AST2000 Lecture Notes

Part 3B
Stellar birth

Questions to ponder before the lecture

1. A star starts its life as a huge cold gas cloud which contracts. But not all gas clouds will contract
to form a star. Which conditions do you think are necessary for a gas cloud to form a star?
2. How large and how cold do you think a typical star forming gas cloud could be at the beginning
of the contraction process?
3. A star forming gas cloud becomes very hot and starts shining even before nuclear reactions
start? Why? Where does the energy come from?

An artistic view of the dusty protoplanetary disk around a massive young star (Image: ESO/L. Calçada/M. Kornmesser)

1
 AST2000 Lecture Notes

Part 3B
Stellar birth

The Virial Theorem

1 The virial theorem 1
hKi = − hU i.
2
We have seen that we can solve the equation of
motion for the two-body problem analytically and where hKi is the mean kinetic energy of the sys-
thus obtain expressions describing the future mo- tem and hU i is the mean potential energy of the
tion of these two bodies. Adding just one body to system. The mean value here can be taken over
this problem, the situation is considerably more all the bodies in the system. We will not go into
difficult. There is no general analytic solution to details about the condition for when a system is
the three-body problem. In astrophysics we are in equilibrium. The equilibrium condition usu-
often interested in systems of millions or billions ally applies to systems which are bound. In this
of bodies. For instance, a galaxy may have more course it will be clear from the context when the
than 2 × 1011 stars. To describe exactly the mo- virial theorem is applicable.
tion of stars in galaxies we would need to solve
the 2 × 1011 -body problem. This is of course im- 1.1 OPTIONAL: Deducing the virial the-
possible, but we can still make some simple con- orem
siderations about the general properties of such
a system. We have already encountered one such
general property, the fact that the center of mass
maintains a constant velocity in the absence of
external forces. A second law governing a large
system is the virial theorem. The virial theorem
has a wide range of applications in astrophysics,
from the formation of stars (in which case the
bodies of the system are the atoms of the gas) Figure 1: The N-body system.
to the formation of the largest structures in the We will consider a system of N particles (or bodies) with
universe, the clusters of galaxies. mass mi , position vector ~ri , velocity vector ~vi and momen-
tum p~i = mi~vi (see figure 1). We will take the origin of our
The full derivation of the virial theorem is not a system to be the center of mass for reasons which we will see at
part of this course, but is given below (optional) the end. For this system, the total moment of inertia is given
for those who are interested. The virial theorem by (remember from your mechanics classes?)
N N
is a relation between the total kinetic energy and X X
I= mi |~ri |2 = mi~ri · ~ri .
the total potential energy of a system in equilib- i=1 i=1
rium. It says: In mechanics one usually takes the moment of intertia with
respect to a given axis, here we take the moment of intertia

2
with respect to the origin. The time derivative of the moment where the sums have been interchanged (you can easily con-
of inertia is called the virial, vince yourself that this is the same sum by looking at the ma-
trix in figure 2). We can also interchange the name of the
N
1 dI X indices i and j (this is just renaming the indices, nothing else)
Q= = ~i · ~ri .
p
2 dt i=1 N X
X
B= f~ji · ~rj .
To deduce the virial theorem we need to take the time deriva- i=1 j<i
tive of the virial
N N From Newton’s third law, we have f~ij = −f~ji ,
dQ X d~ pi X
= · ~ri + ~i · ~vi ,
p
dt i=1
dt i=1
N X
X
B=− f~ij · ~rj .
where Newton’s second law gives i=1 j<i

d~ ~i
pi /dt = F Totally, we have,
N
~i being the sum of all forces acting on particle i. We may
F
X
~i · ~ri = A + B
F
write this as i=1
N N
dQ X
~i · ~ri +
X N X N X
= F mi vi2 , X X
dt i=1 i=1
= f~ij · ~ri − f~ij · ~rj
i=1 j<i i=1 j<i
where the last term may be expressed in terms of the total
N X
kinetic energy of the system K = i 1/2mi vi2
P X
= f~ij · (~ri − ~rj ). (2)
N i=1 j<i
dQ X
~i · ~ri + 2K.
= F (1)
dt i=1
Did you follow all steps so far? Here, the force f~ij is nothing
else than the well known gravitational force,
We will now try to simplify the first term on the right hand
mi mj
side. If no external forces work on the system and the only f~ij = G 3 (r~j − r~i ),
rij
force which acts on a given particle is the gravitational force
from all the other particles, we can write where rij = |~rj −~ri |. Note that the force points in the direction
N N X of particle j. Inserting this into equation (2) gives
X X
~i · ~ri =
F f~ij · ~ri , N N X N
i=1 i=1 j6=i
X
~i · ~ri = −
X mi mj 2 XX
F G 3
rij = Uij ,
i=1 i=1 j<i
rij i=1 j<i
where f~ij is the gravitational force on particle i from particle
j. The last sum is a sum over all particles j except particle where Uij is the gravitational potential energy between particle
j = i. The double sum thus expresses a sum over all possible i and j. This sum is the total potential energy of the system
combinations of two particles i and j, except the combination (do you see this?), the sum of the potential between all pos-
where i = j. We may view this as an N × N matrix where sible pairs of particles (note that one pair of particle should
we sum over all elements ij in the matrix, except the diagonal be counted only once, this is why there is a j < i in the latter
elements ii. We divide this sum into two parts separated by sum). Thus, we have obtained an expressions for the two terms
the diagonal (see figure 2), in equation (1) expressing the time derivative of the virial
N N X N X dQ
X X X = U + 2K.
~i · ~ri =
F f~ij · ~ri + f~ij · ~ri dt
i=1 i=1 j<i i=1 j>i
| {z } | {z } Finally we will use the equilibrium condition. We will take the
≡A ≡B mean value of this expression over a long period of time,
dQ
h i = hU i + 2hKi,
dt
where Z τ
1
hi = lim dt.
τ →∞ τ 0
For the term on the left hand side, we find

1 τ dQ
Z
Q(τ ) − Q(0)
Figure 2: The matrix visualizing the summation. lim dt = lim ≡ 0,
τ →∞ τ 0 dt τ →∞ τ
We now rewrite the sum B as for a system in equilibrium. The last equality here is the def-
N X N X
inition of the equilibrium state in which the system needs to
be for the virial theorem to hold: the mean value of the time
X X
B= f~ij · ~ri = f~ij · ~ri ,
i=1 j>i j=1 i<j derivative of the virial must go to zero. In order for this to be

3
fulfilled, the quantities Q(τ ) and Q(0) need to have finite val- Before the advent of the theory of relativity, the
ues. If, for instance, the system is bound and the particles go source of the energy that powers stars was sought.
in regular orbits, the virial Q will oscillate regularly between
two finite values. In this case, the last expression above will go
One suggestion was that the stellar energy was
to zero as τ → ∞. If Q had not been limited, which could hap- gravitational energy that is being radiated away
pen for a system which is not bound, then Q could attain large as the cloud of gas retracts. A star starts out as
values with time and it would not be clear that this expression a huge cloud of gas which starts collapsing due
would approach zero as τ → ∞.
to its own force of gravity. Gas falls towards the
Using the above equation and the equilibrium condition we see center of the cloud and releases gravitational en-
that a bound system in equilibrium obeys
ergy in the form of electromagnetic radiation as
it falls. As long as the cloud keeps collapsing,
energy is radiated away and could possibly ex-
The Virial Theorem
plain the energy production in stars. To check
1 if this is a plausible explanation, we will need to
hKi = − hU i.
2
calculate the total energy, kinetic plus potential,
that the star could possibly radiate away during
In order to obtain hKi and hU i we need to take the average of
the kinetic and potential energy over a long time period. In the
its collapse and compare this with the energy out-
case of the solar system, this is easy: The orbits are periodic so put from the Sun. To calculate the total energy of
it suffices to take the average over the longest orbital period. such a cloud, we need to invoke the virial theorem.
Please note that we have done the calculations in the center A collapsing cloud of gas is a bound many-body
of mass frame. If we did it from a different frame of reference,
our system of particles would move at a constant speed with system and the virial theorem should apply.
respect to us and the distance to the system would increase
indefinitely. All the distances would grow to infinity and the
We will assume that the cloud is spherically sym-
time derivative of the virial would not go to zero. metric with radius R and mass M . We need to
Averaging a system over a long time period may be equal to
calculate the total energy, kinetic plus potential,
averaging the system over the ensemble. This is the ergodic of such a cloud. Thanks to the virial theorem,
hypothesis. Mathematically it can be written as it suffices to calculate only the potential energy.
Z τ N The total energy is given by
1 1 X
lim dt → lim .
τ →∞ τ 0 N →∞ N i=1 1 1
E = K + U = − U + U = U,
If a bound system has a huge number of particles (N → ∞),
2 2
it is equivalent to seeing the system over a long period of time where K is kinetic energy and U is potential en-
(τ → ∞). Thus, we can apply the virial theorem to a galaxy ergy. Using the virial theorem K = −U/2, we
by taking the mean of the kinetic and potential energy of all
stars in the galaxy in a given instant. According to the ergodic replace K by U and obtain an expression for the
hypothesis, it is not necessary in this case to take the mean total energy given only in terms of the total po-
of the kinetic and potential energy over a very long period of tential energy.
time. Since the time scales for changes for such huge systems is
very long, it is much easier to simply take the average over all We see that if we are able to calculate the total po-
stars. The ergodic theorem thus says that we can replace the tential energy of the cloud, we would also obtain
mean value from being a time average to be an average over
all bodies in the system.
the total mechanical energy (kinetic+potential).
To obtain the total potential energy, we will start
by considering the potential du of a tiny particle
2 Applying the virial theorem to a of mass dm inside the cloud at a distance r from
the center. We have learned (see the lectures on
collapsing cloud of gas dark matter) that the gravitational forces from a
spherical shell of matter add to zero inside this
To show the power of the virial theorem we will shell. Thus we need only to consider the gravita-
apply it to a system with very many particles and tional attraction on the mass dm from the sphere
show how properties of this complex system may of matter inside the position of the mass. This
be calculated. In the exercises you will find two is a sphere of radius r with mass M (r). Being a
more examples of applications of the virial theo- sphere, Newton’s law of gravitation applies as if
rem to problems of a very different nature. it were a point mass located at the center with

4
mass M (r). Thus the potential energy between
the particle dm and the rest of the cloud (the part 3GM 2
U =− . (3)
inside the particle) is 5R

M (r)dm From the virial theorem, the total energy is then

du = −G .
r (check!)
1 3GM 2
We integrate this equation over all masses dm in E= U =− .
the shell of thickness dr at distance r from the 2 10R
center. We assume that the mass density in the This is the total energy of a cloud of gas with
shell is given by ρ(r). We then obtain the poten- mass M and radius R. The energy that the Sun
tial energy dU between the shell and the spherical has radiated away during its lifetime can be writ-
mass M (r) inside the shell. ten as

M (r)4πr2 ρ(r)dr Eradiated = E(bigR) − E(R ),

dU = −G .
r where ’big R’ refers to the radius of the cloud
To obtain the total potential energy U, we need when it started collapsing and R is the current
to integrate this expression over all radii r out to radius of the Sun. The total energy of the cloud
the edge of the cloud at r = R, goes as ∝ 1/R, so for the initial cloud this quan-
Z R tity can be approximated to zero. Thus we are
U = −4πG M (r)ρ(r)r dr. left with
0
3GM 2
Eradiated = ,
10R
We would generally need to know the density ρ(r)
in order to obtain M (r) and to integrate this where M is the mass of the Sun. Inserting num-
equation. The scope here is to obtain an approxi- bers for the mass and 41
radius of the Sun we obtain
mate expression giving us an idea about the mass E radiated ≈ 1.1 × 10 J. Assuming that the Sun
and radius dependence of the energy and to ob- has been radiating with the same luminosity L
tain an order of magnitude estimate. For this (dE/dt) during its full lifetime, we can calculate
purpose, we assume that the density is constant the age of the Sun,
with a value equal to the mean density of the Eradiated
cloud, ∆t = ≈ 107 years.
L
M
ρ= .
(4/3)πR3 If gravitational collapse was indeed the source of
solar energy, the Sun couldn’t have lived longer
This gives M (r) = (4/3)πr3 ρ and we can inte- than about 10 millions years. Several geological
grate the equation findings have shown that the Earth and there-
 2 Z R fore also the Sun has existed for about 500 times
M as long. Thus using the virial theorem we have
U = −4πG (4/3)π r4 dr,
(4/3)πR3 0 shown (using some assumptions) that gravita-
tional collapse cannot satisfactory explain the
generation of energy in the Sun.

5
Fact sheet: Fritz Zwicky was the first to use the virial theorem to infer the
existence of unseen matter, which he referred to as ”dunkle Materie” dark
matter. He used the theorem in 1933 to calculate the mass of the Coma clus-
ter of galaxies (aka. Abell 1656) and found that it was much larger than the
mass expected from the luminous matter. The cluster contains more than one
thousand galaxies, most of them ellipticals. It lies in the constellation of Coma
Berenices, at a mean distance of roughly 100 Mpc. The central region is dom-
inated by two giant elliptical galaxies, which are easily spotted in the above
image. The bright blue-white source above the center is a foreground star in
our own galaxy. (Figure: J. Misti)

3 The Jeans criterion where m = µmH is the mean mass per gas par-
ticle. We repeat the definition of mean molecular
A star forms from a cloud of gas, a so-called weight
m
molecular cloud, undergoing gravitational col- µ= .
lapse. These molecular clouds consist mainly of mH
atomic and molecular hydrogen, but also contain This is simply the mean mass per particle mea-
dust and even more complex organic molecules. sured in units of the hydrogen mass mH (check
The question is whether a cloud will start col- now that expression 4 for N makes sense to you!
lapsing or not. The virial theorem tells us that This is important!). So the condition 2K < |U |
the condition for stability is 2K + U = 0. If the becomes simply
kinetic energy is large compared to the potential
3M kT 3GM 2
energy (and using the virial theorem we see that < .
we need to compare 2K to U ), the system does µmH 5R
not stabilize, the gas pressure is larger than the We can write this as a criterion on the mass
gravitational forces and the cloud expands. On
the other hand, if the potential energy is dom- 5kT
M> R.
inating, the cloud is gravitationally bound and GµmH
undergoes collapse. For a cloud to collapse we This minimum mass is called the Jeans mass MJ
thus have the condition (why?), which we can write in terms of the mean density
2K < |U |. of the cloud as

We will now use the expression for the total po-

tential energy of a gas cloud which we deduced to The Jeans mass
be (equation 3),  3/2  1/2
5kT 3
MJ = ,
3GM 2 GµmH 4πρ
U =− ,
5R
where M is the mass of the cloud and R is the where we used ρ = M/((4/3)πR3 ) assuming con-
radius. From thermodynamics, we learn that the stant density throughout the cloud. Thus, clouds
kinetic energy of a gas is given by with a larger mass than the Jeans mass M > MJ
3 will have 2K < |U | and therefore start a gravita-
K = N kT, tional collapse. We can also write this in terms
2
of a criterion on the radius of the cloud. Using
where N is the number of particles in the gas, k again the expression for the density we have the
is the Boltzmann constant and T is the tempera- Jeans length (check again that you can deduce
ture. As we did in earlier lectures, we can write this expression from the expression above).
N as
M
N= , (4)
µmH

6
 The Jeans length
 1/2
15kT
RJ = .
4πGµmH ρ

A cloud with a larger radius than the Jeans length

R > RJ will undergo gravitational collapse. The
Jeans criterion for the collapse of a cloud is a good
approximation in the absence of rotation, turbu-
lence and magnetic fields. In reality however, all
these factors do contribute and far more compli-
cated considerations are needed in order to calcu-
late the exact criterion.
The collapsing cloud will initially be in free fall,
a period when the photons generated by the con-
verted potential energy are radiated away without
heating the cloud (the density of the cloud is so
low that the photons can easily escape without
colliding with the atoms/molecules in the gas).
The initial temperature of the cloud of about
T = 10 − 100 K will not increase. After about
one million years, the density of the cloud has
increased and the photons cannot easily escape.
They start heating the cloud and potential energy
is now radiated away as thermal radiation. In the
beginning of this part, we made an approximate
calculation of the time it would take the Sun to
collapse to its present size assuming a constant
luminosity. We found a collapse time of about 10
million years. Proper calculations show that this
process would take about 40 million years for a
star similar to the Sun. The contracting star is
called a protostar.
When the core of the collapsing protostar has
reached sufficiently high temperatures, thermonu-
clear fusion begins in the center. The luminosity
starts to get dominated by the energy produced
by nuclear fusion rather than converted potential
energy from the gravitational collapse. The pro-
tostar keeps contracting until hydrostatic equilib-
rium is reached and the star has entered the main
sequence.

7
Fact sheet: A close-up of one of the famous ”Pillars of Creation” in the Eagle
Nebula (M16), a nearby star-forming region some 2000 pc away in the constel-
lation Serpens. This pillar of cool interstellar hydrogen gas and dust is roughly
4 light-years long and protrudes from the interior wall of a dark molecular
cloud. As it is slowly eroded away by strong ultraviolet light from nearby stars,
small globules of even denser gas buried within the pillar are uncovered. These
globules are most easily seen at the top of the pillar. They are dense enough
to collapse under their own gravity, forming young stars and possibly plane-
tary systems. This color image is constructed from three separate images taken
through filters specially designed to isolate the light from different gases. Red
shows emission from singly-ionized sulfur atoms, green shows emission from
hydrogen, and blue shows light emitted by doubly-ionized oxygen atoms. (Fig-
ure: NASA, ESA, STScI, J. Hester and P. Scowen)

4 Exercises 1. Looking in the telescope we see that the clus-

ter is spherical, the galaxies are evenly dis-
Exercise 3B.1 tributed inside a spherical volume. The dis-
tance to the cluster is 85 Mpc. You observe
In a way we can look at the virial theorem as a the radius of the cluster to be 320 . What is
generalization of Kepler’s third law to a many- the radius of the cluster in Mpc?
body system. Show that for the two-body prob-
lem, the virial theorem is identical to Kepler’s 2. All galaxies in the cluster appear to be very
third law in the Newtonian form (as deduced in similar to the Milky Way, both in the num-
the exercises in part 1B). Assume circular orbits. ber of stars and the type of stars. The galax-
Start with the virial theorem, insert expressions ies look so similar to each other that we can
for the energies and show Kepler’s third law. (You assume that all the galaxies have the same
won’t get more help here. . . ). mass m. We know that the Milky Way has
about 2 × 1011 stars. Assuming that the
mean mass of a star equals the mass of the
Exercise 3B.2 Sun, what is the estimated total luminous
Fritz Zwicky was the first to note that there is mass m of these galaxies?
some missing matter in the universe. In 1933, 3. Use the virial theorem to show that the mass
several years before the discovery of the flat ro- m of a galaxy in the cluster can be written
tation curves in the galaxies, he used the virial as PN 2
theorem to calculate the mass of galaxies in the vi
Coma Cluster. A cluster of galaxies is a cluster of m = PN i=1 P ,
G i=1 j>i 1/rij
a few hundred galaxies orbiting a common center
of mass. The Coma Cluster is one of our neigh- where rij is the distance between galaxy i
bouring clusters of galaxies. He found that the and galaxy j and vi is the velocity of galaxy
mass of the Coma Cluster calculated using the i with respect to the center of mass.
virial theorem was much larger than the mass ex- 4. You will find a file with data for each of the
pected from the visible luminous matter. In this galaxies here:
problem we will try to follow his example and esti- http : / /www. u i o . no/ s t u d i e r / emner /
mate the mass of galaxies in a cluster of galaxies. matnat / a s t r o /AST2000/ h17 /
We will consider a simulated cluster of about 100 u n d e r v i s n i n g s m a t e r i a l e −2017/
galaxies. We will assume that the cluster consists f i l e r −t i l −oppgaver / f i l e r −t i l −
of these 100 brightest galaxies and assume that oppgaver−i −d e l −3b/ g a l a x i e s .
the remaining galaxies are too small to affect our txt
calculations significantly.

8
The first column in the file is the observed sum in Python you can construct two
angular distance (in arcminutes) from the FOR-loops, one over the index i and one
center of the cluster along an x-axis. The over the index j. Inside the two FOR-
second column in the file is the observed an- loops, you add the expression inside the
gular distance (in arcminutes) from the cen- sum for indices i and j to the final result.
ter of the cluster along an y-axis. (the x- Hint 2: To find the distance between
y coordinate system is chosen with an arbi- two galaxies i and j, it is convenient to
trary orientation on the plane of observation find the x, y and z coordinates of each
(which is perpendicular to the line of sight)). galaxy in meters.
The third column is the measured distance to
the galaxy (from Earth) in Mpc. The fourth (d) Your measured velocities are based on
column is the position of the spectral line at the Doppler effect and are therefore ra-
21.2 cm for the given galaxy in units of m. dial velocities. Because the inclinations
(a) Using these data, what is the radial ve- of the velocities with the line of sight is
locity of the cluster with respect to us? not 90◦ , your estimate is a minimum es-
Remember that the velocity of a galaxy timate of the mass. We will now use the
can be written as fact that you have many galaxies and
that you know that the orientation is
v(gal) = v(cluster) + v(rel), random to get a more exact estimate.
As a first step you will need to find
where v(gal) is the total velocity of the the mean of sin2 i (where i is inclina-
galaxy with respect to us, v(cluster) is tion) taken over many galaxies with ran-
the velocity of the cluster (of the cen- dom orientations: What is the expected
ter of mass of the cluster) with respect mean value taken over many galaxies of
to us and v(rel) is the relative velocity the expression sin2 i? We assume that
of a galaxy with respect to the center of the inclination is random (with a uni-
mass of the cluster. The relative veloci- form distribution). Remember that the
ties with respect to the center of mass mean value of a function f (x) is defined
are random, so for a large number of statistically by
galaxies the mean
N
R
1 X dxf (x)P (x)
vi (rel) → 0 hf (x)i = R ,
N i=1 dxP (x)

goes to zero.
where P (x) is the statistical distribu-
(b) Make a plot showing how this cluster ap- tion, i.e. the probability of having a
pears in the telescope: draw the x-y axes value x. The denominator here is to
(using arcminutes as units on the axes) ensure that the integral over the distri-
and make a dot at the position for each bution P (x) is 1 as it needs to be (see
galaxy. Remember that in Python you part 1A). In this case, the distribution is
can plot for instance a circle at each data uniform, meaning that there is an equal
point by using plot(x,y,’o’). probability for getting any value of the
(c) Use these data and the expression above inclination i. We may thus set P (x) = 1.
for the mass of a galaxy from the virial The integration in this general expres-
theorem to obtain a minimum estimate sion is done over all possible values of
of the total mass of a galaxy in the clus- x.
ter. How does it compare to the esti-
mate you obtained for luminous matter (e) Can you use this to obtain a more accu-
above? Hint 1: To make the double rate estimate of the mass?

9
Exercise 3B.3 pushing the cloud together, could this pos-
sibly lead to a decrease in the minimum
A Giant Molecular Cloud (GMC) has typically a
temperature of T = 10 K and a density of about mass required for collapse (give arguments in
terms of K and U )? Argue why a decrease
ρ = 3 × 10−17 kg/m3 . A GMC has been observed
in minimum mass is more probable than an
at a distance of r = 200 pc. It’s angular exten-
sion on the sky is 3.50 . Assume the cloud to be increase. (Hint: does K really increase for
spherical with uniform density. all particles when you compress the cloud?).

1. What is the actual radius of the cloud? 5. Could the supernova thus have contributed
to the collapse of a cloud which has a mass
2. What is the mass of the cloud? less than the Jeans mass?
3. Is the mass larger than the Jeans mass? Is
6. The galaxy has a fairly uniform distribution
the cloud about to collapse and form a pro- of hydrogen in the galactic disc. If a pressure
tostar?
wave is moving around the center of the disc
4. A supernova explodes in the vicinity of the in a spiral like shape, would this explain why
star emitting a pressure wave which passes we observe galaxies as spirals and not as a
through the cloud. If an external pressure is disc?

10