Part1 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 48

University of Canterbury

Stellar Structure
and

Evolution
Astr 323/Phys 323
Plan:
1. Structure equations

2. Microphysics

3. The Sun & the main-sequence evolution


The structure equations
• Stars are self-gravitating objects of hot plasma;

• emitting energy in the form of photons from the


surface;

• spherical symmetry (absence of rotation and mag-


netic fields);

→one-dimensional problem with radius r being the


natural coordinate (Euler description).

Mass and radius

Eulerian description: mass dm in a shell at r and of


thickness dr is

dm = 4πr 2 ρdr − 4πr 2 ρvdt

From the partial derivatives of this equation one can


derive:

∂ρ 2
−2 ∂(ρr v)
= −r
∂t ∂r

∂ρ
→ = −∇(ρv)
∂t

(continuity equation in 1-dimensional form and Eule-


rian description)
Lagrangian description: mass elements m (mass in a
concentric shell).
⇒ r = r(m, t)

Variable change (r, t) → (m, t):


∂ ∂ ∂r
=
∂m ∂r ∂m
and
µ ¶ µ ¶ µ ¶
∂ ∂ ∂r ∂
= +
∂t m ∂r ∂t m ∂t r

∂r 1
= (1)
∂m 4πr 2 ρ

This is the first structure equation (mass equation or


mass conservation).

Contains transformation Euler- → Lagrange-description


∂ 1 ∂
=
∂m 4πr 2ρ ∂r
Gravity

Gravitational field
∇2Φ = 4πGρ
(G = 6.673 · 10−8 dyn cm2 g−2).

In spherical symmetry:
µ ¶
1 ∂ ∂Φ
r2 = 4πGρ
r 2 ∂r ∂r

∂Φ Gm
g= ∂r
→ g= r2
is solution of Poisson’s equation.

Potential
R∞ Φ vanishes for r → ∞.
− 0 Φdr is the energy required to disperse the com-
plete star to infinity.
Hydrostatic equilibrium

On layer of thickness dr two forces:

gravity −gρdr and pressure △P = − ∂P


∂r
△r.

If shell is at rest (hydrostatic equilibrium):

∂P Gm
=− 2 ρ
∂r r

in Lagrangian coordinates:

∂P Gm
=− (2)
∂m 4πr 4

Second structure equation (hydrostatic equilibrium).


Estimates for central values of the Sun:

Replace derivatives in the hydrostatic equation by dif-


ferences between center (Pc ) and surface (P0 ≈ 0) →

2GM 2
Pc ≈
πR4

(M/2 and R/2 were used for mean mass and radius)

Sun: Pc = 7 · 1015 (cgs units).


µP
With ρ = RT
and ρ̄ = (3M )/(4πR3) ⇒
8 µ GM ρ̄
Tc = < 3 · 107 K
3 R R ρc
Motion:

∂P Gm 1 ∂ 2r
+ =−
∂m 4πr 4 4πr 2 ∂t2

1. P = 0 → free-fall Gm/r 2 = r̈.


p p
2. τf f = R/|r̈| ≈ R/g.
p
3. G = 0, τexpl ≈ R ρ/P (isothermal sound speed)

4. hydrostatic timescale τhydro ≈ 12 (Gρ̄)−1/2

5. Examples: τhydro =
• 27 minutes for the Sun
• 18 days for a Red Giant (R = 100R⊙)
• 4.5 seconds for a White Dwarf (R = R⊙ /50)

Conclusion: stars return to hydrostatic equilibrium


within an extremely short time.
Energy reservoirs:

1. Thermal (or internal) energy (for an ideal gas)

R
P = ρT
µ
R 2
= cP − c v = cv
µ 3

The thermal energy per mass unit is u = cv T

and therefore the total energy the integral over mass


of u
Z M Z M
3R 3R
Et = cv T dm = cv T dm = hT iM
0 0 2µ 2µ

For the Sun, with hT i ≈ 107 K → Et,⊙ ≈ 5 × 1048 erg

2. Gravitational energy
Z M
GMr GM 2
Eg = − dMr ≈ −
0 r R

For the Sun, Eg,⊙ = −4 × 1048 erg

Generally,
−Eg ≈ Et

Why?
The Kelvin-Helmholtz timescale
¯ ¯
|Eg |
L ≈ ¯ dE → τ := ≈ ELt .
¯ g¯
dt ¯ KH L

GM 2 GM 2
|Eg | ≈ 2R
⇒ τKH ≈ 2RL
.

Sun: τKH = 1.6 · 107 yrs.

⇒ Sun could shine only for about 10 million years,


if gravitational potential or the thermal energy would
have been its only energy source!
The Virial Theorem

Integrate eq. (2) after multiplying it by 4πr 3:

Z M Z M
∂P Gm
4πr 3 dm = − 4πr 3
dm
0 ∂m 0 4πr 4

Right hand side is obviously total gravitational energy


Z M
GM
Eg = − dm
0 r
defined to be 0 at infinity. −Eg is energy released at
assembly of star from parts at infinity.

Left hand side solved by integration by parts:

Z M Z M µ ¶
3 ∂P ∂r
3
¤M
12πr 2
£
4πr dm = 4πr P 0 − P dm
0 ∂m 0 ∂m

On right hand side, the term in brackets vanishes, if


P (M ) ≈ 0, and the integrand becomes [eq. (1)] 3P/ρ

So finally:
Z M
P
3 dm = −Eg
0 ρ
Meaning of left hand term:

For ideal gas:


P R
= T = (cP − cv )T = cv (γ − 1)T
ρ µ

γ = 5/3 for mono-atomic gas


(4/3 for photon gas).

With this P
ρ
= 32 cv T = 23 u
R
u: specific internal energy ⇒Ei := udm and
Z M
P
3 dm = 2Ei
0 ρ
In summary, we obtain the Virial Theorem
⇒ Eg = −2Ei (3)

For a more general ideal gas, 3(P/ρ) = ζu (ζ = 3(γ −


1)); for monatomic gas ζ = 2 (γ = 5/3) and for
photon gas 1. →

ζEi + Eg = 0.

Total energy W (< 0 for bound system):


ζ−1
W = Ei + Eg = (1 − ζ)Ei = ζ
Eg

But luminosity L of star must come from this energy


reservoir:

L = − dW
dt
⇒ L = (ζ − 1) dE
dt
i
.

Ėg
ζ=2 ⇒ L=− = Ėi , (4)
2
Interpretation:
star loses energy → decrease in gravitational energy
(contraction) → but same amount goes into increase
in internal energy (heating)

Stars become hotter, because they lose energy!

Note: remember assumptions!


Hydrostatic equilibrium, ideal gas.

White dwarfs lose energy and get cooler! Why?


Energy conservation

dLr = 4πr 2ρǫdm,

ǫ (erg/gs): specific energy generation rate


Sources for ǫ:

~ : nu-
• in a stationary mass shell: ǫ = ǫn (ρ, T, X)
clear energy generation;

• in non-stationary mass shell: interaction with sur-


rounding via P dV

µ ¶
∂Lr
ǫn − dt = dq = du + P dv
∂m
∂Lr ∂u P ∂ρ
= ǫn − + 2
∂m ∂t ρ ∂t
∂T δ ∂P
= ǫn − cP +
∂t ρ ∂t

ǫg : gravothermal energy
µ ¶
∂T δ ∂P 1 ∂T ∇ad ∂P
ǫg = −cP + = −cP T − (5)
∂t ρ ∂t T ∂t P ∂t

Energy loss due to plasma neutrinos: −ǫν . energy


conservation equation reads

∂Lr
= ǫn + ǫg − ǫν . (6)
∂m
Global energy conservation

The change in the total energy reservoir equals the


loss of energy due to escaping neutrinos and photons:
d
Ẇ = (EG + Ei + En) = −(L + Lν ),
dt

Integration of the energy equation (6) over m should


recover this. Some terms can easily be identified:

L = ∂m dm, Lν = ǫν dm, ǫn dm = − dE
R ∂Lr R R
dt
n
.

This leaves the integration of ǫg :

We use the formulation


∂u P ∂ρ
ǫg = − + 2 .
∂t ρ ∂t
The first term is easily identified as −dEi /dt.
RM
The second term, 0 ρP2 ∂ρ ∂t
dm needs more work:

From the virial theorem derivation we know that


(EG = grav. energy)
Z M
P
EG = −3 dm,
0 ρ
R P ∂ρ
and compute dE
R Ṗ
dt
G
= −3 ρ
dM r + 3 ρ2 ∂t
dMr
Taking the deriative w.r.t. t of the hydrostatic equa-
tion and integrating it over Mr , we get that

M
1 ∂P dEG
Z
−3 dMr = 4
0 ρ ∂t dt

using this in the previous equation we arrive at


Z M
P ∂ρ
Ėg = 4Ėg + 3 2 ∂t
dMr
0 ρ
or
Z M
P ∂ρ
2 ∂t
dMr = −E˙G
0 ρ

Therefore
Z M
ǫg dM = −Ėt − E˙G
0

and all terms in the global energy conservation equa-


tion have been recovered!
The nuclear timescale

τn := En/L

Nuclear energy reservoir: mass of fuel times the erg/g


of fuel.

Sun is in hydrogen burning, which has a heat release


of q = 6.3 · 1018 erg g−1

or a total nuclear reservoir of 8.75 · 1051 erg for the


whole Sun.

⇒ τn = 7 · 1010 yrs

τn ≫ τKH ≫ τhydr

This is the timescale most important for most stars


in most evolutionary phases. ∂L ∂m
r
≈ ǫn to high preci-
sion is an equivalent statement. It implies that ǫg ≈ 0
or that the star is said to be in thermal equilibrium.
Together with the mechanical equilibrium this is also
called complete equilibrium, because all terms involv-
ing dt are missing. Of course, complete equilibrium
cannot be achieved accurately, as will be discussed
later.
Energy transport
The energy created in the centre must be transported
outwards. This is possible only along a temperature
gradient.

T -gradient in Sun: △T /△r ≈ 107/1011 = 10−4 (K/cm).

Energy tranport is possible by radiation, convection


and conduction, the latter process being usually unim-
portant except in the case of degenerate electrons.

Formal equation for the temperature gradient:

∂T T Gm
=− ∇
∂m P 4πr 4

Determine ∇ in all cases!

Transport by radiation

Radiation intensity I weakened by intervening matter


according to
d ln I 1
dI = −Iκρdr ⇒ − = κρ =:
dr l

~ (cm2 /g).
Opacity κ(T, ρ, X)

Values for solar interior:


ρ¯⊙ = 1.4 g/cm3, κ⊙ ≈ 1 cm2 /g ⇒ l⊙ ≈ 1 cm!
Because of this very short mean free path the radiation
field is highly isotropic, the stellar interior everywhere
in Local Thermal Equilibrium, and the monochromatic
intensity well described by Planck-function Bν (T ).

Anisotropy (of Planck radiation):


△F △T △T l T l l
F ∼ T4 → =4 =4 ≈4 ≈
F T △r T RT R

For the Sun, this is of order 10−10! Therefore radia-


tion diffuses outwards only extremely slowly (timescale
about 106 · · · 107 years).

Radiation transport by diffusion:

In analogy to particle diffusion equation:


Diffusive flux ~j of particles (per unit area and time) is

~ = − 1 vlp ∇n
~j = −D∇n ~
3
(D is called the diffusion constant; v the diffusion ve-
locity; lp is the particle free path length and n the
particle density).

We now use U := aT 4 for the radiation density in place


of particle density, l = 1/(κρ) for the photon mean free
path, and c instead of v. In a 1-dimensional problem,
we get for ∇U~

∂U ∂T
= 4aT 3
∂r ∂r
and for the radiation flux F (replacing ~j)

4ac T 3 ∂T
F =− ,
3 κρ ∂r

or F = −Krad ∇T .
4ac T 3
Krad = 3 κρ
is the radiative conductivity.

Its value for the solar center is ≈ 3×1015 erg K−1 s−1 cm−1 ,
compared to values of 107 for typical metals or 104
for concrete.

⇒ Transport by photon gas is extremely efficient; even


a small T -gradient allows excellent heat transport [L⊙
transported by △T /△r ≈ 107/1011 = 10−4 (K/cm)]

With Lr = 4πr 2F , we obtain

∂T 3 κρLr
=−
∂r 16πac r 2T 3
or, in Lagrangian coordinates

∂T 3 κLr
=− (7)
∂m 64acπ 2 r 4T 3

However, this is – so far – only true for monochromatic


radiation. κ = κ(ν).
The Rosseland mean opacity
We would like to have
∂T 3 κ̄Lr
=−
∂m 64acπ 2 r 4T 3
with κ̄ being a suitable mean over frequency of κ(ν).
This mean turns out to be
R ∞ 1 ∂Bν
1 0 κν ∂T

:= R ∞ ∂Bν
κ̄ 0 ∂T

where
¶−1
2hν 3
µ µ ¶

Bν (T ) = exp −1
c2 kT

is the Planck-function for the energy density flux of a


black body. (U = aT 4 = (4π/c) Bν dν).
R

κ̄ is the Rosseland mean for opacity, mostly called κR


or simply κ.
Note that the Rosseland mean is dominated by those
frequency intervals, where matter is almost transpar-
ent to radiation, i.e. where transport efficiency is high-
est.
Note: Eq. (7) can also be derived rigidly within radi-
ation theory (see, e.g. Cox & Giuli).
Conduction
In regions of degeneracy, the mean free path of elec-
trons is very large, because the probability for mo-
mentum exchange is very small due to the fact that
all energy levels are occupied!

The flux of energy is the sum of the radiative and the


conductive heat flux:
F = Frad + Fcond = −(Krad + Kcond )∇T

If we introduce formally κcond :


4ac T 3
Kcond = ,
3 κcond ρ
this allows to replace κ in (7) by
1 1 1
= +
κ κrad κcond

The mechanism with the smaller opacity κ manages


the transport!

Effevtive ∇ in transport equation:


3 κLr P
∇ = ∇rad =
16πacG mT 4
Perturbations and stability: convection
For radiative transport, if κ and/or Lr is large, ∂T
∂r
will
be large and the question arises whether the stratifi-
cation is stable against small displacement?
Consider moving mass elements; assume no heat ex-
change with surrounding (adiabatic movement); Dif-
ferences between element and surrounding denoted by
D, e.g. DT = Te − Ts

Illustration of blobs moving in unperturbed sur-


rounding

The figure illustrates the picture we have in mind: the


temperature excess DT is positive, if the element is
hotter than its surrounding. DP = 0 due to hydro-
static equilibrium. If Dρ < 0 (ideal gas), the element
is lighter and will move upwards. Take an element
and lift it by △r:
·µ ¶ µ ¶¸
∂ρ ∂ρ
Dρ = − △r
∂r e ∂r s
The stability condition therefore is
µ ¶ µ ¶
∂ρ ∂ρ
− > 0.
∂r e ∂r s

If it is fulfilled, the drop in ρ of the element during


the upward movement is smaller than that of the sur-
roundings ( ∂ρ ∂r
< 0!), so the element will gradually be-
come as dense as the surrounding or denser and will
finally experience a downward force and return.
With the EOS d ln ρ = αd ln P − δd ln T − ϕd ln µ, the
stability condition changes to
µ ¶ µ ¶ µ ¶
δ dT δ dT ϕ dµ
− − >0
T dr s T dr e µ dr s
(Here we have used the fact that DP = 0.) Multiply
this by the pressure scale height
dr dr P
HP := − = −P = >0⇒
d ln P dP ρg
µ ¶ µ ¶ µ ¶
d ln T d ln T ϕ d ln µ
< +
d ln P s d ln P e δ d ln P s
ϕ
∇s < ∇ad + ∇µ
δ
ϕ
∇rad < ∇ad + ∇µ
δ
The last equation holds in general cases and is called
the Ledoux-criterion for dynamical stability. If ∇µ = 0,
the Schwarzschild-criterion holds. Note that ∇µ will
stabilize.
If the stability criterion is violated, convective insta-
bility will set in and convection will transport energy.
Illustration of the various temperature gradients
in a convectively unstable region

In an unstable layer, the following relations hold:


∇rad > ∇ > ∇e > ∇ad (8)

∇ is the actual gradient that will result in the convec-


tive layer. The first inequality arises from the fact that
only part of the flux can be tranported by radiation,
since convection is carrying some in any case. The
last is due to the fact that the element will cool more
than just adiabatic, because some energy will be lost
by other means (radiation, conduction). The middle
one is just the stability criterion for the blob not to
be pushed back. The task of a convection theory is
it to derive ∇!
The chemical composition
The chemical composition of a star enters ρ, κ, ǫ. It
can change due to nuclear burning, diffusion, convec-
tion and other mixing processes.
Notation:
mi ni
P
relative element mass fraction: Xi := ρ
,i Xi = 1;
special cases:
hydrogen X, helium Y , “metals” Z = 1 − X − Y .
typical values:
X ≈ 0.7 · · · 0.75; Y ≈ 0.24 · · · 0.30; Z ≈ 0.0001 · · · 0.04
mi = µi mu and ni are particle masses and densities.
Changes due to nuclear reactions:
∂Xi mi X X
= [ rji − rik ]
∂t ρ j
k

rji indicating that species i is created from species j,


and rik that i is destroyed to create k. In most cases
more than one input or output particle is involved, e.g.
p +12 C →13 N.
Energy released is ǫij = 1ρ rij eij . rij number of reactions
per second, and eij the energy released per reaction,
per particle mass it is qij = eij /mi.
In the simplest case, for the conversion of hydrogen
into helium, we get
∂X ǫH ∂Y
=− =− ,
∂t qH ∂t
eH,He ≈ 26.7 MeV/reaction = 4.72·10−5 erg and qH,He =
2.5 · 1019/4 = 6.44 · 10−18 erg/g.
Changes due to convection:

Usually, the convective timescale τc ≪ τn , and there-


fore the mixing can be assumed to be instantaneous.
The composition after mixing is simply the mean com-
position of the convective layer.

If there is nuclear burning happening in the convective


layer (e.g. in the convective core of a star), the change
in composition is
³P ´
mi
R P
∂Xi c.z. ρ j rji − k rik dMr
= R ,
∂t c.z.
dMr

(c.z. indicates the convection zone) but


∂Xi
=0
∂Mr
.

Complication, if convection zone moves into a chem-


ically different region.

Above equation is sometimes considered the 5th equa-


tion of stellar structure theory.
Changes due to diffusion:

Treated by diffusion equation; the general expression


for the diffusive velocity is:
1 ³~ ~ ln T + kP ∇
~ ln P
´
~vD = − D ∇n + kT ∇
c

D is the diffusion constant, n the particle density, and


ki suitable scaling constants (for D).
~
Fick’s law is ~jD = n~vD = −D∇n, where ~jD is the dif-
fusive particle flux.

And the continuity equation states


∂n ~ ~jD = ∇(D
~ ~
= −∇ ∇n) = D∇2 n
∂t

With corresponding equations for the other two terms,


and kT and kP describing relative diffusion speeds w.r.t.
concentration diffusion, the first equation is obtained.

The three terms correspond to concentration, ther-


mal, and pressure diffusion, the latter better known
as sedimentation. It often is the dominating effect.

The calculation of the diffusion constants is non-trivial,


because it involves interactions on the quantum and
electromagnetic level.
Estimate for diffusive timescale in Sun:
Diffusion is random walk; therefore distance ldiff re-
lated to number of scatterings N by

ldiff = λ N
with λ being the mean free path and N = ττdiffs .
τs is the mean time between interactions. From this
we can write
µ ¶2
ldiff
τdiff = τs
λ
λ is related to the particle density n and the interaction
cross section σs by
1
λ= .
nσs
Assuming elastic scattering by Coulomb-forces, σs is
of order 10−20 cm2 for hydrogen in the solar interior
and therefore λ ≈ 4 · 10−7 cm.
p
The thermal velocity of the particles is hvi = 3kT /m.
We can compute τs from
λ
τs = ≈ 8 · 10−15 s.
hvi

Sun: τdiff ≈ 1013 yrs for ldiff = R⊙ !


However, within the solar age ldiff is of order 1% of the
solar radius, and therefore small effects are present.
In general, diffusion is unimportant except for (a) high
precision models (Sun, high-resolution spectroscopy),
and (b) thin surface layers of high temperatures (hot
horizontal branch stars, young white dwarfs).
The equations - summary
m is the Lagrangian coordinate;
r, P, T, Lr are the dependent variables;
Xi are the composition variables;
~
ρ, κ, ǫ, . . . are physical functions, all depending on (P, T, X).

The four structure equations to be solved are:


∂r 1
= (9)
∂m 4πr 2ρ
∂P Gm 1 ∂ 2r
= − − (10)
∂m 4πr 4 4πr 2 ∂t2
∂Lr ∂T δ ∂P
= ǫn − ǫν − cP + (11)
∂m ∂t ρ ∂t
∂T GmT
= − ∇ (12)
∂t 4πr 4P
with ∇ depending of the type of energy transport:
3 κLr P
∇ = ∇rad = (13)
16πacG mT 4
∇ = ∇con (≈ ∇ad ) (14)
Finally, for the composition, we have (schematically)
 
∂Xi mi X X
= rji − rik  (15)
∂t ρ j k

which may include a diffusive term (here: representa-


tive for concentration diffusion)
· ¸
∂Xi ∂ ∂X i
= (4πr 2ρ)2D (16)
∂t ∂m ∂r
Solving the structure equations

• We have I+4 equations for the I species under


consideration plus the 4 dependent variables →
system can be solved!

• The equations are non-linear partial differential


equations in t and Mr (m). We need additional
boundary conditions at Mr = 0 and M (for the
spatial part) and intial conditions for t = 0 (for
the temporal part). Stellar structure is an initial
and boudary value problem.

• We call a stellar model the spatial solution for the


structure at a given time t0 (r(Mr , t0), T (Mr , t0),
. . . XI (Mr , t0)).

• Initial values: needed for all variables at t = 0.


How to obtain them? Possible ways: a previously
computed model, a simplified model, a good guess.

• If the initial model is a bad guess, but close enough


to the real solution, it will converge over τKH to
this.

• The problem can be separated into a spatial and


a temporal part:
Step 1 solve eqs. (9)–(12) for fixed t1 (Xi(Mr , t1 ) given)
Step 2 solve eqa. (15) and (16) for fixed spatial struc-
ture from t1 to t2 = t1 + △t using ǫ(Mr , t1 )
Step 3 update composition:¡ ∂Xi ¢
Xi (t2) = Xi (t1) + ∂t t △t → Step 1
1
Boundary conditions

at center: Mr = 0 → r(0) = 0, Lr (0) = 0

at surface: different possibilities

(1) “zero” b.c.: for Mr = M : P (M ) = 0, T (M ) = 0;


gives inner parts of stars approximately correct, but
outer parts are unrealistic; cannot be compared to
observations

(2) photospheric b.c.: b.c. taken at photosphere,


i.e. at optical depth τph = 2/3, where T = Teff

Stefan-Boltzmann law:
L = 4πσR2 Teff
4
(17)

This is the first photospheric boundary condition. It


relates three of the four dependent variables.

Temperature relation for gray ( ∂κ


∂ν
= 0) Eddington-
atmosphere:
3 4
T 4(τ ) = T (τ + 2/3)
4 eff

mass in atmosphere < 10−10 M⊙ → M (τ ) = M (τph)


We still need a second relation that involves pressure.

The optical depth definition is


Z ∞
dτ = κρdr → τph = κρdr
R

for simplification, replace κ(r) by κ̄, and τph = 2/3 →


Z ∞
2
= κ̄ ρdr
3 R
From the hydrostatic equation one can integrate the
pressure gradient to get
Z ∞
GM ∞
Z
Pph = gρdr = 2
ρdr
R R R
which results in the second outer b.c., for Pph :
2 GM 1
Pph = (18)
3 R2 κ̄

This is the second relation for the outer b.c., and the
system is closed and can be solved.

Eq. (18) can be obtained by more realistic integrations


using κ(r) = κ(P, T ) and g(r), or even by using non-
gray atmospheres. In general, gray atmospheres are
accurate enough.
Note on influence of outer b.c.:

If outer stellar layers are radiative, the solutions will,


independent of the outer b.c., quickly converge to the
identical inner structure. Hot stars have radiative en-
velopes.

If they are convective, in particular with ∇ > ∇ad , they


remain different for different outer b.c., and influence
the interior. In the extreme case (almost fully convec-
tive stars) the central conditions depend on the atmo-
spheric conditions. Cool stars have deep convective
envelopes.
Note on numerical methods for solving
structure equations:

1. Integrator-type methods (like Runge-Kutta)

Idea: A(x + △x) = A(x) + △x ∂A


∂x

Complication: b.c. not at one boundary, but two of


them at each.

Start from Mr = 0 and Mr = M using boundary val-


ues for two variables and trial values for the other
two. Integrate all equations from boundaries to some
matching point. If variables at matching point do not
agree, go back with new trial values. Iterate until
match is good enough. Fitting method

Advantage: does not need previous model; works well


for homogenous chemical composition; useful for ini-
tial main-sequence models and for pre-main sequence
models. Accuracy control of solution.

Disadvantage: slow or no convergence if steep gra-


dients are present; convective envelopes require ex-
tremely well known outer b.c.
2. Newton-Raphson type methods (like Henyey-method)

Idea: divide star in N shells, and define variables on


grid points; replace differential equations by difference
equations, e.g.
∂P △P GM̄r
→ =−
∂Mr △Mr 4πr̄4
The △P etc. are the differences between values at
neighbouring grid points and on the r.h.s. suitable
means over the shell between these neighbouring grid
points are used. Equations solved by Newton-Raphson-
type methods.

Advantage: reasonable convergence; can handle all


evolutionary phases (depending on grid resolution);
changing outer b.c. part of iteration process

Disadvantage: results in system of coupled linear equa-


tions, which requires inversion of 4(N − 1) × 4(N − 1)
matrix (for 4(N − 1) unknowns; N of order 1000); re-
quires good starting values for each model at each grid
point for Newton-solver to converge. (What about
very first model?)

Henyey-method: clever handling of matrix, which can


be split up into many 8x4 matrices. Method of
choice for almost all stellar evolution codes

Use model at t1 as first guess for new model at t2,


even if Xi (Mr ) have changed. Number of iterations
usually < 10, but can go up to ≈ 100.
Simple Stellar Models
Polytropes
Definition: relation between P and ρ of type P ∝ ρν
Polytropic relation:
P = Kρ1+1/n
K: constant of polytrope; n: polytropic index; both
constant throughout star.
The polytropic relation is not identical to an equa-
tion of state of that type! It requires only that T is
changing with radius in such a way that the polytropic
relation is maintained.
Examples:

1. Isothermal ideal gas: T = const1 → n = ∞

2. Adiabatic stratification:
T = const P ∇ad (in case of convection)
For ideal gas, it follows that
1−∇ad
P = const2 ρ1/(1−∇ad ) and → n = ∇ad
.
For ∇ad = 0.4 → n = 3/2.

A polytropic equation of state automatically results


in a polytropic stellar interior, e.g. the completely de-
generate electron gas:
– non-relat. complt. deg. → P = const ρ5/3 → n = 3/2
– relativ. complt. deg. → P = const ρ4/3 → n = 3

In this case, the constant K is known (from equa-


tion of state); if only stratification is polytropic, K is
unknown and depends on T !
Emden’s equation
With a polytropic relation the two “mechanical equa-
tions” (9) and (10) can be solved independently, by
combining them into one.
From (10), written in Euler-coordinates we get an
equation for Mr , which, when we differentiate it w.r.t.
r gives
µ 2 ¶
dMr 1d r dP
=− = 4πr 2ρ
dr G dr ρ dr

The left hand side is the Euler version of eq. (9) and
we obtain Poisson’s equation.
Using the polytropic equation, the derivative of P is
dP 1 dρ
= K(1 + )ρ1/n
dr n dn
and using this in Poisson’s equation yields
µ ¶ µ ¶
n+1 1 d 1
2 n −1 dρ
K r ρ = −4πGρ
n r 2 dr dr

With the following substitutions and definition:


1
K(n+1)λ n −1
r → αx ρ→ λy n α2 := 4πG

we get Emden’s equation

d2 y
µ ¶
1 d dy 2 dy
x2 = + = −y n (19)
x2 dx dx dx 2 x dx
Discussion of solutions of Emden’s equation –
Lane-Emden-functions

• λ is one free parameter; choose λ = ρc


³ ´
dy
• at center x = 0 and y(x = 0) = 1; dx =0
0

• R = αx0 corresponds to first root of y(x0 )


3−n
³ ´
2 dy
• M = −4C(K)ρc x0 dx
2n

x0

• for n = 3, M independent of ρc (important for


White Dwarfs)
• solved either2 numerically or by power series:
x nx4
y(x) = 1 − 6 + 160 . . .
• analytical solutions for n = 0 (see above),
n = 1: y(x) = sin(x)
x
, n = 5: y → 0 for x → ∞
• for n < 5 polytropes have finite radius
• if equation of state is polytropic, K is fixed, and
a mass-radius-relation results
• otherwise Lane-Emden-functions scalable with one
more free parameter
• from polytropes Chandrasekhar could derive the
critical mass (of his name) for completely degen-
erate stars, above which even electron pressure
can no longer balance the gravitational pull:
5.76
MCH = 2
M⊙
µe
17.7
Simple stellar models
Homology

for chemically homogeneous (i.e. undeveloped, zero-


age, main-sequence) stars.

Basic homology assumption:

Two stars are called homologous to each other, if


m1 m0 r1 r0
at M1
= M0
→ R1
= R0

this gives scaling functions:


¡m¢ ¡m¢
r M = Rfr M
¡m¢ ¡m¢
Lr M = LfL M
¡m¢ ¡m¢
P M = PcfP M
¡m¢ ¡m¢
T M = TcfT M

where fi are independent of M , but the constants (R,


Pc , etc.) are dependent on M and µ.

Assume simple laws for the physics functions:


R
P = ρT (20)
µ
ǫ = ǫ0 ρλ T ν (21)
κ = κ0 ρn T −s (22)
From the (radiative) structure equations we get simple
scaling relations, for example for pressure (x := m/M ):

dP dfP dx Pcx d ln fP 1 Pc P M d ln fP P d ln fP
= PC = = =
dm dx dm fP d ln x M Pc mM d ln x m d ln x

and equating this with the structure equation for P


we obtain the following relation, and similar ones for
the other variables:

dP P d ln fP Gm P m
= =− → ∼ 4 (23)
dm m d ln x 4πr 4 m r
dr r d ln fr 1 r 1
= = → ∼ 2 (24)
dm m d ln x 4πr 2ρ m r ρ
dT T d ln fT 3κ Lr T Lr
= =− → ∼ 4 3(25)
dm m d ln x 64πac r 4T 3 m r T
dLr Lr d ln fL Lr
= =ǫ → ∼ǫ (26)
dm m d ln x m
From (23) and (24) expressions for P and ρ as func-
tions of r and m can obtained; taking the ratio of these
and using the same ratio from the ideal gas EOS it
follows that
P m T
∼ ∼
ρ r µ
(or rT = µm) and with equation (25) for Lr
⇒ Lr ∼ µ4 m3
in particular, for x = 1 or m = M , we have
L ∼ µ4 M 3 ,

which is the mass-luminosity relation for main se-


quence stars. It does not depend on energy genera-
tion, but the proportionality is determined mainly by
opacity!
Since also Lr ∼ mǫ ∼ mρλ T ν , we obtain (again for
x = 1, using ρ ∼ m/r 3 and T ∼ µm/r)
ν−4 λ+ν−2
R ∼ µ ν+3λ M ν+3λ

For λ = 1 and ν ≈ 5 (pp-cylce) ⇒


R ∼ µ0.125 M 0.5

For λ = 1 and ν ≈ 15 (CNO-cycle) ⇒


R ∼ µ0.61 M 0.78

These are mass-radius relations for the two main


nuclear cycles on the main sequence. A representative
value for ν is 13, which gives 0.75 for the M -exponent.
This slide is about the main-sequence and lines of
constant radius in the HRD, and about MS-lifetimes
derived from the relations obtained above.

→ see 2nd exercise sheet!


Central values on the main sequence:
Set λ = 1 and µ=const.
P (x) T (x)
Pc ∼ P ∼ fP (x)
and Tc ∼ T ∼ fT (x)
ν−1
M M2 Pc
Also T ∼ R
, P ∼ R4
, ρc ∼ Tc
, and R ∼ M ν+3 ;

4
Tc ∼ M ν+3 (27)
2(ν−5)
Pc ∼ M − ν+3 (28)
2(ν−3)
ρc ∼ M − ν+3 (29)
2

Tc ∼ ρc ν−3
(30)

Tc is rising with M ; but ρ is falling with M for ν > 3


(M > 0.8 M⊙ )!
Two different types of stars on the main sequence
< 1.5 M⊙
M∼ > 1.5 M⊙
M∼
Teff low high
core radiatve convective
envelope convective radiative
H-burning pp-chain CNO-cycle
ν > 10 <7
gas pressure small significant
Theoretical and observed M -L-relation:

Theoretical and observed M -R-relation:

Theoretical log ρc -log Tc -relation:

You might also like