You are on page 1of 20

1977ApJ. . .212. .

513M

The Astrophysical Journal, 212:513-532, 1977 March 1


© 1977. The American Astronomical Society. All rights reserved. Printed in U.S.A.

THERMONUCLEAR ION-ELECTRON SCREENING AT ALL DENSITIES.


I. STATIC SOLUTION
Henri E. Mitler
Center for Astrophysics, Harvard College Observatory and Smithsonian Astrophysical Observatory
Received 1976 May 4
ABSTRACT
Thermonuclear reaction rates are accelerated in a plasma over their vacuum values, by the
factor/s. The usual expression for it, In fs = H12($) = AU/kT, is shown to be valid only in lowest
order—i.e., in the weak-screening regime ; higher-order corrections are found here. That is, a new
expression is found for /s, in the fluid approximation, which agrees in lowest order with those
found by Salpeter. At high degeneracy (strong screening), the corrections are very large. Graboske
et al. have derived H12(0) (the lowest-order term) more generally, i.e., without the fluid approxi-
mation; the higher-order terms nevertheless must modify that calculation as well. It is also found
that the Graboske et al. results are not valid at temperatures below which the plasma crystallizes.
This is the region where the Salpeter-Van Horn results must be used. Although the present
results are better than those of Graboske et al. in this region, they also are inadequate there.
However, they generally merge quite smoothly with the Salpeter-Van Horn values. Finally, these
results are used to find a new carbon-ignition curve; it is found that ignition of a carbon core
would occur at somewhat higher densities and temperatures than calculated before, but not
enough to resolve present difficulties with explosive carbon ignition. A correction factor is also
given for resonant reaction rates.
Subject headings: nuclear reactions — plasmas — quantum mechanics — stars: interiors

I. INTRODUCTION
Nuclear reactions in stars take place at energies in the kilovolt range. This is much below the Coulomb barrier,
and therefore the reactions only occur through quantum-mechanical tunneling through this barrier. The reacting
ions polarize the plasma, however: electrons are attracted and ions repelled, into a statistical equilibrium distribu-
tion around each ion. Hence the potential seen by a reacting particle is modified from the Coulomb potential ; in the
Debye-Hückel approximation, it is modified to a Yukawa potential. This is a thinner barrier than the Coulomb one,
tunneling is easier, and the reaction rate increases over what it is in vacuum. The ratio of the actual rate to that
which would prevail with a Coulomb potential is called the “screening factor,”/s (it might more appropriately be
called the “acceleration factor”). The classical calculation offs was made by Salpeter (1954, hereafter referred to
as S). It is quite good at low densities, but poor for very high densities. Many attempts have been made to improve
the calculation in the high-density (strong-screening) region (see, e.g., Schatzman 1948; Cameron 1959; Harrison
1964). They generally have had the virtues of elegance and simplicity, but are inadequate because of the nature of
the various approximations made. This is clear from the fact that they give quite different answers. Wolf (1965)
and then Salpeter and Van Horn (1969, hereafter referred to as SVH) made careful calculations of the acceleration
factor in the regime where the medium, including the reacting nuclei, forms a crystal lattice; the SVH calculation
is correct, but even in that paper they continue to use the S result, with only small corrections, for the strong-
screening but noncrystalline regime. The definitive calculation of Ff12(0) = àU/kT (where At/ is the effective
potential-energy shift due to the polarization of the plasma) is that of DeWitt, Graboske, and Cooper (1973) and
Graboske et al. (1973), hereafter referred to as DGC and GDGC, respectively. However, they also make the standard
assumption that In fs = i/i2(0) ; this is not correct, as discussed in § II and shown in detail in the rest of the paper.
The present calculation is carried out in the usual way—i.e., with the two-fluid approximation to the plasma.
However, the potential and the various integrals involved are calculated more carefully than usual, and that suffices
to give reasonable answers. A check on the correctness of the present calculation is given by the fact that in the
cryogenic region it agrees with the SVH crystalline result better than other calculations do. The rest of the paper is
organized as follows : A very brief outline of the standard treatments is given in § II, with an indication of their
inadequacies ; the equilibrium potential and charge density around the target nucleus are found in § III ; the resulting
interaction energy is found in § IV. In § V we derive the action integral therefrom, and in § VI the reaction rates
and resulting screening factors. Explicit results and comparison with previous authors’ work are given in § VII,
while the inadequacies of this treatment are considered in § VIII. The screening for resonant reactions is given in
§ IX, and a summary and discussion of the astrophysical significance is given in § X.
513

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


1977ApJ. . .212. .513M

514 MITLER Vol. 212


II. STANDARD CALCULATIONS
For nonresonant nuclear reactions between nuclei of charges Zx and Z2 at low energies £ (i.e., below the Coulomb
barrier), the reaction cross section has the form

<J{E) = Ö

where

and S{E) is a slowly varying function of E, generally referred to as the astrophysical “constant.” The function 7j(E)
arises from a calculation of the penetrability of the Coulomb barrier; if the penetrability is calculated in the WKB
approximation, then for s waves we can express it in terms of the action integral for the Coulomb potential :
(2m)112 1/2
SC(E) dr = 77(77 — 28 — sin 28) ,
ñ
where
sin 8 = (R/Rc)112 ,
R being the nuclear radius and Rc the classical turning radius. So long as R « Rc, 8 « 77/2, and SC(E) is well
approximated by
SC(E) £ 7777 — 4778 = 7777 — £, (1)
with
£ ä 0.52495(ARZ1Z2)112
(A is the reduced mass of the nuclei in amu, R is in femtometers). Then setting
_ n&z 1/2
n
012 1 (kT) -3/2
“ 1 + S12 i-
\7rmJ)
the unscreened reaction rate is (see, for example, Clayton 1968, page 300)

r 12(C) = ôi2 JJ $(E) exp + 2SC(E) + 2£ dE. (2)

This integration cannot be performed analytically, so one usually resorts to the saddle-point method. This yields

r„ « %e-*=S(£.)[l + erf + 2S,(£.)] (3)

where Em is the position of the peak of the integrand. (In the literature, the integration is usually extended from
— 00 to +00, effectively taking erf =1.) The peak for S(E) = SC(E) is at the Gamow energy,
Em = E0= [77Z1Z2e2fcr(2m)1/2/2Ä]2/3 = 1.2204(Z1Z2r6)2/M1/3 keV.
For the screened reaction rate, we have precisely the same integration, but with the action integral calculated
with the screened potential. The latter is found in any of several ways: thus, assume statistical equilibrium; then
the particle density for the /th species, at the distance r from a given nucleus Z^ is
«iO) = «i exp [—Z¡e</>(>)¡kT\. (4)
Then the Poisson equation yields
V2<£ = — 477/>(r) = — 477Z1e8(3)(r) — 477e ^ Z^r) , (5)

where is the number density of the ith ion species. (The sum includes Z = — 1, for the electrons.) For large r,
</>(r) is small, and keeping only the linear terms in the expansion of the exponentials is adequate. It follows that
V^äicoV, (6)

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


1977ApJ. . .212. .513M

No. 2, 1977 THERMONUCLEAR SCREENING 515


where
kT ]l/2 1/2
*o 2 8.8924 x 10 -gl" zV-eT* cm
bm
‘ = [: e e (Z* + 1) _
Lp( * + i)J
is the Debye length, and i¿e is the mean molecular weight per electron. The mean electron number density is nei and
Z* is the mean charge number in the plasma, defined by

z* = 2 zfah Zitii.
i>0 f i>0
The appropriate solution of equation (6) is of course
<l>(r) = (Z1elr)Qxp(-K0r) . (7)
It is then found, however, that the action integral cannot be evaluated; hence <£(r) is itself approximated by expand-
ing the exponential and taking only the first two terms :
<£0) £ Zi^/r - Z^kq . (8)
This permits a straightforward integration, and it is immediately found that the screened reaction rate is given by
r12(S.C.) * r 12(C)eAL7//cr ^ r12(C)/s, (9)
with
AU = Z1Z2e2KQ .
The Debye approximation is inadequate for small r, however, and the linear approximation to that (eq. [8]) is
even worse. In particular, we note that the approximation (x_ 1 — 1) to 1 ¡xex yields a barrier which is too thin, and
hence gives too large a screening factor. These errors become very large in the strong-screening region—that is,
at high densities the Debye sphere becomes small and the approximations break down. Nevertheless, Salpeter
(1954) assumed,/atzte de mieux, that fs is still given by equation (9), but with AU found in a more appropriate way:
He took AU to be the difference in electrostatic energies between an initial configuration where the nuclei are
widely separated, and a final configuration where they are fused.
In fact, the appropriate energy should be the work done in going from the initial to the final configuration. This
is easily shown in the weak-screening case : When we discard the infinite self-energy of the nucleus (assumed to be a
point), we find that the Coulomb energy of the charge distribution around a nucleus Z1 is, in the Debye approxi-
mation,
w1 = i J p(r)4>{r)dT = -¿(Z^/cc, • 00)

Hence the electrostatic energy difference is


W12 = fZiZÄ,
which is 1.5 times the correct result. In bringing these ions and associated plasma clouds together, however, the
work done is the difference in the Helmholtz free energies, F:

F= r
ideal gas 0D

But AFiaoal a: 0, and the same holds for the exchange energy and the correlation energy. Hence, since equation (10)
implies AtF = Kj^T, we can immediately integrate, and find
AF = ^AW = Z1Z2e2K0 ,
precisely the desired result.
Salpeter (1954) used the ion-sphere approximation for a highly degenerate plasma. That is, each ion is im-
mersed in a completely uniform sea of electrons; the other ions are taken to form a uniform positive background,
which precisely cancels the negative background of the electrons, outside a sphere around the ion. This simple
model yields
9 Z2e2
Wsal = (12)
10 Rx ’
where
Fx = (3Z1l47meY13 s RZ1113 (13)

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


516 MITLER Vol. 212
is the radius of the Wigner-Seitz sphere about Zx. Since Rx is independent of T, so is lFSal, leading inunediately to
AF ä LW, with

=
“TO[(Zl + Za)5 3
' " Zl6/8
~ Z2<5 3]
' •
Hence Salpeter’s use of the electrostatic energy difference (see, e.g., SVH, § Ulb) is justified.
Cameron (1959) used this ion-sphere model, instead, to obtain a potential, analogous to equation (7), in the high-
degeneracy limit; this yields the screened potential

0 < r < jRx .

The interaction energy with a charge Z2 then would be


ZiZ2e2
Z2e<t>i(r) =

He drops the cubic term, which is generally small, and thus obtains a value for At/. Harrison (1964) took the cubic
term into account in an approximate way by adjusting the product Z1Z2e2 appropriately. In either case, however,
it is clear that the product Z2e<l>1(r) is not the same as Z^^r), as it ought to be. Hence those calculations cannot
be correct. We shall return to this lack of symmetry in § IV.
In the rest of this paper we show that it is possible to derive a reasonably good expression for fs even at very
high densities (where electron degeneracy is very high), still within the two-fluid approximation to the plasma.
We do this without making the assumptions that A{7 (in ln/s = AU/kT) is given by AF, or that Z± and Z2 » Z.
We begin by making a better calculation of the charge density p(r) and electric potential around the ion Z^

III. THE POTENTIAL


Model A. We again assume statistical equilibrium. Hence equations (4) and (5) still hold, with a small modifica-
tion for degeneracy (see eqs. [18] ff.). For large r the linear approximations are still valid, so that
V2^ Ä K2<f> .
The appropriate solution at large distances is <£0utO*) = Cee~Kr¡r. For small r the ion density goes strongly to zero,
according to the nonlinearized equation (4), whereas the electron density diverges. Of course the electron density at
the origin is in fact finite. We make the simple approximation that ne(r) x ne = ne(oo)—i.e., the same assumption
of uniformity as made in S. Indeed, this turns out to be quite good, as we shall see. Thus we assume that there
exists some radius r± such that
PmOO » Ztf^Xr) - ene (r < rj (14)
and
K2 k2Cp p~Kr
Pout(0 ~ V (r>ri). (15)

This density distribution and the results flowing therefrom constitute model A. Next, we match the values at rx
and normalize the density. We then easily find

*1 = K/l = - 1

and
47mer1
c= exp (k/i) . (16)

The Poisson equation with p(r) given by (14) and (15) has the solution
fair) = Z1e/r - C0 + Cxr* (r < rj

= ~~ e~Kr (r>r1). (17)

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


1977ApJ. . .212. .513M

No. 2, 1977 THERMONUCLEAR SCREENING 517

Fig. 1.—The various charge densities associated with example 1, versus x = kv. Dashed curve, p0(x)9 the initially assumed
density; it is normalized to unity at the origin. Solid curve, the Stewart and Pyatt result, p(x); pD(x) is the density resulting from
Debye-HUckel theory. Dotted curve, ps(x), the Salpeter ion-sphere assumption.

where

Ci = frrewo and C0 = ^(2 + x±) .

This may be compared with the Debye solution: we see that near r = 0 this has just the form of the linear approxi-
mation (8) to the Debye solution, but where the energy shift is

E z eC Z Z
- - ‘ ° - ‘ H rbn) ’
rather than AU.
We can see how this density distribution compares to the Debye and the ion-sphere distributions, for a particular
example: Suppose we have a supernova shock in a carbon star, with p = 2 x 109 g cm-3, T = 3 x 108 K. We
have = 2, /c-1 = 19.88 fm, x± = 5.721 (i.e., r± = 114fm, comparable to the interionic distance, 134 fm), and
C = 103.9, or 17.32 times Zx = 6. The various densities are plotted in Figure 1 ; p0(x) is the initially assumed density
(given by eqs. [14] and [15]), pB(r) is that obtained from Debye theory, and ps is the Salpeter ion-sphere approxi-
mation. It is apparent that the length k~1 takes on a meaning different from the usual meaning of “Debye length” :
as seen from equations (14) and (15), it is the range of the effective charge density beyond r-^. Rather than speak of
the Debye sphere, we note that the fluid (statistical) approximation is valid if the number of particles in the shell
of radius r1 and thickness /c“1 is large. Let us call this the Debye shell. The mean number of ions in a shell of
thickness /c-1 around r1 is ^(47rr12)/c"1«i. With Vx given approximately by Rx (see eq. [13]), the expected number of
ions in the shell is
_3_Zi
<N>
Z ’

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


518 MITLER Vol. 21
where Z is the mean atomic number in the plasma. For the fluid approximation to be valid, we must have <iV> » 1.
For this example, <V> ~ 0.33. Thus the fluid approximation can no longer be justified here, and statistical fluctua-
tions become important. This is discussed in § VIII, where model B is introduced. Even though the fluctuations are
important in such cases (where the degeneracy is high and we approach the ion-sphere configuration), let us
nevertheless use this formulation as a zeroth-order approximation, as did Salpeter, Cameron, Schatzman, and
Harrison (Salpeter was careful to note that the condition Zi » Z is required in order for his expression to be valid).
Next, we see whether a significant improvement can be made over p0(r). Stewart and Pyatt (1966) have solved the
problem of the charge distribution around an ion more precisely: They assumed a Thomas-Fermi distribution
around the ion Zl5 and found

ne(r) = neF1/2j^||jp + Fuzi?) ? (18)

where a is the degeneracy parameter and Fu2 is a Fermi integral. Equation (4) still holds for the ions. Define
y = e<l>(r)lkT; it is then easy to show that

^ - e‘z‘’] • (19)

This is the Stewart-Pyatt result. When the electrons are strongly degenerate, the asymptotic expression for F1/2(a),
F1I2(*)xW2, (20)
yields
a « EF/kTy> 1.
When y « 1, we may take just the first two terms of the Taylor expansion of F1/2(j + a) about a:
Fli2(y + a) = Fi/2(a) + ja + • • • , (21)
where it is not difficult to show that
ä=F-m(p)l2Fll2(a). (22)
(This has been done in an equivalent form by Kidder and DeWitt 1961.) When Z*y « 1, we may then expand
equation (19), and find
p(r ) = -eney(Z* + a).
Equation (15) is then satisfied if we define
47re2ne
K2 (Z* + ä) =
kT
For low degeneracy, Æ £ 1, while for high degeneracy, ä £ 3/2a. We see that at high electron degeneracy, when
a « 1, the relaxation length /c_1 is somewhat increased over the Debye length kq"1; this is because the electrons’
effectiveness in screening is very much reduced—but not that of the ions. (Stewart and Pyatt erroneously use k0
in eq. [16]—that is, Z* + 1 rather than Z* -f ä—and this leads to error in the calculation of Es, which [as we shall
see] would affect the calculation offs substantially.) Finally, we can give a simple approximation for a: it is easy
to show that for a completely degenerate electronic distribution, the exponent e<f>¡kT in equation (4) becomes
Hence for a partially degenerate distribution, one might guess that this term becomes e<j>[(kT)2 + (2£'F/3)2]“1/2.
Comparing this with equations (18), (21), and (22), this is seen to be equivalent to approximating ä by :

« ~ a* = [i + (Ü)2]”1'2 = {i + ít^W«)]4'3}-1'2 •

This approximation is in fact correct within 2% over the entire range.


Finally, we compare p{r) as given by (19), with po(r). For the example chosen above, a = 1006; that is, the
electrons are extremely degenerate. Then ä # 0.00149 and /c-1 -> 19.88 fm. We note that (19) is highly nonlinear.
However, if we start with the density given by p0(r)—i.e., equations (14) and (15)—we obtain the potential given
by equation (17); this is used to find y(r), and that can in turn be used in equation (19) to give an improved pi(r).
This process can then be iterated, and quickly converges to p(r). We do this,1 and show the various densities in
1
The p(r) shown in Fig. 1 is actually pi(r), the result of iterating only once.

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


1977ApJ. . .212. .513M

No. 2, 1977 THERMONUCLEAR SCREENING 519


Figure 1 : We see that, except for small regions near the origin and near xl9 the assumed distribution p0(x) is already
in excellent agreement with p(x). At the origin, p(x) diverges—this is an artifact of Thomas-Fermi theory. In fact,
of course, the actual density there, p(0) = — ene(0), is easy to find : we solve the Schrôdinger equation for an electron
in a Coulomb field, find its “penetrability,” average over the (relativistically degenerate) distribution, and readily
find that the buildup of electrons at the origin is only 14% : ne(0) = 1.144«e. This result is generally true, even for
low degeneracy—that is, the deviation from uniformity (out to rj is small. Thus we have shown that p0(r) is
indeed an adequate approximation to p(r), and we shall use it hereafter. It follows that <£(r), as given by equation
(17), is also adequate. Thus, we have found reasonable general expressions for p(r) and <£(r). We must next find
the interaction energy.
IV. THE INTERACTION ENERGY
If Z2e<l>(r) were symmetric in 1 and 2, we could go ahead and use it as the interaction energy between ions Z± and
Z2. This is not the case, however, just as it is a difficulty also for the ion-sphere potential. Only in the Debye case
(the small-x limit) do we have symmetry. This lack of symmetry arises because the potential (17) is that seen by a
(hypothetical) infinitesimal test charge which carries no associated plasma cloud with it. When a real (finite)
charge Z2e approaches the ion Z^, their polarization charge clouds do not simply superpose, as in Debye theory,
but combine into a common (and larger) cloud. The work done by the particles in approaching each other is
AF(r), the difference in free energies, and thus to lowest order the interaction energy is
U12(r) X + AF(0) .

The free energy. We use the results for p{r) and <£(r) from § III (that is, eqs. [14], [15], and [17]) to find the electro-
static self-energy of the configuration:
r qc 2y 2
W, = i J PiirW^dr = (f*!3 + 2x±2 + 4Xl + 3) , (23)

where we have discarded the infinite self-energy of the point ions. For a classical (i.e., nondegenerate) gas,
3K3Z1l47rn0 « 1, so *! « 1, and equation (23) yields
WiX-UZ^)2*,
the Debye result. For the extremely degenerate case, on the other hand, x± » 1 instead, and
9Ci2 2 5 _ 9 ^Zi5'3
JFi Ä — = 10 R
where R is the mean interelectronic distance (see eq. [13]). This is the standard (Salpeter) ion-sphere result. For the
example in §111, the exact equation (23) yields Wx = —0.1331 Z12e2K, whereas equation (12) yields IFSai =
— 0.1340Z12e2/c, which is almost identical to Wx. Thus, we can use JFSai, rather than the complicated JFl given by
equation (23), in the ion-sphere regime. It is appropriate to note here that JFSal is a good approximation to by
virtue of the variational principle : a first-order error in the wave function \¡j results in only a second-order error in
<#>• . . t
Next, we need to find the free energy from this, as given by (11). If we convert to #c3 as our independent variable,
this is integrable, and we find

Fl=
-y^r[^ + 1)6,3-i-Ki],
where

^= (24)
4irn0
This is found to have the proper limits for both large and small Ci- Hence
10/^2
AF = F1+2 - F, - F2 = [((, + (2 + l)5'3 - + l)6/s - (L + l)6'3 + 1] • (25)

This is the generalization of Salpeter’s expression, and is an approximation to the general expression, equation (24)
of DGC.
For K small, so that ^ £a « 1, equation (25) yields the Debye solutions; while for Çj, £2 » 1 we recover the
Salpeter strong-screening solution. Only in the Debye limit do we find AF = — C0Z2e, as would be obtained from

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


1977ApJ. . .212. .513M

520 MITLER Vol. 212


Z2e<¿i(0), from equation (17). The quantity C0Z2e is the shift in potential energy due to the screening charges
around and is what an infinitesimal test charge would experience. The real charge Z2e, however, modifies the
ion-electron distribution around Zi ; therefore, as the ion Z2 approaches Zj, it changes C0Z2e (continuously) to
— AF. This is easily seen: assume Z2 « 1, so that (2 « 1 ; then equation (25) leads to
2
AF= _^£2[(1 + £1)2/3- 1],

and this is precisely — CoZ2e.


We thus have the first two terms in the expansion of the effective interaction energy:

t/i2(r) = ^l2 + AF(0).

In order to obtain the next term in the expansion, we should carry out a two-center calculation. We have not done
so, but we may get a good estimate of the next term by noting that for small separations r between Zi and Z2, their
common cloud is distorted, in a manner analogous to a fissioning nucleus. Thus, assume that the distorted charge
cloud is equivalent to a uniform charge density distributed as an ellipsoid of revolution. Then for a charge Z, radius
R,

R(6) = f[i + 2 únicos 0)1 • (26>


L n=0 J
If we make an isovolumic change in the shape, then the Coulomb self-energy of the distribution is
„ 3 Z2e2 (a.
E =
' 5-Rg(e)’
2
where to lowest order g(0) £ 1 — fl2 /5. (See Swiatecki 1956 for the general expression, especially eq. [56].) For
a uniform spherical distribution of charge density p0 and radius R0 about the nucleus Z0, we have the ion-sphere
potential, i.e.,
<Kr) = hrpoQRo2 - r2) + Z0e/r.
The self-energy of this distribution is ^ J p^dr = —0.9 (Z0e)2IR0. On the other hand, the self-energy of the distri-
bution about Z0e in the Debye limit is -0.75(Z0e)2K. Hence the equivalent uniform radius is R0 = 1.2 k'1 (as
long as K is small).
When the two ions are separated by a small distance r « Rx or R2, the distribution will be approximately de-
scribed by equation (26), with a2 # pr/R, where R = (RxR^112 is the mean interionic radius, and ^ = 4/tt — 1 =
0.27324. Thus
g(6) = 1 - pL2r2K2¡1.2 (kR < 1.2) , (27)
2 2 2
AF(r) ä F12(0)g(0) - Fx - F2 = AF(0) - F12(0) . (28)

and

U12(r) « Urir) = + AF(r) (r < rt) . (29)

At large distances, we have a Yukawa solution, as usual:


ZlZ
U12(r) X U2(r) = ^2Cse-Br (r > rs) (30)

such that their logarithmic derivatives fit smoothly at some radius rs. For brevity, we define
B = |AF| s |AF(0)|, A^ZxZae2, a = B¡Xk, and jS = -^2F12(0)/7.2A« .
When k is large, so that kR > 1.2, we must use
g{0) x 1 - ÍM2(r/F)2 (kR > 1.2)

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


1977ApJ. . .212. .513M

No. 2, 1977 THERMONUCLEAR SCREENING 521


instead of (27). Hence k < 1.2/R and r < rs imply
U12(r) = X(ljr-aK + ßKar2) ; (31)
while when k ^ 1.2/R,
U12(r) = A[l/r -aK + lA4ßK(rlR)2].
Note.—From equation (24), kR = £li3 [Ç s ((iC2)1,2> of course].
When kR < 1.2 (i.e., C < 1.23), we find rs and Cs by equating expressions (3) and (31) for U12(r), and the slopes
there as well. We find that x = kvs is the solution of the equation
l — a — ax + 3ßx2 + ßx3 = 0 (32)
and
Cs = e*(l - 2/3x3)(l + x)-1.
Now generally,

-= =0-001037 ttl ±U± D5'3 - Kl - ÎÎ2 - 1


a 7.2 ¿\F (Ci + Í2+ l)6'3 - (Cl + l)5'3 - (Í2 + l)5'3 +1 '
For the case £i = £2 = C, we have ß/a then running from 0.0021 for £ « 1 to 0.0028 for £ » 1. Thus, we always
have ß « a, so we can solve for x (from eq. [32]) by iteration; we find

(33)
a a \ a / <x
When kR > 1.2, ß is replaced by ß' = (\.2lKR)2ß in equations (31)-(33).
V. THE ACTION INTEGRAL
Now that we have a reasonably good expression for U(r), we must find the action integral for the WKB approxi-
mation to the penetrability. For the screened Coulomb potential, the turning radius will be denoted by We
assume a point nucleus (we will show that this simplification introduces only a minor error), and want

S(E) = [U12(r) - E]ll2dr . (34)

Consider the turning point for a given energy E at low degeneracy—i.e., when kR < 1.2 (and & < rs) : U12(ß) =
E. We define the dimensionless energy u = E/Xk; then for r < rs,
Ui(r) — E& A(l/r — a* + ßK3r2 — ku] (35)
vanishes at r = ^. Now we ask for the energy corresponding to &(E) < rs; from equation (35), this condition is
found to yield

E < Em= Xkus ä YZT/x ^f1 - JjO - “)(4 - «)j » (36)

to first order in ß. Having Em, we may consider the calculation of S(E) for E > Em (i.e., & <, rs). From equation
(35) we see that at the turning point,
E-\ = E + B- jSAic3^2 = A/^ . (37)
Heneé we may write the argument in the form

£* [U(r) - E]ll2dr = [7 “ 1 + ß^2 ~ ^2)] ^ •

[It is useful to do this because the ratio of ß\K3(r2 — Si2) to the first two terms is generally small; the largest value
it can take on, at the endpoint, is 1.] Hence for E > EM, we may carry out a binomial expansion of the integrand;
we define z = r/Si and a{E) = (jcSiyß. Then we can write

[U(r) - E]U2 = (A)1'^!^)1'5^ _ CT(l + Z)2]m.

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


1977ApJ. . .212. .513M

522 MITLER Vol. 212


Expanding and integrating, we find

= ]■ «
We may use equation (37) to find 0t in terms of E explicitly, for small 0t\ rather than working in terms of E, it
will be more useful to work in terms oi E' = E + B. Since âi and therefore a are not simple functions of E', we
define the small, energy-dependent dimensionless quantity
<(E) = ßiXKjE'f, (39)
and readily find

3t(E) = A(i + e + 362 + 12e3 + ...) (40)


and

With the help of equations (37), (39), and (40), we then find
7x9x11 2
S{E) = S+(E) = + € + € + E > Em . (41)
kTy/E' \ 16 210
For the case of high degeneracy, i.e., kR > 1.2, everything follows just as before, but with ß' replacing ß everywhere.
Note that for large E, equation (41) gives an integral which approaches the Coulomb value, given by equation (1).
For small E, on the other hand, it goes to a constant, whereas SC(E) diverges. This is the case, of course, because
the finite range of the potential leads to a finite S(0). We may also note in passing that equation (41) has the same
form as equation (29) of SVH.
Next, we turn to the calculation of S(E) for the case E < EM. Then &Z > rs, and the action integral must be
broken up into two parts:

S_(F) = [U.ir) - E]ll2dr - Ç [i/2(r) - E]™dr,

where U^r) is given by equation (29), U2(r) by equation (30). The second integral is complicated, in general.
However, we can find it easily for £ = 0; thus, consider ^.(O). This can be written as ^-(O) = S+(0) - S1 + S2,
where

Sim(2mpjyUi(r)]ll2dr

and

and is the turning point for E = 0. These integrals are readily evaluated, and to the lowest order in €0 s e(0),
we have

S_(0) = 5+(0) + ^¿)1/2 j'?l|H£!j1,2|(27rCs)i/2 erfc

1 - .h _J Es \i/2 (EmEs)112]

where Es = (a - ßx2)\K. Again, if kR > 1.2, we must use ß’ instead of ß. We still need S(E) for 0 < F < Em. The
result is very complicated; on the other hand, even though equation (41) is valid only for E > EM, it has a reason-
able functional form for E < EM as well ; hence it would be convenient if we could simply modify (41) a little in the
region 0 < E < EM, so that it approximates the correct answer. We do this as follows: For E < let (41) be
modified to
S(E) = £+(£) + kSf(E) ,
where
AS=S_(0)-£+(0);

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


No. 2, 1977 THERMONUCLEAR SCREENING 523
S+(0) is the (incorrect) value found at is = 0 by using (41), and f(E) is a function of E such that/(0) = l,f(EM) = 0.
We will make a reasonable estimate for/(E) in the next section.
VI. REACTION RATES AND ACCELERATION (SCREENING) FACTORS
We can now calculate the integral

W= exp (43)

in the rate equation (2). Just as in the Coulomb case, so long as the integrand has a peak, the saddle-point method
can be used, as in § II. The energy at the peak, Em, is readily found by setting
S'(E) = -l/(2fcr).
This yields

1 + 2kTASf(E) = (I?)3'/ +IT + e2 + • • • I •

We define
0 = -2kTkSf'(Em)
and write the above equation as

f + + (44)

where now € = €(Em) and Em' = Em + B. Thus, if 0 = € = 0, we would have the Gamow peak shifted down by B.
With 9 and € / 0, the shift is somewhat smaller. Note that for Em > Em, 0 0.
The simplest approximation forf(E) which permits us to match the slopes as well as the values of S(E) and S+(E)
at Em is
f(E) = (1 - EIEMf . (45)
Then

0a X 4kT AS
._ (46)
Em
and with this substitution equation (44) must be solved for Em. Having obtained Em, we also have 0; we then use
them in equation (3) (so long as Em > 0), and find, for the integrand in equation (43) at E — Em,

i>(€, 0) ^ exp{-[^ + 2S(Em)

( rp ' _ R 2Fn3m
exp + 2^SREm) 1 +
+ kTVE
n A I6e +
)]}

- [-{&<> - «)-»[■ 44«+"]+l^+f<1-- ¡1«—] -


(47)
When the degeneracy is low enough that Em ^ Em, then equation (41) is valid, 0 = 0, and equation (47) im-
mediately yields

2S
# + <«-&(3 + 5* + TÍ¥«2 +
'")+rí’ E,ZE„. (48)

Finally, to first order we find


3 35 \
S"(Em) 1 +
4E0kT ( 12 7

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


1977ApJ. . .212. .513M

524 MITLER Vol. 212


and thus from the saddle-point method [see eq. (3)] we find, for all Em > 0,

(4,)
^ -1 *){'+»f [& •
Of course, the standard result (i.e., without screening) is recovered if we set | AF| and e = 0.
The pycnonuclear regime.—If the shift |AF(0)| is so large that the peak occurs for Em < 0, then it is more appro-
priate to expand the exponential about Is = 0, rather than about Em (which is, of course, what the saddle-point
method does). Thus

(50)
y/E' s/E + B y/Bl 2B^ %\B) \
Now, Em < 0 corresponds to extreme degeneracy; but then «->0, and hence as we see from equation (36),
Em-+0 ; therefore, equation (41) is correct for all energies. Then equation (43) becomes

r i \E +^ 2Eom /i u. 5 7x9x11
W: exp + € + 210
kTVE' Jo i [kT kT^E' 16 ■)]h-
Next, we can use the mean-value theorem, and take e(E) e(E) = e. Then with the expansion (50),
312
2Ê0
W: exp + (51)
Jo I [kT kT-s/B (1 2B +
2, B2)\)dE '
where

(Ê0f<2 = £o3/2(l +^+ 7X 9 Xl1 e 2


2 xo ‘) •

(We can estimate the mean value Eto be the expectation value of E when weighted_by the integrand in [51], with
E0-+E0. However, the precise value is evidently not important; when B » E0, E ä kT.) The integral (51) is
easily carried out, and we find

W x B^y *(erîc G) exp (Ga - 2H), (52)

where
(53)
»-Mir
and

(54)
»■u^r-ferwAir
The expansion (50) is valid only for E < B. Thus it is clear that W as given by (51) and (52) is an underestimate of
the correct value. In this region—i.e., Em < 0—we see that the temperature dependence of Wdiminishes, while the
density dependence increases. This is the pycnonuclear regime (Zel’dovich 1958; Cameron 1959; Harrison 1964).
(SVH use the term “pycnonuclear” for reactions in the crystalline regime—see below.) In the low-temperature
limit we find that

exp 31.291z-^) (55)

with B (in the exponent) in keV. At low temperatures (and for Zx = Z2 = Z),
il/3
B^ 1.8(22/3 - l)e2Zs'3^y2y

and therefore the exponent is proportional to p~li6', hence as the density rises, so does the reaction rate. This form
of the reaction-rate integral was first given by Harrison (1964), though with somewhat different constants, resulting
from a somewhat less accurate assumed potential. This region where the temperature dependence becomes negli-
gible, Harrison refers to as the “ cryonuclear ” region. Note that the result (55) appears paradoxical : upon multiply-
ing by S(E)Q12 (to get r12), we find that r12 oc T~112, so that the reaction rate appears to go through a minimum

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


1977ApJ. . .212. .513M

No. 2, 1977 THERMONUCLEAR SCREENING 525


and then to diverge, as T —> 0. However, we note that when the temperature is sufficiently low, the plasma
“crystallizes” (see, e.g., SVH), and these calculations cease to be valid. Crystallization takes place when the
Coulomb energy between ions in the plasma becomes significantly greater than their thermal energy. Following
SVH, we. define the ratio

P _ -ffoopi. _ (Z*e)2 r, /1/3 0.22748


“ OMT kT \6 )
They find that crystallization occurs for F ~ rcrlt. The precise value of rorlt is not known; we adopt the value used
by Graboske (1973): rcrlt = 143. Whatever value it in fact has, the phase transition from fluid to solid is expected
to occur over a narrow strip of the (p, replane.
We can now calculate the acceleration factors : for the case Em > 0, equation (49) gives us both the screened and
unscreened reaction rates, and it follows that
1/2
1 + erf f£m /3kr\ i'l
f _ ^(S.C.) _ Mn) [kT \4E0 ) ] H*, 6)
h (56)
~ r12(C) ~ $(¿0) P(0,0) ’

where from equations (47),

and Em'IE0 is given by equation (44). When Em > EM, 0 = 0 and the expression simplifies considerably: equation
(48) then yields

F(e, 0) Eo_ 7x547


exp (-
P(0,0) \kT kT 3 x 29
Except right near Em = 0, where the error function vanishes, all the factors in front of the exponential are close to
unity. Hence we may write

In/s ä ¿ [l AF| - I £oe(l + j, Em > Em > 0 . (58)

Thus we recover the standard weak-screening expression for fs, but with corrections. Thus, although it is true that
fs = exp (AF/kT) in the weak-screening (low-density, high-temperature) limit, it does not follow that that is the case
in a// regimes, as assumed by Salpeter (1954), Cameron (1959), DeWitt, Graboske, and Cooper (1973), etc.
Equation (58) is valid, moreover, not only for weak screening, but for strong screening as well. When the screening
is very strong, so that 0 < Em < EM, we must use the more general expression (57), where 6, Em, and EM are given
by equations (44), (46), and (36). Finally, when Em < 0,/s no longer has this form at all. Equations (49) and (52)
now lead to

„ ¡B\5li (erfc G) exp [G2 - 2H + 3E0lkT]


Js (59)
~ S(E0) \Eo) 1 + erf (3E0l4kT)112 ’ "
We again note that generally §(E) ~ S(E0) and (3E0l4kT)112 » 1. We also find that usually G » 1. We may then
use the asymptotic expression for the error function to write

The model we have presented, resulting in equations (56)-(59) for /s, constitutes model A. This completes the
formal presentation. Calculations based on model A are presented in the next section.

VII. MODEL A RESULTS AND COMPARISON WITH OTHER WORK


For our calculations, we have continued to assume the C + C reaction in a pure 12C medium; the results for
C + O mixtures will not be very different. The reaction rates and screening factors have been calculated numerically

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


1977ApJ. . .212. .513M

526 MITLER Vol. 212

Fig. 2 Fig. 3
Fig. 2.—Reaction rates for the C + C reactions (in terms of the dimensionless quantity K\ plotted as a function of temperature,
for p = 1.36 x 106 gem“3. The curve labeled GDGC is obtained by multiplying the vacuum rate by /s, as calculated from the
Graboske et al. theory. The one marked model A is obtained by multiplying by fs as calculated in the present work, and the one
marked SVH is obtained directly from the finite-temperature reaction rate calculated by Salpeter and Van Horn.
Fig. 3.—Same as Fig. 2, for the case p = 2.51 x 108 g cm“3. The curve labeled model B, in addition, is obtained from the GDGC
values by applying the correction given in the text to the expression fs = exp Li/i2(0)].
according to equations (56), (57), and (59). For purposes of comparison the reaction rates are shown in terms of
the dimensionless quantity
ll2r
= (3\ i2 mc — (3tnc2\112 W(kT)~m
~ W § rtitiz ~ \ ?r / (1 + 8ia)
lmc2\w(E0\w 2 / 2>E0\
\kT) (fcr) i + s12^exp\ kr)’ ^
The program also computes this reduced reaction rate for the Graboske et al (GDGC) theory and for the SVH
theory, using equation (60) and the values of /s found from their formulations. (For the SVH calculations their
“relaxed” approximation has been used, with temperature correction but without the partition function correc-
tion.) The results are shown in Figures 2, 3, and 4 (the densities have been chosen to agree with the values used in
Table 1).
We immediately note that—as one would expect—the GDGC results and those from the present work (model A)
merge at high temperatures. Perhaps the most striking observation is that there is no continuity or overlap between
the GDGC and SVH results. That is quite independent of the new results presented here. Graboske (1973) argues
that the steep inverse temperature dependence found by him at low temperatures results from the very rapid rise
in fs with decreasing temperature. That is certainly its mathematical origin, but then the question arises as to the
adequacy with which fs has been calculated. That a reaction rate should fall with increasing temperatures violates
our physical intuition, but it can justifiably be argued that the latter is a poor guide in such unusual conditions.
However, the SVH calculations in the crystalline regime are relatively reliable, since they did not make the a priori
assumption that AÍ7 (in fs = exp AU/kT) is the difference in free energies. But if the SVH calculations for the
crystalline and near-crystalline regions are accepted, then the inverse-T dependence given by the GDGC theory
must be spurious.

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


1977ApJ. . .212. .513M

No. 2, 1977 THERMONUCLEAR SCREENING 527

Fig. 4.—Same as Figs. 2 and 3, for p = 2.0 x 1010 gem-3. The discontinuity in the model A curves arises from switching to the
pycnonuclear calculation at 2sm = 0, rather than at B = E0. The latter procedure gives continuous curves, the transition occurring
at a higher T, close to the minima of the GDGC curves. The discontinuity has been displayed purposely, to show the inaccuracy
inherent in the present calculations (in that region).
Next, we note that the model A curves generally lie much closer to the SVH curves than do the GDGC ones in
the region where those calculations overlap. The discontinuous (dashed) parts of the model A curves arise (at
Em = 0) because of the difference in the values of W calculated according to equations (49) and (52). The calcula-
tions, in fact, match near the point where B = E0; this occurs when Em is still positive, i.e., for somewhat higher
temperatures. The discontinuity has been left in because its size is a measure of the relative error involved in these
TABLE 1
Acceleration Parameter /s, as Computed from Various Expressions,
for a Carbon Plasma at Several Densities and Temperatures
Example Number
Parameter 1
3
P (g cm“ ) 1.36E6 2.51E8 2.0E9 2.0E10 2.0E10
Te 502 251 300 300 100
S 2.30 1.34E4 7.9E6 7.3E14 3.8E44
SVH:
a) 2.30 1.2E4 4.7E6 1.22E13 8.3E34
¿).... 2.30 5.1E3 1.1E6 1.06E12 1.2E34
H 2.20 7.8E3 3.2E6 1.1E14
C 3.26 6.8E5 5.6E9 1.0E21
GDGC 2.06 4.2E4 6.1E7 7.3E16 5.6E50
Correction factor. 1.000 1.034 1.19 5.62 5.21E8
Present work:
Model A 2.12 1.25E4 6.4E6 1.3E14 7.3E35
ModelB 2.06 4.1E4 5.1E7 1.3E16 1.08E42
F 0.8 9.0 15 32 97
Note.—The headings correspond to: S = Salpeter 1954; SVH = Salpeter and Van Horn 1969; H = Harrison 1964; C =
Cameron 1959; GDGC = Graboske et al. 1973. Model B is described in § VIII.
a) The original SVH factor.
b) The same factor with their correction due to the change in the partition function.

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


1977ApJ. . .212. .513M

528 MITLER Vol. 212

Fig. 5 Fig. 6
Fig. 5.—The (dimensionless) reaction rate as a function of r, for three values of density, in a pure 12C medium. These have been
obtained by interpolating smoothly between the model A and the SVH values, where necessary (from Figs. 2, 3, and 4). The medium
is crystalline to the left of the dashed curve. We note that for p = 2.0 x 1010 gem-3, we are already well into the pycnonuclear
region—i.e., in the cryonuclear region—before crystallization occurs. Thus “pycnonuclear” is not equivalent to “crystalline.” Below
the knee, the temperature dependence is negligible.
Fig. 6.—The acceleration parameter fs as a function of temperature, for three values of density (for the C + C reaction), corre-
sponding to Fig. 5. We note that in the thermonuclear range, these are given pretty closely by power laws in T.

calculations. Even if this discontinuity were removed, it is clear that model A also yields a minimum in the reaction
rate as a function of r(even though the fall in K with T is not as steep as with the GDGC model). This minimum
occurs at the transition to the pycnonuclear regime, and is in accord with the discussion just below equation (55).
We note that in the region of the minima, the expected number of particles in the Debye shell is too small to justify
these calculations : <V> < 1. It follows that collective effects cease to be valid, and | AF(0)| as given by equation (25)
is therefore too large. Hence there must be an overestimate of the reaction rate r12(SC). That is, the ion-sphere
model breaks down. This explanation, however, may not be correct, since the GDGC work explicitly takes fluctua-
tions into account. As one might expect, the SVH finite-temperature correction becomes significant just as we leave
the Fm < 0 region.
It is, finally, seen that a smooth interpolation can always be made between the model A and the SVH results.
When this is done, the curves of Figure 5 result. The corresponding values of fs are plotted in Figure 6. It is also
worth noting that the transition from almost no temperature dependence to a very steep temperature dependence
is usually quite abrupt. We might therefore adopt Harrison’s terminology and refer to cryonuclear and thermo-
nuclear regimes.
The points at which crystallization takes place are also shown in Figure 5; it is seen that for sufficiently high />,
the pycnonuclear regime is attained before crystallization takes place. With increasing density, crystallization occurs
in regions deeper in the cryonuclear range. We see that fs is a steeply decreasing function of temperature, but a
steeply increasing function of density.
Next, we compare our results with those of other authors for some selected cases: in Table 1, we list fs as com-
puted from the expressions given by various workers as well as from those derived in the present work, for five
examples. The results listed in the next to last row are from model B, which will be described in the next section.
Example 1 corresponds to a carbon flash in a 0.8 M© star—i.e., degenerate ignition (see Beaudet and Salpeter
1969). The screening is intermediate, and the results of the various calculations agree fairly well. Example 2
corresponds to a late stage of carbon burning, where the energy generation rate is equal to the neutrino energy loss
rate (from Rose 1969). For this and the rest of the examples, screening is very strong. The various estimates now

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


1977ApJ. . .212. .513M

No. 2, 1977 THERMONUCLEAR SCREENING 529


diverge significantly, with a factor of ~ 100 between the largest and the smallest. The value for model A is inter-
mediate, and agrees most closely with that calculated from S. Example 3 corresponds to a thermal runaway in a
star of 4-9 Af©, with a 0.8 M© carbon core—i.e., explosive carbon ignition in a supernova shock (Arnett 1969).
Arnett used the Salpeter strong-screening expression for /s. Here the Salpeter, Van Horn, and Harrison results
agree within a factor of 4, but are (again) smaller than those of Cameron and GDGC ; our model A result is closest
to that of Harrison. Example 4 was computed to see what the effect is of just increasing the density by an order of
magnitude from what it is in example 3. These physical conditions might be found in an extreme white dwarf, or
during collapse to a neutron star. Here the various calculations begin to diverge radically, with nine orders of
magnitude difference between the largest and smallest estimates. The model A result is closest to that of S, as might
be expected, since it approaches the ion-sphere model. Thus one order of magnitude increase in p results in the
acceleration factor going up by eight orders of magnitude. The ratio F is 32, and hence the medium is not yet
crystalline. However, we are just at the edge of the crystalline range for T6 = 100: F = 97. Hence example 5 was
calculated. As we see from Figure 5, at /> = 2 x 1010 and 108 K we are also in the cryonuclear range.
We conclude that model A is indeed an improvement over the simpler models of S, H, or C, and in some ways
even over the GDGC work. We will discuss the latter in the next section (the penultimate row of Table 1 refers to
a modification of the GDGC work). Finally, we see that the correction factors (see, for example, eq. [58]) become
significant only for very high degeneracy.
VIII. INADEQUACIES OF PRESENT TREATMENT; MODEL B; NEEDED WORK
We must now consider the validity of the various approximations we have made.
First, the WKB approximation to the penetrability,
P(E) * exp [-23(E)],
is not terribly good. However, the ratio of the penetrabilities is in fact quite good—the errors tend to cancel each
other. Thus, for example, for a hypotheticalp-p reaction at = 25, /> = 2 x 106 gem“3, the transmission co-
efficients (screened and unscreened) found via WKB are 5.8 times too large (as found from an “exact” calculation
of the transmission coefficient made by finding the wave functions); the ratios, however, differ by only 0.7%.
Second, the errors incurred by assuming a point nucleus also approximately cancel, again as found from explicit
calculations.
Third, the use of the saddle-point method. Since the shape of the effective interaction is so similar to that of the
Coulomb interaction, being mainly shifted by Es, the corrections2 to r12(SC) and to r12(C) are essentially the same
and hence, once again, the errors offset each other in fs. Exact (numerical) calculations of the wave functions and
resulting penetration factors show that the error arising from all three of these approximations is only a few percent
for the /a = 2 x 106 example.3
Fourth, we ask for the effect of using /o0(r), rather than the exact density distribution, p(r). If we use p(r) =
p0(r) + Ap(r), such that J £sp(r)d*r = 0, then the potential is changed from <f>(r) to <h(r) = <t>(r) + A<£(r) and, as
we see from equation (34), S(E) is changed to

S' + AS = J [U(r) -E + Z2eA<Kr)]ll2dr .

From the mean value theorem, the entire effect is equivalent to the shifts
ES->ES + AES ~ Es — Z2eA<l>(r)
and
fs ->fs' - fs exp [-Z2€Acf>(r)].
For the example used in § III, a simple calculation yields AES = 0A6%kT, corresponding to an 18% increase infs,
from 2.06 x 109 to 2.44 x 109. In view of the other uncertainties, this is negligible.
Fifth, consider the effect of ion-ion correlations, which we have neglected so far. Short-range order does indeed
come into play at high densities. DGC discuss these, and display g(r) [which is essentially Pi0n(r)] f°r very dense
plasmas in their Figure la. The principal effect is to enlarge the region about Zx from which other ions are excluded,
and “pile up” the excess ions about the edge of this ion sphere. For the almost-crystallized case F = 100 in a
carbon plasma, a crude calculation shows that// # 26/s, where fs is the value found from model A, for example—
a considerable increase. For the example we have been considering, on the other hand, F = 15, and their Figure la
permits one to estimate
ft ä 1.29/s * 3.14 x 109,
2
Usually denoted by F(r).
3
This work was done by Raymond Davis, Jr.

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


530 MITLER Vol. 212
a much more modest increase. The effect is no larger because the normalization condition on Ap(r) keeps A<£ small.
In view of the disagreement between GDGC and SVH, however, it is not clear that the ion-ion correlations in fact
increase fs even that much.
Sixth, consider the effect of the fluctuations when the particle number in the Debye shell is small. If there are
many fluctuations during the time of passage of the ion from the edge of the ion sphere, to the surface of the
nucleus, Rn, then the time-averaged density distribution will still be the equilibrium one. We expect the fluctuation
frequency to be the ion plasma frequency. Hence an additional criterion for the validity of the fluid model is

where vm is the velocity corresponding to the energy at the peak, Em. For most cases of astrophysical interest,
Rn « Ri. Hence we may write (61) as

£m « ^ Z2«2 = 2.94 x IO"5(J)1/3 Z2 MeV . (62)

(When Em < 0, E should be used in the left-hand side of eq. [62].) For our canonical example, this criterion is not
fulfilled, however: Em approximately equals the right-hand side. Hence in this case the quasi-static deviations of the
ionic distribution from the smooth distribution used in this work must be taken into account. That is precisely what
DGC and GDGC did, and we next consider that work.
Model B.—li is clear from equations (28) and (29) that ^i2(0), as calculated by them, takes the place of BjkT.
That suggests a possible improvement: we replace | AF(0)| by H12{0)kT, whenever the former appears. The calcula-
tions then follow precisely as before, with the correction terms in € still there, as in equation (58). This is model B—
that is, just model A with the substitution B = |AF(0)| H12{^)kT. Thus for Em > 0, we have precisely the
GDGC result, but modified by the correction factor. When Em < 0, equation (59) is the appropriate expression
to use, with (again) H12{G) taking the place of BjkT. The results of using this procedure in examples 1-5 are given
in the penultimate row of Table 1, and the case p = 2.51 x 108 is plotted in Figure 3. We see that model B is not,
after all, an improvement over model A!
Seventh, quantum-mechanical effects for the nuclei are negligible at the densities we consider; crystallization of
the medium occurs long before we get to near-nuclear densities, where degeneracy of nucleons and other such
effects come into play.
Eighth, and finally, the fact that the interacting nuclei are moving rapidly in the plasma has not been taken into
account: the potential found here, as in all such calculations, is taken to be an equilibrium (static) potential. In
fact, since the ions near the peak usually move supersonically (Em » kT), the ion clouds about the reacting nuclei
will not have an opportunity to achieve equilibrium, thus lowering the degree of screening. This effect is expected
to be quite significant at high densities. Calculations have been made to take this effect into account, and will be
published as Paper II.
IX. RESONANT REACTIONS
When the cross section has a sharp resonance which contributes substantially to the reaction rate, the screening
factor for it is different from the above expressions. Let us refer to it as /r.
The cross section (in vacuum) is
rxipr/io
<r„(£) = Fa,-
- Erf + il\2
where w is the statistical weight, and ri? r/? and F are the widths of the incoming channel, the outgoing channel, and
the total width, respectively; Er is the resonance energy. In the plasma, however, we have seen that there is work
AFdone on the particle on approaching the target nucleus. This is equivalent to raising the energy reference level,
and hence in the plasma the resonances are shifted to Er' = Er — AF.
Now if the peak is very sharp, then the integration is trivial, and we find the reaction rate to be

r12(SC, res) = 2*Q12(^f)G(T, Er) exp (63)


(-§)
Nonresonant reaction rates have to be multiplied by a correction factor F(r) to account for the deviations of the
Gamow peak from the assumed Gaussian shape. In the same way, G(Er, T) here is a correction factor which
accounts for the deviation of the resonance peak from being a pure delta function. We find that so long as the width
is small,

G(Fr, T) = 1 - JL +
+
2 (IlV (64)
2ttEt 32 \kT)

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


No. 2, 1977 THERMONUCLEAR SCREENING 531
The energy dependence of Tf is weak for exothermic reactions, and we ignore it. But the incoming width has a
barrier-penetration factor: 1^(2?) oc e~23tn. Hence if « r/5 this dependence appears in equation (63), but (clearly)
must be replaced by the screened penetration factor. It follows from (63) and the form it would take in vacuum
that

+ 2 2S
f- - TÜSSf “ “■> [w ”" - «'>] <r' ^•
where r¡r is evaluated at Er. On the other hand, when Fj æ Fr (so that Fj » F^, TilTr ä 1 in equation (63); and
fr * exp (AF/kT) (F( » F,) .
In general, we have
TijEr') Tr(Er)
fr = rtfr) r (E ’) exp
r r

where the widths at primed energies are calculated with the screened penetrability. These expressions are equivalent
to those found by SVH, though the numerical values will differ because of the different S(E).
X. SUMMARY AND DISCUSSION
We have found a good (fluid) approximation to the equilibrium charge density surrounding a target nucleus in a
plasma ; from that, the expression (25) for the free energy change (which is a generalization of Salpeter’s expression) ;
and therefrom, finally, the (equilibrium) interaction energy between the target nucleus and an incoming nucleus.
Then using the WKB and saddle-point methods, we have found the penetrability of this barrier, and thus the
increase of the reaction rate over what it would be through a Coulomb barrier in vacuum, i.e., the factor/s. We
find that the peak of thermonuclear reactions is shifted down from the Gamow peak by roughly | AF(0)|, the work
done in bringing the nuclei together (that is, the difference in free energies between the initial and final configura-
tions). This is the standard result, first obtained by Salpeter (1954). At high densities, however, the correction factor
given by equation (58) becomes significant. When the shift is so large that there is no longer a peak in the inte-
grand, we are in the pycnonuclear region; then the saddle-point method breaks down, and the expression (58) is
altogether invalid. Instead, fs is then given by equation (59).
In both these expressions, a better result may in principle be obtained if we use the expression H12($) calculated
by GDGC, in place of EJkT. (It must nevertheless be stressed that it is not proper to merely set/s = exp [i712(0)].)
When exp [i712(0)] is modified by the correction terms found above, we have model B. The correction factors operate
in a direction such as to lower the value of fs below exp [/^(O)]. Whether the result is better in fact, however, is
questionable in view of the discrepancy between the GDGC and SYH results near crystallization.
The examples in this paper were chosen for a pure carbon plasma, because the deviations from the usual expres-
sions only become important at very strong screening, which generally occurs in late stages of stellar evolution—
e.g., in carbon stars. The calculations are equally applicable to mixtures, however.
The difference between the present results and those of Graboske et al. are not significant until very strong screen-
ing is attained—i.e.,/s > 50. Assuming that the SVH results are valid, it is clear that model A gives much better
results than any of the other calculations. In the region near the pycnonuclear transition, it is always possible to
make a smooth extrapolation from the SVH curve (of reaction rate, for example) to that given by model A, or a
smooth interpolation between the two. Hence the present pycnonuclear calculation, equation (59), is to be neglected
entirely. A conservative estimate of the probable accuracy of these calculations (including the interpolation pro-
cedure) is given by the probable uncertainty in the SVH calculation. Thus a reasonable estimate is

In/s
The applications of these results have been discussed in some detail by GDGC; they rightly point out that because
of the great temperature dependence of nuclear reactions, uncertainties (or changes) in the astrophysical constant
or the acceleration parameter are easily compensated for by a relatively small change in the temperature. (In the
cryonuclear region, the same result is achieved by varying the density.) As they point out, the places where the
new screening values could show up are: (1) The minimum mass of stars burning a given element; (2) The point
of onset of burning high-Z fuel—e.g., C. This affects the structure of highly evolved stars, and also affects the
evolution of thermal runaways and the point where they occur—i.e., degenerate ignition. This is true either for a
flash (where we relax to a higher-temperature configuration) or for a detonation, where material is blasted away.
Since we find smaller acceleration factors, a blast wave will either be weaker, or start at higher (p, T) values, or
both. In particular, in Figure 7 we show the carbon-ignition curve calculated by Graboske (1973). The function
is double-valued in temperature because of the apparent rise in the reaction rate at low temperatures yielded by their
calculations. Assuming carbon ignition at the same reaction rates as he does, the present results yield the curve

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


1977ApJ. . .212. .513M

532 MITLER

Fig. 7.—The carbon-ignition curves in a pure carbon medium, according to Graboske (1973) and according to the model
developed here, as modified by the SVH results. As usual, carbon ignition is assumed to occur at the point where the energy
generation rate due to the 12C + 12C reaction, €1212, equals the rate of loss of energy by neutrino emission, €v. The straight line is
the evolutionary track of a carbon core, as calculated by Paczyñski (1971), and extrapolated by Couch and Arnett (1975). The
intersections correspond to expected runaway ignition of the core.

labeled “Model A 4* SVH.” This curve is no longer double-valued. However, at p = 4.8 x 109gcm"3, the
cryonuclear rate is sufficiently high to “ignite” the carbon even at zero temperatures. In the same figure the evolu-
tionary track of a dense core is shown (from the evolutionary calculations of Paczyñski 1971, as extrapolated by
Couch and Arnett 1975). It is clear from that curve that for the Zc = 1.0 case, C ignition would move from p =
1.6, Tq = 2.5, to p = 1.9, Te = 2.6 as a result of the decrease in the acceleration factors calculated here—a very
modest change. Thus the problems of pulsar remnants and of Fe overproduction remain.
The author wishes to thank Mr. Raymond Brady, Jr., for many useful discussions and for many numerical
calculations.
REFERENCES
Arnett, W. D. 1969, Ap. Space Sei., 5, 180. Kidder, R. E., and DeWitt, H. E. 1961,/. Nucl. Energy,Part C:
Beaudet, G., and Salpeter, E. E. 1969, Ap. /., 155, 203. Plasma Phys., 2, 218.
Cameron, A. G. W. 1959, Ap. J., 130, 916. Paczyñski, B. 1971, Acta Astr., 20, 47.
Clayton, D. D. 1968, Principles of Stellar Evolution and Rose, W. K. 1969, Ap. J., 155, 491.
Nucleosynthesis (New York: McGraw-Hill). Salpeter, E. E. 1954, Australian J. Phys., 7, 373 (S).
Couch, R. G., and Arnett, W. D. 1975, Ap. J., 186, 791. Salpeter, E. E., and Van Horn, H. M. 1969, Ap. J., 155, 183
DeWitt, H. E., Graboske, H. C., and Cooper, M. S. 1973, (SVH).
Ap. J., 181, 439 (DGC). Schatzman, E. 1948, J. Phys. Rad., 9, 46.
Graboske, H. C. 1973, in Explosive Nucleosynthesis, ed. D. N. Stewart, J. C., and Pyatt, K. D. 1966, Ap. J., 144, 1203.
Schramm and W. D. Arnett (Austin: University of Texas Swiatecki, W. J. 1956, Phys. Rev., 104, 993.
Press, 1973). Wolf, R. A. 1965, Phys. Rev., 137, B1634.
Graboske, H. C, DeWitt, H. E., Grossman, A. S., and Zel’dovich, B. 1958, J. Exp. Theor. Phys. (English transi.
Cooper, M. S. 1973, Ap. J., 181, 457 (GDGC). Soviet Phys.—JETP), 6, 760.
Harrison, E. R. 1964, Proc. Phys. Soc. London, 84, 213.

Henri E. Mitler: Harvard University, Division of Engineering and Applied Physics, 40 Oxford St., Cambridge,
MA 02138

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

You might also like