperature, density, and given point in the turbulent domain a distinct pattern is
other fluid properties. Shocks are nonlinear waves whose repeated more or less regularly in time, and at a given
behavior is much more complex than that of (linear) instant a distinct pattern is repeated more or less regularly
sound waves, which are periodic adiabatic pressure in space. Turbulent motions in which distinct patterns can
perturbations of the flow. Because of their nonlinear be discerned are described quantitatively by introducing
character linear superposition does not apply in shock the notion of scale of turbulence: a certain scale in time and
wave interactions. Flows entering a shock at an oblique a certain scale in space. Turbulent motion, in addition to
angle are deflected away from the normal direction. The its scale, is characterized by the violence or intensity of
deflection is due to a decrease of the normal component of the turbulence fluctuations,
√ which is defined by the root-
the fluid velocity entering the wave. Hence kinetic energy mean-square value w ≡ w 2 of the random component w
is dissipated into heat in a shock wave, increasing the of the momentary value of the velocity u = v + w about the
entropy of the fluid. Conservation of mass, momentum mean flow velocity v . Often the intensity of the turbulence
and energy must also hold across shock waves. This is defined by the relative intensity w /|v |.
fact can be used to derive the Rankine–Hugoniot relations
relating the fluid states on both sides of the shock wave. Flow instabilities
One further discriminates between viscous and Fluid flows are subject to a variety of flow instabilities, the
inviscous or ideal flows depending on the amount of most common ones being the buoyancy driven Rayleigh–
viscosity of the fluid. The viscosity of a fluid varies, Taylor instability and the Kelvin–Helmholtz instability
in general, with the thermodynamic state of the fluid. occurring in shear flows.
Except for superfluids the viscosity of a fluid is never Rayleigh–Taylor instabilities occur, for example,
exactly zero, but it may be sufficiently small to consider when two fluids of different densities are superposed
the fluid for all practical purposes as being inviscous or one over the other or accelerated towards each other (see
ideal. Quantitatively the viscous character of a fluid is figure 1). According to a linear stability analysis the
determined by comparing the size of the viscous drag arrangement of two fluids is stable if the lighter fluid is
acting on the fluid with its inertia. The relevant flow on top of the heavier fluid. Otherwise the arrangement of
parameter is the Reynolds number Re ≡ |v |l/ν, where l fluids is unstable.
is a characteristic linear dimension of the flow, and ν is Rayleigh–Taylor instabilities are ubiquitous in astro-
the kinematic viscosity of the fluid. Viscous flows have physical flows. A particular example is Rayleigh–Taylor
Re < 1. Most astrophysical flows have Re  1 (Reynolds instabilities induced by the propagation of a shock wave
numbers larger than 1010 are not uncommon), because of through the envelope of a massive star which explodes as
the large size of the astrophysical objects. a (core collapse) SUPERNOVA (see figure 2). Since the energy
A common feature of terrestrial and in particular given to the shock wave in a supernova explosion is much
of astrophysical flows is their tendency to undergo a larger than the binding energy of the mass outside the col-
spontaneous transition from laminar (i.e. smooth, regular lapsed core, gravity is dynamically unimportant for the
type of flow) to turbulent flow. This occurs if the Reynolds propagation of the shock wave through the stellar enve-
number of the flow exceeds a critical value Recrit beyond lope. In this case the Rayleigh–Taylor instability is driven
which effects due to inertia dominate over friction by such by pressure and density gradients of opposite signs, the
an amount that the flow becomes unstable to turbulence. ‘effective acceleration’ being the (negative) pressure gra-
From laboratory experiments one finds critical Reynolds dient.
numbers for fully developed turbulence in the range Kelvin–Helmholtz instabilities occur whenever the
1000
Recrit
2000. velocity field of the flow has a gradient perpendicular
According to J O Hinze ‘Turbulent fluid motion is to the (local) flow direction, i.e. in the presence of shear
an irregular condition of flow in which the various fluid flow. The Kelvin–Helmholtz instability gives rise to a roll-
quantities show a random variation with time and space up of the shear layer. It is responsible for the formation
coordinates, so that statistically distinct average values of mushroom shaped flow structures seen, for example,
can be discerned’. Turbulent motion is not irregular in in large atmospheric explosions, and in connection with
either space or time alone, but both. Turbulence can be Rayleigh–Taylor instabilities (see figures 1 and 2). In the
generated by friction at fixed walls (‘wall turbulence’) or latter case the rising bubbles and sinking spikes resulting
by flow of layers of fluids with different velocities past from the Rayleigh–Taylor instability cause a shear flow
or over one another (‘free turbulence’). In real viscous with respect to the surrounding matter.
fluids, viscosity effects will result in the conversion of
kinetic energy of flow into heat, i.e. turbulent flow is Fluid dynamics and nuclear burning
dissipative in nature. Hence, if there is no continuous Much of our understanding of astronomy and astrophysics
input energy into the flow, the turbulent motion will decay. is inseparably connected to our understanding of fluid

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 3
Fluid Dynamics E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

Figure 2. Simulation of Rayleigh–Taylor instabilities induced by

the propagation of a shock wave through the envelope of a
massive star which explodes as a (core collapse) supernova. The
initially spherical helium shell (bright grey) of the progenitor
star has been shredded by the instability. The snapshot is taken
about 4 h after the onset of the explosion.

dynamics and nuclear burning. Thermonuclear processes

and gravitational binding are the two major sources of
energy for astronomical flows. The release of energy in
an astrophysical PLASMA gives rise to changing pressure
gradients and so to fluid motion. Such fluid motion, in
turn, will transport ashes and fuel to and from the flame
zone. This gives rise to a subtle and difficult problem.
The degree of difficulty of the problem is closely related
to the nature of the coupling between the nuclear burning
and the fluid flow. Special cases, in which this coupling
is simpler, can be treated more reliably. First, nuclear
reactions release energy in the form of radiation and kinetic
energy of particles; at stellar densities this energy acts to
heat the medium in which the reactions occur. The heating
can modify the pressure, and thereby modify the flow.
Second, nuclear reactions change the composition of the
material, which in turn may modify its equation of state
and its opacity, which in turn may modify the pressure
and the heating, and again the flow. Conversely, the flow
may remove ashes and bring fuel to the flame, countering
Figure 1. Rayleigh–Taylor instability in a dense fluid (bright) the natural tendency of the flame to exhaust its fuel.
supported against gravity (directed upward!) by a lighter fluid Many different types of burning occur in astrophys-
(dark). The horizontal interface between the two fluids (initially ical problems. Burning can range from quiet hydrostatic
located at 200) has been perturbed by a small sinusoidal burning, such as occurs in the center of most stars, to the
perturbation. The typical structure of a narrow, high-density explosive burning which can occur, for example, in mod-
finger penetrating into the low-density region and a broader,
els of thermonuclear supernova explosions. Burning can
low-density bubble at the edge of the grid penetrating into the
high-density region is clearly visible. The mushroom cap occur in laminar or turbulent flows, the fluid motion being
structure on the top of the high-density region, which is created produced either by external causes and/or by the burning
by the Kelvin–Helmholtz instability, is also apparent. itself. Hence, the interaction between burning and flow

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 4
Fluid Dynamics E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

must be considered. flame is subsonic, flame and flow do interact. In particular,

Hydrostatic burning is the least violent form of if burning occurs in a convective or turbulent environment,
burning and occurs in stars that are hydrostatic. In this depending on the relevant length scales and time scales,
case, the reaction rate is slow enough that energy can physically quite different combustion regimes must be
be transported from the burning region by radiation or considered.
conduction as rapidly as it is generated. A similar case
is a subsonically convective flame region, in which heat Relativistic fluid dynamics
and ashes are removed by slow convection, and new fuel Some astrophysical phenomena require a relativistic
brought in. formulation of fluid dynamics. A special relativistic
Detonations are the most violent form of burning one formulation must be used if the flow velocity reaches
encounters. Astrophysical detonations generally occur a significant fraction of the speed of light, or if the
only under degenerate conditions. In its simplest form, internal energy of the fluid becomes comparable to its
detonation wave theory is easily understood. As a shock rest mass. Such conditions occur in jets from quasars,
propagates into unburned fuel, it compresses and heats the active galactic nuclei and some galactic compact binaries,
material beyond the ignition point. Immediately behind and in gamma-ray burst sources. General relativistic
the shock is the reaction zone, in which the fuel burns. flows involving strong gravitational fields and relativistic
In simple detonation theory the width of the reaction speeds are encountered in accretion flows onto compact
zone is neglected, and the detonation front is treated as objects like neutron stars or black holes, in the formation
a sharp discontinuity. In this case jump conditions can or merging process of neutron stars and black holes, and
be derived for the change in the fluid variables across the are most likely at the heart of the processes leading to the
front in much the same way as is done for a simple shock. formation of relativistic jets and gamma-ray bursts.
Note, however, that in order to obtain the post-detonation An important difference between Newtonian and
state one must first choose a detonation velocity, because relativistic fluid dynamics is the presence of a maximum
unlike the case for simple shocks, the front velocity is velocity, i.e. the speed of light in a vacuum in the latter case.
not determined from the jump conditions. The simple A flow can be relativistic in two ways. In a kinematically
detonation theory assumes that the reaction rate is infinite relativistic jet the beam gas moves with relativistic velocity,
and that the reaction zone has zero width. A slightly while in a thermodynamically relativistic jet the specific
more complex treatment is used in the Zeldovich–von internal energy of the beam gas is large compared with its
Neumann–Doering model. This theory assumes that the rest mass energy.
shock, which is taken to be infinitely thin, is followed by a Special relativistic astrophysical flows are encoun-
reaction zone of finite width. The final state obtained after tered in extragalactic jets (see ASTROPHYSICAL JETS). These are
the fuel is completely burned is exactly the same as for the highly collimated supersonic outflows from ACTIVE GALAC-
simpler theory described above. TIC NUCLEI. Extragalactic jets are thought to be formed as
Deflagrations represent a much less violent form of a consequence of mass accretion onto a central rotating
burning than detonations, but in many ways are more super massive BLACK HOLE in an active galactic nucleus be-
complex. They result when the burning is unable to ing fed by interstellar gas and gas from tidally disrupted
produce sufficient overpressure to create a shock which stars. Most models of the (initial) collimation and acceler-
is strong enough to ignite the fuel. The motion of the ation of astrophysical jets involve magnetohydrodynamic
front is usually very subsonic. Burning is initiated by the processes at or near the inner edge of a rotating magne-
diffusive transfer of heat from the hot ashes behind the tized accretion disk with corona. The combined action of
front into the cold fuel. Although, for the case of a thin thermal and magnetic pressure gradients, of centrifugal
front, deflagrations must obey the same jump conditions force and of gravity drives an outflow, which does colli-
as detonations, the propagation velocity now depends on mate itself. General relativistic effects seem to be crucial
the rate of heat transfer. Compared to detonations another for a successful launch of the jet. A Newtonian treatment
major difference is that the pressure and density decrease of extragalactic jets is justified only at distances larger than
behind the deflagration front, and in the reference frame
several kiloparsec from the central black hole. At smaller
in which the front is stationary, the velocity increases. In
scales, and in particular at parsec scales, jets propagate
the simplest theory deflagrations are described as flow
at relativistic speeds. Hence, special relativistic hydrody-
discontinuities, which propagate subsonically. If the flow
namic simulations must be used in order to describe their
is laminar, a physically more accurate description is that
propagation correctly (see figures 3 and 4).
of a conductive or diffusive laminar flame, whose width
and speed is determined by the rate of energy generation
Equations for ideal special relativistic flows
and the transport of energy and composition. In most
The equations of special relativistic fluid dynamics for an
situations, however, even this description is too simple,
ideal fluid are most conveniently expressed as a set of
because deflagrations or subsonic flames are subject to a
conservation equations
variety of instabilities, which can significantly influence
their shape (wrinkling of the flame) and propagation ∂D
speed. Moreover, because the propagation speed of the + div(D v ) = 0
∂t

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 5
Fluid Dynamics E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

Numerical integration of the fluid dynamic

equations
In astrophysical applications one usually has to solve the
hyperbolic set of the inviscous and compressible Euler
equations instead of the parabolic set of the viscous and
compressible Navier–Stokes equations, because (except
in shock waves) viscosity and heat conduction due to
physical processes are often negligibly small. However,
because of truncation errors and because of finite spatial
resolution, all numerical methods proposed for the
solution of the Euler equations possess some finite intrinsic
numerical viscosity, whose form and size depends on
Figure 3. Color-coded contour plots of the logarithm of the the numerical method, on the grid resolution, and to
proper rest-mass density (top) and pressure (bottom) of a hot, some extent of the flow problem itself. The effect of the
relativistic jet (vb = 0.99c, Ma = 1.72). The maximum values are numerical viscosity on the solution can only be quantified
coded in white. Decreasingly smaller values are coded in green, by resolution studies, i.e. by performing a series of
bright blue, dark blue, red and the minimum values are coded in
simulations of the same flow problem using different grid
black. The jet has a ‘naked’ beam and almost completely lacks
internal structure. The beam is surrounded by lobes (near the jet resolution. Thus, when simulating astrophysical flow
head) instead of a cocoon and is in pressure equilibrium with the one is in the strange situation that one tries to solve the
shocked ambient medium. This figure is reproduced as Color inviscous Euler equations, but instead solves some viscous
Plate 20. variant, which is also different from the Navier–Stokes
equations.
The evolution equations are a set of coupled nonlinear
first-order (in time) partial differential equations. One
way of solving this set of equations is to discretize them
in time and space. By this discretization the partial
differential equations are transformed into a set of coupled
nonlinear algebraic equations, which can be solved on a
computer with appropriate numerical techniques. The
algebraic equations are not uniquely determined, because
a given differential equation can be discretized in different
ways. Obviously, one tries to choose those algebraic
systems which minimize the truncation errors, and which
possess some additional properties, like being stable and
Figure 4. Same as figure 3, but for a cold, relativistic jet consistent. This is at the heart of the ‘art of computing’.
(vb = 0.99c, Ma = 6.0). The jet has a beam with internal oblique Guided by the conservation properties underlying
shocks and an extended turbulent cocoon. This figure is the hydrodynamic equations applied mathematicians and
reproduced as Color Plate 21.
physicists have been able to develop accurate and stable
high-resolution finite volume schemes In finite volume
∂S i schemes the (finite) computational domain is discretized
+ div(S i v ) + (gradp)i = 0, i = 1, 2, 3 into a finite number of zones or cells (typically 102 to 103
∂t
∂τ per spatial dimension). The hydrodynamic scalar fields
+ div[(τ + p)v ] = 0 (e.g. density) and the components of the vector fields (e.g.
∂t
momentum) are approximated by a set of discrete values,
for the conserved rest-mass density D, momentum density
which are the (approximate) cell averages of the respective
S = (S 1 , S 2 , S 3 ) and energy density τ , respectively. These
variable. Time is discretized too, the evolution being
variables, which are defined in a fixed frame, are related
approximated by a series of finite time steps (typically 103
to quantities in the local rest frame of the fluid through
to 105 per simulation).
D = ρW Besides finite volume schemes, another completely
S = ρhc−2 W 2 v different method is widely used in astrophysics for
integrating the hydrodynamic equations. This method is
τ = ρhW 2 − p − Dc2
smoothed particle hydrodynamics, or SPH for short. In
where ρ, p, v , W , h and c are the proper rest-mass density, SPH the fluid or gas is described by a finite set of ‘particles’
the pressure, the flow velocity, the flow Lorentz factor (typically 103 to 105 ), which move with the flow. SPH is
W = (1 − v 2 /c2 )−1/2 , the specific enthalpy h = c2 + ε + p/ρ, a free-Lagrange method, in which spatial gradients are
and the speed of light in a vacuum. evaluated without the use of a grid.

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 6
Fluid Dynamics E N C Y C LO P E D IA O F A S T R O N O M Y AN D A S T R O P H Y S I C S

Bibliography
The literature on fluid dynamics is extensive. Thus, the
following references can only provide entry points into
the field.
The derivation of the fluid dynamic equations
using the phenomenological approach can be found, for
example, in the book by

Shu F H 1991 The Physics of Astrophysics (Mill Valley, CA:

University Science Books)

The kinetic theory approach to the equations of fluid

dynamics is described in detail by

Hirschfelder J O, Curtiss C F and Bird R B 1964 Molecular

Theory of Gases and Liquids (New York: Wiley)

A general introduction to fluid dynamics can be found in

Landau L D and Lifshitz E M 1982 Course of Theoretical

Physics, Vol.VI, Fluid Mechanics (Oxford: Pergamon)

For a modern description of numerical techniques for the

solution of the equations of fluid dynamics one can consult,
e.g.

Laney C B 1998 Computational Gasdynamics (Cambridge:

Cambridge University Press)

or with special emphasis on astrophysical fluid flow

including also various astrophysical applications

LeVeque R J, Mihalas D, Dorfi E A and Müller E 1998

Computational Methods for Astrophysical Fluid Flow ed
O Steiner and A Gautschy (Berlin: Springer)

Ewald Müller

Copyright © Nature Publishing Group 2001

Brunel Road, Houndmills, Basingstoke, Hampshire, RG21 6XS, UK Registered No. 785998
and Institute of Physics Publishing 2001
Dirac House, Temple Back, Bristol, BS1 6BE, UK 7