You are on page 1of 10

194 8MNRAS.108. .

324B

INWARD INTEGRATION OF THE STELLAR EQUATIONS

C. M. Bondi

(Received 1948 June 4)

Summary
A method of inward integration of the equations of stellar structure is
applied and extended to cover the cases of radiation pressure and electron
opacity.

1. Introduction.—A paper by Hoyle * gives a method of inward integration


of the equations of stellar structure. It is shown there that for a region in which
the transfer of energy is by radiation, the integration of the equations can, with a
certain degree of approximation, be reduced to a one parameter problem. The
purpose of this paper is to show how this method can be applied and to indicate the
results that it gives for a radiative envelope fitted to a convective core.
Comparison with the numerical results obtained by Cowling f and later by Hoyle
and Lyttleton J integrating outwards from the surface of the core can also be made.
It is shown that the approximations do not lead to serious errors, that the integra-
tion of the equations is stable, and that the method is one which may effectively be
used more generally. Thus, corresponding to definite assumptions as to com-
position and opacity, one is enabled to predict the luminosity and radius of a star of
given mass. It should be noted that in general, when the radiation pressure is
negligible, each solution of the equations corresponds to a one parameter series of
possible configurations with the mass as parameter and that, therefore, even
though the assumptions for composition and opacity may not be valid for a specified
star it is highly probable that it would be valid for some set of stars.
2. We assume in the present paper that the composition is uniform throughout
the star, with mean molecular weight ¡jl. We first take the case actually considered
by Hoyle §, namely a convective core with a radiative envelope in which the opacity
is given by Kramers’ law, radiation pressure being neglected in both the core and
the envelope. The Kramers’ law does not give the opacity due to photoelectric
effect with complete accuracy, as the guillotine factor is neglected, but it does give
the main variation, and the accuracy it gives is sufficient at this stage.
In Hoyle’s paper, a numerical constant ©2 is introduced, the significance of
which will appear later. A standard radius r2 is defined in terms of the radius of
* F. Hoyle, M.N., 105, 23, 1945.
t T. G. Cowling, M.N., 96, 42, 1935.
X F. Hoyle and R. A. Lyttleton, M.N., 102, 177, 1942.
§ Loc. cit.

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

No. 4, 1948 Inward Integration of the Stellar Equations 325


the star rs by
02-i
^2
The crucial assumption is now that with a suitable choice of @2, the mass
outside ig can be neglected This implies * that for r>r2 (in the notation of
Hoyle’s paper)

where z — r/r^ and T2 are the values of p and T respectively at r2.


In the radiative region r^>2, the following equations then apply
d
Ml = -A.2'®pK
dz 4 5Ü2
’ (1)
dM
= <}>pz2, (2)
dz
dT = _(=) r
dz 2
T^Z*' (3)

•where T =TjTi, p=pjp2 and M = MjM^ — MjMg. Here <f> is defined in terms of
the variables at r2 by
4^492
<f> = (4)

-while it follows from Kramers’ opacity formula, in the customary notation, that

(i -X^L p|
2
1677 acr2 (5)
tP
An additional relation is given by Bethe’s law. As is shown by Bethe f, the
energy generation depends so critically on temperature that throughout the
radiation region L(r) = L. Then

L = 47tE [r^f^dr (6)

taken throughout the convective core. _ _


We define n by the equation « +1 =íJ,(log P) íf(log T). Then
65
, GpMT 'pi
(« + l)©2 = =2 (7)
P2 r2T|'5
and, since « = 3-25 for
^ =:-P.
4-25® Gu Ms
2 (8)
& r2T2
Hence
n + i= 4-25MT6'5p-2. (9)
The actual integration of equations (i), (2) and (3) was carried out by means of
the differential analyser of the Cambridge University Mathematical Laboratory
and it was found convenient to express the equations in terms of T and the variables
sy Uy W defined by z = i¡Sy p=uTz'25y M= WT. s increases from 1 at r2 to the
value s1 at the surface of the core.
* Cf. Chandrasekhar, An Introduction to the Theory of Stellar Structure, Chicago (1938), p. 30c,
«equations (66) and (67).
t H. A. Bethe, Phys. Rev~, 55, 452, 1939.
23*

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

326 C. M. Bondi Vol. io8

The equations (i), (2), (3) then become


du dT
T-j-s +4-25«-^- =4‘2$@2uWT, (xo)

dW d(\ogT) <j>uT2™
(11)
ds+Wds~ i4 ’
dT
n 2 (12)
* =02M ’
or, in the form imwhich the differential analyser actually performed the integration,

w = ©2 J^2 JWd^uds— j^logT)] ,

W=-<f>jT^s-id juds - J Wd(log T),

T, = 0o \ud juds,

with u = T= W=i at s = I.
The integrations were performed with ©2 = 2. It is clear that less error would
be involved if 02 were >2, but taking ©2 = 3, the required ranges of the variables-
s and T become considerably larger and it was found that the increase in ranges
made the use of the differential analyser less appropriate.
The conditions at the surface of the core are that the mass, pressure, temper-
ature and temperature gradient should be continuous.
Thus (fl + i) must be continuous. In the convective core ft = 1*5. This
implies that the fit will occur when í = where ^ is defined as the point at which
the variables take the values Wx satisfying 2^=^2$W1T1u{^2i from (9).
The procedure is therefore to take a series of values of </> and for each value
find the corresponding sv The correct value of <f> will be that for which the
temperature gradient at s1 is continuous.
To find the temperature gradient at s1 for the radiative region in terms of the
Emden units, in which the solutions for the convective core are tabulated *, it is
necessary to find the radial coordinate x± in Emden units, corresponding to
T —
If y± is the Emden temperature at xly we have Tx =yx and ~ px =y\ where the
Po
suffix o refers to the central values. Hence, since

[S^To/SttG^poP
and from (4) and (8)
„2 _ 8-5^2
2
477G/XP2 ’
we have

(IS)

xx is found from this by successive approximation.


The temperature gradient of the radiative envelope at s1 in the Emden units is

-V
1 • (14)

* Cf. British Association Mathematical Tables, Vol. II, 1932.

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

No. 4, 1948 Inward Integration of the Stellar Equations 327


This has then to be compared with the y' of the Emden solutions.
The results obtained were as follows :—
Temperature
Temperature Gradient
$ l£1 X± Gradient (at sly
(at slf Emden) radiative
region)
*0*40 3-138 3-56 0-0493 0-546 1-220 — 0-3263 “0*3349
■0-405 3-072 3-51 0-0512 0*553 1*247 — 0-3302 —0*3303
0-41 3-016 3-47 0-0530 0*559 1*271 — 0*3336 -0*3254
0-42 2-945 3-40 0-0552 0-565 1-298 — 0-3372 — 0*3166
•°'43 3-853 3-32 0-0580 o*572 i*334 — 0*3418 —0-3122
0-44 2798 3-26 0-0601 o*577 i*357 -0-3441 — 0-3081
<5*45 2-722 3-19 0-0624 0-582 1*388 — 0*3481 — 0-3006
0-50 2-473 2-94 0-0722 o*599 i*5i4 — 0*3601 — 0*2694
«0-55 2*288 2-74 0-0812 0-617 1*637 — 0*3682 — 0*2450

Thus the required value of </> is 0-405.


The values of u, T for a number of values of 5 for (f> = 0*405 are given below :

5 u 5 T u
i i i 2*0 2*60 o*77
1*2 i*39 0-98 2*5 3*io o*66
i*4 i*74 o*93 3*0 3*47 o*57
i *6 2*07 o*88 3*07 3*5i o*55
1-8 2*35 0*825

The mass-luminosity relation resulting from these figures is considered in


Section 5.
3. Similar integrations have also been carried out for the case in which the
opacity is entirely due to electron scattering. That is, the opacity is equal to
*(1 -f-XT), where /c is a constant which in c.g.s. units is equal to 0*19.
The equation corresponding to (5) of Hoyle’s paper is therefore

dT* /c(i +X)pL


%ac
dr 477T2

In this case p/T3 tends to a limit at the surface and r>¿ is defined such that p is
proportional to T3 for r>r2. T& p2 are defined as before and M(r2)~Ms.
If we now put
4”jp2
<f> = (15)
M2 ’

a _ l*!1 +X)L p2 (16)


'-'2 —. ,
iSnac
"we get
G/x MY3 p, ©„-I
(w + l)©2 = (17)
p r2ff
and, since n = 3 when r = r2,

4©2=Æ-^. (18)
^ r2r2

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

¡28 C. M. Bondi VoL ioS

Defining z, p, T, Mas before and putting p=uT3,z — i/s, the equations become

Tdu/ds + ^u dT/ds = 8uM, (19)

dM/ds = —(f>u T3, (20)

dT/ds = 2u, (21)


for the case ©2 = 2.
These were solved by the differential analyser in the form

u = 4\(M-u)d(logT),

m=

T=2 Juds.

As before, the integrations were carried out for various values of </> and in this
case the fit on to the convective core occurred at the point sv where Mv %
satisfy the equation 2*5 =4M1/w1.
Expressing as before the equality of the corresponding radial distance, we have

*a = i &r0'5') ■
u T (22)

To express the continuity of dTJdr, we have to equate

'5%y
yti
( 4<f>/ n
to the value of the temperature gradient given by the Emden tables at
The results obtained were as follows:—
—Temperature
—Temperature gradient
Wl *1 gradient (at si,
(at Emden) radiative
region)
0*5 2-67 3*39 0-293 0-469 1-165 0-318 0*504
0*6 2*21 2-84 0-323 0-519 1- 0-349 0-415, 3
0*68 1-99 2-55 0-344 0-551 1*564 0-364 0-359
07 i*95 2*50 0-348 0-557 i*599 0-366 0-350
o*8 i*79 2'28 0-371 0-593 i -8o6 0*374 0-293
o*9 i*66 2*10 0-390 0-622 2- 6-372 0-235 0
The required value of is therefore about 0-67 and by interpolation we find
that s1 = 2*oi, ^ = 2-58, Mi = 0-342, % = 0-548.
The values of M, w, T for a number of values of s for <£ = 0-67 are given below :

s T u M
i i i i
1-2 1- 0-940 0-850 4
i*4 i*75 0-836 0-692
1-6 2- 0-730 0-550 0
1-8 2*34 0-631 0-434
2*0 2*57 o*553 0*345
2-01 2-58 0-548 0-342

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

No. 4, 1948 Inward Integration of the Stellar Equations 329


4. When radiation pressure is included, the equations are (using electron
opacity)
d ImpT\ GpM
dr\ ¡iß ) r~ (23)

dM
= 477r2/), (24)
dr
df 3k(i+X)L p
dr lÓTiflcr2 T* ’ (25)

These equations can be treated in a similar manner, provided that dßfdr is neglected,
where ß is the ratio of gas pressure to total pressure. This does not in fact
involve treating /? as a constant but merely that one assumes that the term in which
its derivative occurs is negligible.
With this approximation equation (23) becomes
d IMpT\ GpMß
(23')
rfr \ /x / r2
Here
i ua T*
-7; =1 +
3a (26>
- i'
Defining r2 as before and using the same variables T, u, M and j’, the equations
(23'), (24), (25) become
^ / 'ri\ rwi. \ Mu2T3
(27)

dM .uT3
(28)
dy ^ y4 ’
dT
n
(29)
where
j 477’P2r2
9= " TV~> (15)
Ms
3K(i +X)p2L (16)
©,=
i67Töcr2T|
and

h- (,0)
l30
/3(rs) • '
Putting ©2 = 2, the equations (27), (28) and (29) were solved on the differential
analyser in the form

M= -

T=2 juds,

with the initial conditions u = M= r = 1, jS = i/(& +1) at 5 = 1.

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

330 C. M. Bondi YoJ. xo8

In this case

(-ï) Mß
(k + i) — d — =4(ä + i)
¿(log T) ■' ' u
and from (27) and (30),
ß=ul(u+k). (31)
At the surface of the core, ß is continuous and the effective polytropic index is
3(8-7fc)
8-3ß1~3ßr
where ß = ß1zt s=
Thus the surface of the convective core occurs at the point s=.?, at which
M
3(8-7 ßi) +I=4(Ä + I) A
8-3ft-3/3? u,
or in terms of u, M only,

5^i + 4Q%^ + 3^2 _ 4M1(k + i) _


(32)
+ 13%^ + 8Ä2 Ui + k 1
V yj*
As before, this point ^ is found for a series of values of <£. In the core, we assume
that ß is constant throughout. The justification of this is that in fact ß varies like
ys-n por smaii radiation pressure, the radius of the core is small and so the
variation of T through it is also small. For large radiation pressure, although the
radius of the core is large, the value of n is very near to 3, so that the variation of
ys-ji js again small.
The Emden radial coordinate xly corresponding to is thus given by
/8<j>(k+i) m? TJ y / y
11 1 1 (33)
\ Wj + I u1 + k yß ^ )
The temperature gradient at s1 in the radiative region, when expressed in terms
of Emden units is
si f2M1y1n'+1\*
T*{ <(> ) ■
The correct value of is that which makes this expression equal to the y' of
the Emden tables for the corresponding nv
This was carried out for ß(r2) equal to 0*9, o*8 and 0*5, i. e. for &= 9, | and 1.
The solutions for intermediate values of ß(r2) can clearly be found by interpolation.
The results obtained were as follows :—

k=1

Mx Hi X!
© -©
Emden radiative
0-4 2-50 3'3o 0-432 0*552 i-o8 0-291 o-68o
0-6 i*8o 2*34 0-489 0-645 1*56 o*339 0-427
07 1*65 2-10 0-507 0-675 1-79 40-343 0*351
0-71 1*64 2*08 0*509 0*678 1- 0-342 o*343
0-8 i‘55 i*95 0-521 0-700 2- o*335 0-258
Thus the fit occurs for 0 = 0-71.

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

No. 4, 1948 Inward Integration of the Stellar Equations SSI


k—\

ux *1 -m
Emden radiative
°*4 2*01 271 o*557 o-66o I -21 0-303 0-641
1 2*24 0*706 0-568
°-5 '74 0*582 I-4I 0-321
0-6 i*59 2-04 0*595 0-730 1*70 0-331 0-424
0-7 i*49 1-88 o-6io 0755 I *98 0-325 0-326
0'75 i*45 i -8i 0-617 0-770' 2*24 0-310 0-252

Thus the fit occurs for <f) = 0-70.

k=i

1^ *1 (?) -(?)
Emden radiative
o*4 1*435 1*82 0-789 0-859 i*34 0-302 0-765
0-6 i -29 1-56 0798 0-888 1-87 0-305 0-515
0*64 1*275 i*53 o-8oo 0-893 2*40 0-269 0-271

The required value of f is thus 0-64.


5. We now consider the numerical values that our results give for the mass-
luminosity relation and the relations holding between radius and mass. We also
obtain expressions for the central density and temperature. These relations are
in general dependent on Bethe’s law of energy generation (6), which has not been
introduced into the integration up to this point.
(6) may be written in the form

L=+ i)f0I^G^0}’ J" x2 {f)}“ dx.

The integral is almost independent of n± and is equal to 0-07, approximately.


(i) For Kramers’ opacity (j81 = i) using equations (4), (5) and (8),

l2"+5=/0-2877 fetiy
\I7©277/ yV+i’i + îj
/47t\^+12.5 / GfX\ 1^+45 / l60277aC \ 2??+6 —
X Mi0"+31.
W/ \4'250o-^y V3«(i-v2);
Putting 0a = 2, <f>= 0-405, Wj = i'5, 7]= 17*25 and inserting the values of p1;
and yv we have
L = [i39'4939]^7 25(i (34)

Using (4), (5) and (8) we get also

r, = [290-7438]^(i (35)

r2=[16-2783]^ ^-1
8 > (36)

p2 = [í-4ii2]Msr¡-3. (37)
Since
T, P2 *
T
i=yi and =rPi=yi>
Po

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

332 C. M. Bondi Yol. 108

(36) and (37) lead to


To = (38)

/>o = [0-9615] Ai>s3- (39)


It is of some interest to note that the correct values of the luminosity and
radius of the Sun are obtained only if /u, = 0*807 and X = 0*967. With this value of
X and at the relevant temperatures, the Kramers’ opacity law would certainly
apply. If the conventional assumption that the material other than hydrogen is
either helium or heavy elements ionized with mean molecular weight 2, these
values of and X are inconsistent. However, if in fact the assumption of uniform
composition is invalid, this result may be some indication of the mean molecular
weights at the centre and in the outer regions, since X as used in our formulae refers
throughout to its value in the radiative region, whereas /x refers also to the central
regions, entering as it does into Bethe’s law.
(ii) For electron opacity, we put X~i and therefore /x = |. For any appreci-
ably lower value of X, the electron opacity would not be predominant unless the
temperature in the radiative region were of a higher order of magnitude than the
theory indicates.
In this case L can be expressed as a function of Ms without using Bethe’s law.
In fact

Using Bethe’s law to give the radius, we have

r-^ (K+lM* «ITT3 / 2^ y /3,c^(i+X)Y>±g^


\ 47rG/xj81 J VttG/xjSj/ \i67r2acG/jLß1 ) ?2+îi*
Inserting numerical values, these two equations lead respectively to

L = [66-2150] (40)

16-25
5 ó
L 3
=[17-6316] vtiY' ( W T—20,25 (4i>
. h ) \ßj
We have besides these the conditions expressed by equations (15), (16) and (18).
The approximation of zero radiation pressure is valid for all sufficiently small
masses, i. e. equation (26) is approximately satisfied for all such masses, but for a
given non-zero ratio of radiation pressure to gas pressure, equation (26) determines
the mass uniquely.
Using also the equations (Ta/To)^and (f>2lpo)pi=ÿi> we have for k — o
(when = 1*5)
L = [66-3889] M3, (42)'

rs = [v-8652]M<r, (43>

T2 = [16-0036] Msr~ \ (44>

p2 = [ï-6300]Msrs-3 (45)
and
^0 = [0-5877] ñ, (46)

Po = [x-2377]P2- (47)

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System


194 8MNRAS.108. .324B

No. 4, 1948 Inward Integration of the Stellar Equatio?is 333;


34
These are applicable for Ms<4-io g.
In particular for the Sun, Ms = [33-2978] and
¿ = [34-2823] ¿ = [33-5775]
compared with observational values
rs = [10-5856] rs = [10-8421]'
For k v-o, the results are given in the table below :

k=i *=i
M, [35-657I] [35-0693] [34-8I96]
[40-1759] [39-1902] [38-6394]
[12-2545] [11-9127] [II-8224]
[7-6491] [7-6067] [7*5079]
Pc [7-8633] [0-2269] [0-23I9]

6. Comparisons with previous integrations.—The only numerical comparison!


that can be made with previous integrations are concerned with the Kramers’
opacity. Interpolating in the Hoyle-Lyttleton solution * to get the point corre-
sponding to r2 for @2 —2> ^ turns out to be 0-556. Although the agreement
between our value of <f> and this one does not appear to be very good, it must be
pointed out that for ©2 = 2-35, the Hoyle-Lyttleton solution gives ^=0-318.
The value of <f> is thus very sensitive to changes in 02 and hence is accurately
determined only if the values of the variables at r2 for specified 02 are accurately^
determined. This is the weakest point in the connection between the methods of
outward and inward integration. The agreement between the radius of the core
(1-19 in the Cowling and Hoyle-Lyttleton solution and 1-25 in the present solution,
expressed in Emden units) is extremely satisfactory in view of the degree of approxi-
mation used here. In any case, the method of inward integration is far less
troubled by instabilities and is therefore much more straightforward to use.

In conclusion, the writer wishes to express her gratitude to Mr F. Hoyle for


much helpful advice, and to Mr M. V. Wilkes, Director of the Cambridge Univer-
sity Mathematical Laboratory, for placing the facilities of the Laboratory at her
disposal.

Newnham College,
Cambridge :
1948 June 3.
* F. Hoyle and R. A. Lyttleton, M.N., 102, 177, 1942.

© Royal Astronomical Society • Provided by the NASA Astrophysics Data System

You might also like