Notes 12
Notes 12
Notes 12
In the last class we discussed a large number of instabilities, but we haven’t yet discussed
the most important one in most stars: convective instability. That will be the subject of
today’s class.
Convection is a process in which heat is transported by the motion of fluid elements. One
common example where convection occurs is when one heats water on a stove. Initially the
water is still, and heat is transported by conduction through it. However, as the water at
the bottom of the pot gets hotter, eventually the water starts to churn. Hot water from the
bottom of the pot rises and transports heat upwards, while cold water at the top falls. This
process is called convection. Convection is also important in planetary atmospheres, in the
liquid interiors of giant planets and in the liquid iron-rich cores of terrestrial planets.
1
!
dT T P dρ
= (γa − 1)
dr ad
P ρ dr
!
γa − 1 T dP
= ,
γa P dr
where in the last step we substituted for dρ/dr using the adiabatic equation of
state. This value of dT /dr is known as the adiabatic temperature gradient.
We can also express (dT /dr)ad using the equation of hydrostatic balance
dP Gm
=− 2 ρ
dr r
and the ideal gas law P = (R/µ)ρT . (Note that we can use the equation of
hydrostatic balance because we assume that the pressure of the rising fluid element
is the same as the pressure of its neighbors, which are in hydrostatic balance.)
Plugging in for P and dP/dr gives
! ! !
dT γa − 1 µ Gm γa − 1 µ
=− 2
=− g,
dr ad
γa R r γa R
2
where g = Gm/r is the local acceleration of gravity in the star.
An alternative form of the adiabatic temperature gradient is to give it in terms
of a logarithmic derivative of P with respect to T . Dividing both sides by dT /dr
gives !
γa T dP d ln P
= = .
γa − 1 P dT d ln T ad
Bubble
(b) (b) (b)
P f Tf ρ f
2
Now consider what happens to the rising bubble of gas if the temperature gradient
in the star is not equal to (dT /dr)ad . We start with a bubble of gas that is at the
same pressure, density, and temperature as its surroundings, and we perturb it
upward by a distance dr. It stays at the same pressure as its new surroundings,
but it is at a different temperature, and therefore a different density. If the bubble
(b) (b)
is initially at a density ρi , after it rises a distance dr its new density is ρf , where
(s)
Similarly, the initial density of the surrounding gas is ρi , and the density of the
surrounding gas a distance dr higher is
The buoyancy force is just the difference in pressure between its top and its
bottom, and is given by Archimedes principle: the buoyancy force on an object
is equal to the weight of the material it displaces. The density of the material
(s)
displaced is ρf , so the buoyancy force per unit volume is
dρ(s)
!
(s) (s)
fb = ρf g = ρi + dr g.
dr
3
Adding the gravity and buoyancy forces gives the net force,
(b)
dρ(s) dP (b)
!
ρ 1 dρ 1 dP
fnet = − i (b) g dr = − ρg dr,
dr γa Pi dr ρ dr γa P dr
where we have dropped the subscripts because everything in the second equa-
(b) (s)
tion refers to the surroundings, since P (b) = P (s) and ρi = ρi . The term in
parentheses is usually denoted by the letter A:
1 dρ 1 dP
A= − .
ρ dr γa P dr
Since we have computed the net force, we can write down the equation of motion
for the bubble. Since fnet is the force per unit volume, and ρ is the mass per
unit volume, Newton’s second law tells us that the displacement of the bubble dr
obeys
d2 fnet
2
(dr) = = Ag dr
dt ρ
This is the equation of motion for a harmonic oscillator, and it has the usual
solution:
dr = CeiN t ,
with v !
u
q u 1 dP 1 dρ
N = ± −Ag = t − g.
γa P dr ρ dr
The quantity N is the frequency of oscillation, and is known as the Brunt-Väisäla
frequency.
As we found before when considering homologous perturbations, the behavior of
the solution depends on whether the term inside square root is positive or negative,
corresponding to a real or imaginary value for N . If N is real, the solutions
are oscillations, and the system is stable. If N is imaginary, then the solutions
corresponding to an exponentially decaying and an exponentially growing mode,
and the system is unstable.
Convective instability corresponds to the case when N is imaginary. Physically,
we can understand this fairly easily. If A < 0, then the differential equation for
dr looks like a harmonic oscillator, in the sense that the force −Ag dr is opposite
to the displacement. It therefore constitutes a restoring force, which pushes the
system back to stability. The value of A in turn is determined by the balance
between gravity and buoyancy, with A < 0 corresponding to the case where
gravity is stronger. As a result we get a real value for N , and any displaced fluid
element just oscillates, bobbing up and down like a buoy in the ocean.
If A > 0, the net force is in the same direction as the displacement. Physically,
what is going on is that a blob of fluid rises and expands because it is at higher
4
pressure than its surroundings. Although gravity wants to pull it back down, its
high pressure makes it expand so much that it experiences a large buoyancy force
that is stronger than gravity. The net force is therefore upward, and the bubble
accelerates further up. This is an unstable situation, hence N is imaginary.
This physical interpretation makes sense if we examine the terms inside the square
root, and recall that dP/dr and dρ/dr are both negative. If dP/dr is very big
(in absolute value), then the system is unstable. This is because the value of
dP/dr determines how much the rising bubble expands, and thus how large the
buoyancy force is. If dρ/dr is very large (in absolute value), the system is stable.
That is because dρ/dr measures how much denser the rising bubble is than its
new surroundings, and thus how strongly gravity wants to pull it down.
Thus, we have derived the condition for stability against convection: A < 0.
Convection does not occur for A < 0, and it does for A > 0.
C. Convective Stability and the Adiabatic Temperature Gradient
We have now determined a condition for stability in terms of the gradients of P
and ρ, but it is helpful to instead phrase things in terms of temperature, because
this allows us to see how convective stability relates to the adiabatic temperature
gradient we derived a moment ago.
We use the ideal gas law P = (R/µ)ρT , which we showed earlier gives
dP P dT P dρ
= +
dr T dr ρ dr
dρ ρ dP ρ dT
= −
dr P dr T dr
for a gas of uniform composition. Substituting for dρ/dr in A gives
" #
1 ρ dP ρ dT 1 dP
A = − −
ρ P dr T dr γa P dr
!
γa − 1 1 dP 1 dT
= −
γa P dr T dr
5
The signs get a little confusing here. Recall that dT /dr is negative – temperature
falls as one moves upward through a star. Thus this equation means that a system
is stable to convection as long as the true temperature gradient is less negative
than the adiabatic one. To avoid confusion, it is common to take the absolute
value of both sides, which, since both sides are negative, gives
dT dT
< .
dr dr
ad
A star within which the temperature gradient is steeper than the adiabatic tem-
perature gradient is said to be super-adiabatic. What we have shown is that
superadiabatic temperature gradients are convectively unstable.
Finally, it is important to point out that this analysis is for regions of a star
dominated by gas pressure. It can be extended to include radiation pressure in a
fairly straightforward manner, and this extension can be important in the centers
of massive or evolved star where radiation pressure is important. The general
result is that, if radiation pressure is important, convection is more likely.
A. Locations of Convection
As a first step toward this, let us consider where convection is likely to occur in
a star. To do this, it is helpful to write down the temperature gradient that is
produced by radiation alone, and compare it to the adiabatic value. If there is
no convection, then the temperature gradient is given by the equation we have
already derived:
dT 3 κρ Frad
=− ,
dr 4ac T 3 4πr2
where we have added the subscript rad on F to emphasize that this is the flow
carried by radiation, which need not match the total flow if convection is occuring.
Note, here we define F as the heat flow, which is the heat flux times 4πr2 . The
convective stability condition that dT /dr > (dT /dr)ad therefore implies that
!
3 κρ Frad γa − 1 µ
− 3 2
> − g
4ac T 4πr γa R
!
γa 3R κρ Frad
< 1.
γa − 1 4acµg T 3 4πr2
6
Convection begins if this inequality is violated. In practice, it is violated in three
situations.
First, if the stellar opacity (κ) is large, the inequality is violated. Physically, this
occurs because a large opacity means that the temperature gradient must become
steep to carry the same heat flow. The star responds by developing a steeper tem-
perature gradient until it becomes so steep that it exceeds the adiabatic gradient,
at which point convection starts. Since κ generally increases with decreasing tem-
perature, this situation occurs most commonly in the cooler outer parts of stars
than in their cores.
Second, in the ionization zones in a star, γa can become small due to ionization
effects. This also makes the left-hand side large. Due to this effect, we expect the
ionization zones in stars to be highly convective. Again, this occurs fairly near
the stellar surface, since the deep interior is fully ionized.
Third, if the energy generation rate in the star is very sharp function of tempera-
ture, then F rises rapidly as r approaches 0 inside a star. This large heat flow at
a small radius leads to violation of the inequality. This happens only in the center
of the star, and only if the nuclear reactions are very sensitive to temperature,
e.g. the CNO cycle or the triple-α process.
In the Sun, the since the p − p chain dominates, the third type of convective
instability doesn’t occur. The center of the Sun is convectively stable. In the
outer part of the Sun, the first and second types of convective instability do
occur, so the outer part of the Sun is convective. In less massive stars, the gas is
cooler, and the first and second types of convection occur over ever-larger fractions
of the star, working their way down toward the center. At ∼ 0.3 M the star is
fully convective.
In the opposite direction, as one moves to stars more massive and hotter than the
Sun the convection zone at the top of the star disappears, while one driven by the
strong temperature-dependence of the CNO cycle appears at the base of the star
and covers more and more of its mass as the stellar mass increases.
7
star, and it is that flux we want to calculate.
In this picture, the star has some temperature gradient dT /dr, which is more
negative than the adiabatic gradient (dT /dr)ad . Thus when the bubble rises a
distance dr, the gas surrounding it has decreased in temperature by an amount
dT
dT (s) = dr
dr
The bubble, on the other hand, is adiabatic until the point where it stalls and
mixes with its environment. Therefore after it rises a distance dr, its temperature
changes by an amount !
(b) dT
dT = dr.
dr ad
The difference in temperature between the bubble and its surroundings is therefore
" ! # !
(b) (s) dT dT dT
δT = dT − dT = − dr ≡ δ dr.
dr ad
dr dr
The quantity δ(dT /dr) that we have defined is a measure of how superadiabatic
the gas is. At δ(dT /dr) = 0 the temperature gradient is adiabatic and convection
shuts off.
Now suppose that a hot, rising bubble travels a distance ` before it fully mixes
with the surrounding gas and gives up its thermal energy. As the bubble mixes,
the amount of energy per unit bubble volume that it transfers to its surroundings
is !
dT
δq = ρcP δT = ρcP δ `,
dr
where cP is the specific heat capacity of the gas at constant pressure. For an ideal
monatomic gas, cP = (5/2)(R/µ), but we leave the expression as cP because in
convective zones where ionization is important one must use a value of cP that
accounts for ionization energy.
This is the heat per unit volume carried by one bubble. If we want to know the
heat flow associated with the collective motion of all the rising bubbles in the
star, we must multiply by the average speed with which the bubbles move and
the area through which they move:
!
dT
Fc = ρcP δ `v c (4πr2 ).
dr
This expression gives the convective heat flow in the star, which must be added
to the radiative flow to find the total.
The remaining steps are to evaluate ` and v c , the characteristic distance that
bubbles get before dissolving, and the characteristic velocity with which they
rise. Unfortunately at this point we lack a “spherically symmetric” theory of
8
convection that would tell us with certainty. Instead, we are forced to use an
empirical approximation called Mixing Length theory. This is right at the order
of magnitude level, and mostly tells us what we need to know; but it is definitely
not complete. Getting a better understanding of how convection really works is
a major challenge in 3D time-dependent fluid dynamics.
The first approximation of Mixing Length theory is to guess that the typical
distance that a convective bubble travels before breaking up is set by the condition
that the pressure change significantly, so that the bubble must expand significantly
to stay in pressure balance. As long as the bubble expands by a small amount, it
should survive, but once it has to roughly double its volume, it should break up.
To make this definite, we use the equation of hydrostatic balance:
dP Gm
= − 2 ρ = −ρg,
dr r
where g = Gm/r2 is the local gravitational acceleration, which we have defined
for convenience. We are interested in the distance dr that one must travel before
the change in pressure dP is of order P , i.e.
dP 1 dP ρg
1∼ = dr = − dr
P P dr P
Thus we expect a change in the pressure of order unity when dr ∼ P/(ρg). We
define this quantity as the pressure scale height,
P
HP = ,
ρg
and the first basic assumption of Mixing Length theory is that ` ∼ HP . To make
it formal, we write
P RT
` = αHP = α = α
ρg µ g
where α is a dimensionless fudge factor of order unity that represents our igno-
rance.
The second thing we need to approximate is the velocity of the convective bubbles,
v c . To estimate this, recall our equation of motion for the bubble, which we used
in deriving the Brunt-Väisälä frequency:
d2
(dr) = Ag dr,
dt2
where the quantity A is given by
! " ! # !
γa − 1 1 dP 1 dT 1 dT dT 1 dT
A= − = − = δ .
γa P dr T dr T dr ad
dr T dr
9
Thus the equation of motion can be written
d2
!
g dT
2
(dr) = δ dr.
dt T dr
For such a uniform acceleration we can use the old first-term physics standby
formula vf2 = vi2 + 2a ∆x. Since the initial velocity is vi = 0, and ∆x = ` is the
distance traveled, the final velocity is
" !#1/2 " !#1/2 " !#1/2
1/2 g dT g dT RT R T dT
vf = (2a`) = δ `= δ α =α δ .
T dr T dr µ g µ g dr
We are after the mean velocity, which must be somewhere between 0 and vf , so
we again insert another parameter to represent our ignorance. We set
" !#1/2
R T dT
vc = α β δ .
µ g dr
10
becoming too steep. For example, if δ(dT /dr) then Fc increases but then |dT /dr|
decreases, which make δ(dT /dr) decrease.
To make this argument quantitative, suppose that none of the heat flow is carried
by radiation; only convection occurs. This is the limit that gives the maximum
possible temperature gradient, since any additional heat flow due to radiation on
top of convection will only smooth things out further.
If we are outside the part of the star where nuclear burning occurs, then F is
simply the total stellar luminosity L, and under our assumption that there is no
radiative flow, this means that Fc = L. Plugging this into the formula for Fc and
solving for δ(dT /dr) gives
! " 2 3/2 #2/3
dT 1 µ L 1 g
δ = 2 1/2 .
dr α β R 4πr2 ρcP T
This represents the difference between the true temperature gradient and the
adiabatic temperature gradient. We want to know what fraction of the adiabatic
temperature gradient this is, so we divide by |dT /dr|ad = g/CP , which is true for
a monatomic, ideal gas. This gives
4/3 2/3
δ (dT /dr) µ L
1/3
= α−4/3 β −1/3 CP ρ−2/3 T −1 .
|dT /dr|ad R 4πr2
We can evaluate this directly by plugging in, but it is more instructive to examine
the physical meaning of this expression. To the order of magnitude level, r ∼ R
and ρ ∼ M/R3 . If we are dealing with an ideal gas, then cP ∼ R/µ. Finally, recall
that the virial theorem implies that the mean temperature T ∼ (µ/R)(GM/R)
Plugging this in, and dropping factors of order unity,
2/3 !1/3 !2/3
R3
4/3 !
δ (dT /dr) µ L R R R
∼
|dT /dr|ad R R2 µ M µ GM
!1/3
L2 R 5
=
G3 M 5
2 !#1/3
R3
"
RL
=
GM 2 GM
2/3
tdyn
=
tKH
Thus the physical meaning of this expression is that the deviation from adia-
baticity is of order the ratio of the dynamical to the KH timescale, to the 2/3
power. It makes sense that the deviation from adiabaticity should involve this
ratio. Convection is a dynamical instability, where the speeds of motion are set
by the forces of buoyancy and gravity. Thus it should be able to transport heat
on a dynamical timescale. Effects trying to produce a large temperature gradient,
11
like radiation, operate on a KH timescale. Thus the amount by which convection
dominates is determined by the ratio of these timescales.
Numerically, recall that for the Sun tdyn ∼ 3000 s and tKH ∼ 30 Myr. Thus
!2/3
δ (dT /dr) 3000 s
∼ ∼ 10−8 .
|dT /dr|ad 30 Myr
12
Thus the stellar structure equation for the temperature changes to
" ! ! #
dT dT dT
= max ,
dr dr rad
dr ad
" ! #
3 κρ F γa − 1 µ Gm
= − min 3 2
, .
4ac T 4πr γa R r2
The first line is a maximum rather than a minimum because (dT /dr)rad and
(dT /dr)ad are both negative, meaning that taking the maximum is equivalent
to selecting whichever one has a smaller absolute value. The second line is a
minimum because we have factored out the minus sign.
13