43

8. ONE-DIMENSIONAL STEADY FLOW Energy conservation, otherwise known as

Bernoulli, has
We found in Chapter 7 that the sound speed, cs , is the
speed at which a signal propagates in a fluid. Because γ p 1 2 1
+ v = h + v 2 = constant (8.4)
of this, there are important differences between sub- γ−1ρ 2 2
sonic and supersonic flow.
Although most flows are (at least!) two- (recall h = e + p/ρ is the enthalpy). This becomes
dimensional, we can learn a lot from the simpler case of
1 1
1D flows. These are often called channel flows; think h1 + v12 = h2 + v22 (8.5)
of a firehose, the Alaska pipeline, or a wind tunnel. If 2 2
you’re an astrophysicist, you can think about jets (from Energy conservation with heat addition some-
protostars, or accretion disks around black holes). An-
times appears in 1D problems. In this case, Bernoulli is
other astrophysical application is to spherically sym-
modified as
metric flows (such as the solar wind), which are effec-
tively one-dimensional. 1 1
h1 + v12 + q12 = h2 + v22 (8.6)
2 2
A. Two-point Connections in Steady Flow
if some quantity q12 of heat (per gram) is added be-
One approach, especially in the engineering-type lit- tween points 1 and 2. This could describe, for instance,
erature, is to connect flow properties “here” to those to a turbojet, in which ignition of a (hopefull tiny!)
“there”. amount of fuel adds heat to the flow, and accelerates
It’s useful in this game to remember the Adiabatic it (if the flow is subsonic – more on that in the home-
Relations. Steady flows are often assumed to be adia- work).
batic, with a specific heat γ. This gives us useful scaling Two specific reference points in the flow are some-
laws: times useful.
csf 2
   (γ−1)/γ  γ−1
Tf pf ρf The stagnation point is just that, a point on a stream-
= = = (8.1) line where the velocity goes to zero. The subscript
T cs p ρ
“o” may be used here. One refers, for instance, to ho ,
In the above set of expressions, the subscript “f” can the “stagnation enthalpy”. This is often connected to a
refer to any fiducial point (“there”) at which we know “reservoir” – picture a large tub of gas, from which a
T , cs , p or ρ; the unsubscripted quantities then refer to flow emerges through a nozzle.
“here”.
Some useful relations here start from
To be specific, now, consider 1D flow in a channel,
or pipe, of cross-section A. To sound more formal, let γ−1 2
c2s + v = c2so (8.7)
“here” be region 1, and “there” be region 2 (or vice 2
versa, of course). We will want to apply two conserva-
(this is just Bernoulli again, right?), which can also be
tion laws.
written,
To γ−1 2
=1+ M (8.8)
v v T 2
1 2
From this, the ratios ρ/ρo , p/po , come simply from the
region 1 region 2 adiabatic relations (eqns 8.1).
Figure 8.1. A simple 1D “channel”, which has
cross-section area A. We want to connect conditions in
region 1 (ρ1 , T1 , p1 , v1 ) to those in region 2 (ρ2 , T2 , p2 , v2 ). reservoir outflow

Mass conservation has "o"

ρvA = constant (8.2)

Figure 8.2. A 1D flow from a “reservoir” out into the
In our current application, this becomes universe. The gas in the reservoir is assumed to have zero
velocity, and carries subscript o: its conditions are
ρ1 v1 A1 = ρ2 v2 A2 (8.3) ρo , p o , T o .
 44

The sonic point is another useful reference – the

point where v = cs . The subscripts “*” or “s” are used
here. The sonic and stagnation points can be connected
simply:
 (γ/γ−1)
T∗ 2 p∗ 2
= ; = ;
To γ+1 po γ+1
 1/γ−1) (8.9)
ρ∗ 2
=
ρo γ+1 Figure 8.3. Nozzles and diffusers in subsonic and
supersonic regimes. A converging nozzle accelerates
Mass conservation can be used to derive another useful subsonic flow, and decelerates supersonic flow. A diverging
relation, between the sonic point and any general point nozzle decelerates subsonic flow, and accelerates supersonic
in the flow with area A and Mach number M: flow. From Anderson Figure 5.3.
 2   γ+1
A 1 2 γ − 1 2 γ−1
= + M (8.10) • At supersonic speeds, however, (M > 1), a
A∗ M2 γ + 1 γ + 1
decrease of area leads to a decrease of speed,
B. One-dimensional channel flow and conversely. Thus, a supersonic nozzle must
be divergeant; and a supersonic diffuser myst be
Another approach is to consider the structure of the flow convergeant. The reason for this can be noted
throught the channel. In particular, consider flow in a from the force equation, which can be written
channel of area A; work in the limit where the flow is dρ/ρ = −M2 dv/v so that, for supersonic flows,
1D, that is ignore cross-channel variation. We’re partic- the density decreases faster than the velocity in-
ularly interested in the effect that variations in A have creases; so the area in an accelerating flow must
on the flow properties. increase, in order to keep the product Aρv con-
The steady continuity and momentum equations are stant.
2. A SMOOTH TRANSITION ?
1 dρ 1 dv 1 dA
+ + =0 (8.11)
ρ dx v dx A dx What about connecting these? It is possible to accel-
erate a flow smoothly through the sound speed, by a
and
proper combination of converging and diverging chan-
dv dp nels. Figure 8.2 shows an example of this. By consider-
ρv + =0 (8.12) ing (8.13), we can see that such flow must reach M = 1
dx dx
at the throat (where dA/dx = 0). Such a configuration
Using c2s = dp/dρ, these combine as is called a supersonic nozzle, a converging-diverging

1 dv 1 dA nozzle, or a de Laval nozzle.1
M2 − 1 = (8.13)
v dx A dx
which is the basic, governing equation.
1. NOZZLES AND DIFFUSERS

This leads to a striking result: the behavior of the flow

in a channel depends on whether it is subsonic or super-
sonic. In particular, compare the flows in converging
and diverging channels:
• At subsonic speeds (M < 1), a decrease of
area increases the flow speed. A subsonic noz-
zle (which accelerates the flow) must therefore 1
The last name is for Carl de Laval, a Swedish inventor who devel-
have a convergeant profile, and a subsonic diffuser oped such a nozzle, in 1883; he was best known for developing
(which decelerates the flow) must have a diver- a high-speed turbine for driving a cream separator. Astronomers
gent profile. Compressibility does not change this may be familiar with the name as a model for the formation of
qualitative behavior. radio jets by pressure confinement in active galactic nuclei.
 45

Figure 8.4. Transonic flow in a a convergent-divergent We also apply energy conservation, assuming an adia-
nozzle. From Thompson Figure 6.3 batic shock:

We must note, however, that M = 1 is not nec- γ p1 1 2 γ p2 1 2

+ v1 = + v (8.16)
essarily reached at the throat. Other combinations of γ − 1 ρ1 2 γ − 1 ρ2 2 2
initial/boundary conditions allow other flows. Equation
(8.13) also allows smooth solutions where the velocity Now, these equations can be used to express the three
v is an extremum: everywhere subsonic flow can have a post-shock quantities, ρ2 , v2 and p2 , in terms of their
maximum of v at the throat, and everywhere supersonic pre-shock counterparts. Working through the algebra
flow can have a minimum. Figure 8.5 illustrates this. (see chapter 9) gives the normal shock jump conditions:

ρ1 γ −1 1 2
= + 2
ρ2 γ +1 M γ +1
p2 2γM2 − (γ − 1)
= (8.17)
p1 γ +1
v2 γ −1 1 2
= +
v1 γ + 1 M2 γ + 1

Once you have these, the jumps in T and cS can also be

derived from the usual adiabatic and ideal gas relations.
Figure 8.5. More convergent-divergent nozzles, in which In a nutshell: a shock decelerates the flow (v2 < v1 ).
the velocity does not reach M = 1 at the throat. From The density must increase, due to mass conservation
Kundu figure 15.7 (ρ2 > ρ1 ). The “lost” kinetic energy goes to heat: T2 >
T1 and p2 > p1 .
3. NORMAL SHOCKS And a caveat: equations (8.17) hold for the rest
frame of the shock. In many applications,however, the
Most flows that go between subsonic and supersonic shock moves through the fluid. If this is the case, you’ll
aren’t so smooth, however (in fact it takes very special need to remember to do a Galilean transform to the
conditions to achieve the smooth, de Laval transition of shock-rest frame before applying the jump conditions.
Figure 8.4). More typically, mismatched variations in
channel area and/or external pressure lead to shocks in 4. 1D FLOWS WITH INTERNAL SHOCKS
the flow. Chapter 9 is devoted to shock properties; but
we also need some short discussion here. Now, let’s return to 1D flows, and consider what hap-
pens if smooth flow is not reached.
In brief, a shock is a discontinuous transition in the
flow. If you force a supersonic flow to change suddenly The pressure jump maintained across the nozzle can
(compared to the sound travel time), a shock forms at determine the flow possibilities. From (8.12), the sign
the right place and strength to accomodate the change. of dp is opposite to the sign of dv; accelerating flow
has a pressure drop, and vice versa. Consider, say, a
Here we’ll look at normal shocks, which means nozzle connecting an input region of pressure po , and
shocks normal to the flow.2 Put yourself in a frame an exit at pe . What happens? (I follow Anderson in this
in which the shock is at rest; let “1” be conditions up-
discussion; refer to Figure 8.6.)
stream (flow coming into the shock) and “2” be condi-
tions downstream. (a) For pe = po , there is no flow of course. Let pe
drop a bit (say to pa in the figure): the flow is com-
We apply mass and momentum conservation across pletely subsonic, with highest M and lowest p at the
the shock. These two conditions become throat.
ρ1 v1 = ρ2 v2 (8.14) (b) At some specific value of pe , the flow reaches
sonic at the throat. It remains subsonic everywhere
and else, however (pb is not small enough to allow a smooth
sonic transition); this is called choked flow.
ρ1 v12 + p1 = ρ2 v22 + p2 (8.15)
(c) As the exit pressure is further reduced, a region
of supersonic flow occurs downstream of the throat.
2
No, there are no “abnormal” shocks, sorry. Because it can’t continue so all the way to the exit, a
 46

normal shock forms in the diverging nozzle; the flow flow. In particular, a smooth transition from subsonic to
slows down to subsonic at the shock. supersonic flow is possible if the gas stays hot enough
(d) For some specific value of pe , the shock is lo- (extended heating sources are required).
cated exactly at the exit. The fully isentropic, subsonic- The basic solution is due to Parker. Consider a
supersonic flow pattern now exists throughout the entire steady, spherical outflow. Mass conservation in this
duct, except at the exit. case is ρvr 2 =constant; or,
1 dρ 1 dv 2
+ + =0 (8.18)
ρ dr v dr r
while the momentum equation becomes in this case
(noting that gravity from the central star is important),
dv dp GM
ρv + = −ρ 2 (8.19)
dr dr r
Writing dp/dr = c2s dρ/dr, these two equations com-
bine to give the basic wind equation,
c2s dv 2c2
 
GM
v− = s − 2 (8.20)
v dr r r
This does not have analytic solutions over the whole
range of r. It must be solved numerically; examples are
shown in Figure 8.6. However, as with our analysis of
channel flow, we can learn a lot by simple inspection of
(8.20).

Figure 8.6. The effect of the downstream pressure on a

convergent-divergent nozzle. See text for details. From
Kundu figure 15.11

(e) As pe is reduced further, the normal shock is re-

placed by oblique shocks emanating from the edge of
Figure 8.7. Possible solutions of the wind equation,
the nozzle. This is overexpanded nozzle flow.
(8.20). Initial or boundary conditions determine which of
(f) Finally, there is one value of pe that is just right, these solutions will exist in a particular system. The heavy
exactly matching that required by the isentropic flow lines show the two interesting solutions, which pass through
solution. No shocks of any kind exist here, the flow a sonic point (at rs ) and connect smoothly between r → 0
exits quietly into the post-nozzle region. and r → ∞. The outflow solution represents a stellar wind,
and the inflow solution represents steady, spherical
(g) At still lower values of pe , expansion waves will accretion. From Frank, King & Rraine, Figure 2.1.
emanate from thet edge of the nozzle. This is an under-
expanded nozzle flow. (i) The left hand side contains a zero, at v 2 = c2s . If
we want to consider well-behaved flows, that is to say
C. Spherical stellar wind flow those in which the derivative dv/dr does not blow up,
then the right hand side of (8.20) must go to zero at the
A different example of one-dimensional “channel” flow
same point. This defines the condition that must be met
is spherical outflow from a central mass: a stellar wind.
at the sonic point:
Consider a spherical expansion, driven against gravity
by a steady mass and energy source at the origin. In this GM
v 2 = c2s at r = rs = (8.21)
case, the gravity plays the role of the area in channel 2c2s
 47

S2 11111111111111
00000000000000
Whether or not a particular flow satisfies this condition C
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
depends on the starting conditions, such as with what 00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
S1 00000000000000
11111111111111
velocity and temperature it left the stellar surface, and 00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
also what the boundary conditions at large distances 00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
a 00000000000000
11111111111111
are. If it does not start in such a way to satisfy this con- 00000000000000
11111111111111
00000000000000
11111111111111
R(t)
00000000000000
11111111111111
b 00000000000000
11111111111111
dition, it either stays subsonic (corresponding to finite c
d
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
00000000000000
11111111111111
pressure at infinity), or cannot establish a steady flow. 00000000000000
11111111111111
00000000000000
11111111111111

(ii) The solution beyond the sonic point depends on Figure 8.8. Cartoon of the structure of a stellar wind,
and its interaction with the ISM. Left: the shock structure
the temperature structure of the wind. The only solu- within the wind. Right: The outer shell of dense,
tions with dv/dr > 0 for r > rs are those for which snowplowed ISM. From Dyson & Williams figures 7.3 and
c2s (r) drops off more slowly than 1/r; it is only these 7.4.
for which the right-hand side stays positive. In the case
of an isothermal wind, with c2s = constant, (8.20) can
S1 be the inner shock; region “b” be the wind-gas which
be solved in the limit r >> rs : has been through the shock; C be the contact surface be-
v 2 (r) ≃ 4c2s ln r = constant (8.22) tween the wind and the ISM; region “c” be the shocked
ISM; and S2 be the outer shock (moving into the ISM).
Thus, the wind will be supersonic, by a factor of a We expect S1 to be an adiabatic shock (since the wind
few, as r → ∞. The question of how the solar wind is probably hot and low density, and thus will have a
manages to stay nearly isothermal is not solved; it is long cooling time); region “b” will contain hot, shocked
probably due to energy transport by some sort of waves 2
3 mvwind
wind, with Tb ∼ 16 kB ∼ several × 107 K (noting
(MHD or plasma waves, for instance) which are gener-
1
ated in the photosphere and damped somewhere far out that m = 2 mp is the mean mass per particle if region
in the wind. “b” is fully ionized). The outer shock will probably be
isothermal, since the ISM is denser and cooler than the
(iii) Inside the sonic point, the gravity term will
wind. Thus, the shocked ISM will be in a thin shell,
dominate the right hand side of (8.20). Thus, solutions
containing all of the original ISM that lay between S2
with dv/dr > 0 , and v 2 < c2s , will obey
and the star.
dv GM We can’t say anything about the details of the shock
v ≃− 2
dr r transitions until we know more about shocks ... and
This equation looks as if gravity is driving the wind out! that’s next.
This unlikely-looking result comes from the fact that
the flow is nearly subsonic in this region; therefore, the
dp/dr term in (8.19) – which actually drives the wind
out – is nearly equal to the gravity term. References
What happens at the outer edge of such a wind? Re- The basic 1D flow material can be found in sev-
member that the wind is not flowing into vacuum, but eral places; I’ve followed Thompson, Kundu and An-
rather into some external medium (call it the ISM, for derson, mostly. The more specialized spherical-wind
interstellar medium: this theory has been developed to applications are found in, for instance, Frank, King &
describe the solar wind). The pressure in the wind is Raine, Accretion Power in Astrophysics; and Dyson &
dropping with radius (because the density is dropping, Williams, Physics of the Interstellar Medium.
right?). When the wind pressure is close to the am-
bient pressure, the wind must slow down. In fact, we
expect a two-shock structure. At this outer boundary
of the wind, we expect some sort of shock transition,
since the wind is supersonic. Past this shock, the hot,
shocked wind-gas will expand into the ISM (at about
its own sound speed, to start); as long as this expansion
is supersonic relative to the ISM, the expanding hot gas
will drive a “snowplowed” shell of ISM, and a second
shock, out into the ISM.
A cartoon of this region, at some point in time,
would be that in Figure 3.2. Let region “a” be the wind;