Sahas Equation and the Partition Function

Robert Gilmore

Abstract The Saha equation is usually derived using an argument involving one of the classical thermodynamic functions, the Gibbs free energy. Sometimes dynamical arguments are used. Since the Saha result is a statistical relation, it would be desirable to have less roundabout methods for its derivation. Here the Saha equation is derived directly from the partition function.

The dissociation interaction AB A + B is encountered under many conditions: Process Molecular dissociation: Atomic ionization: Equation H2 2H H p+ + e E (eV) 4.476 13.597
E

(1)

The dissociation process is described by the Saha equation. To motivate the result that follows, we rst begin with a very simple example. Suppose we start with three molecules of type AB. Each dissociates into two dierent atoms A and B. The partition function, counting all possible states weighted by their Boltzmann factors, is ZT ot =
3 Z2 Z1 Z1 Z1 Z2 Z2 Z3 Z3 ZAB + AB A B + AB A B + A B 3! 2! 1! 1! 1! 2! 2! 3! 3!

(2)

In this expression ZA = V (2MA kT /h2 )3/2 Za (int). The rst factor comes from an integral over the phase space of the center of mass coordinates of the particle. The second factor, Za (int), is the partition function counting only the internal states of A. Similarly for B and the combined particle AB. Each term in Eq. 3 ZAB (2) accounts for one of the possibilities suggested by Eq. (1). The part 3! is the relative probability that exactly 3 AB molecules are present. The molecules 2 1 1 ZAB ZA ZB 1 are indistinguishable, hence the factor of 3! . The second term, 2! 1! 1! , is the relative probability that two molecules AB are present with one atom each of A and B. And so it goes. The absolute probabilities are obtained by 1

normalizing each of these (four) terms with their sum, ZT ot . It will not be necessary to do this normalization. When lots of atoms AB and their dissociation products are present, the total partition function is a simple generalization of that given in Eq.(2): ZT ot =
NAB N N ZAB ZA A ZB B NAB ! NA ! NB !

(3)

The sum extends over all allowed values of the integers NAB , NA and NB . The integers are not free to vary arbitrarily: if NAB decreases by one, NA and NB must increase by one. Instead of trying to carry out this sum, it is simpler just to look for the most likely state: the state of largest probability. Roughly but accurately, as the integer NAB decreases by integer steps and the two integers NA and NB each increase by integer steps, the relative probabilities will rst rise, then reach a maximum, then nally decrease. At the largest liklihood, two successive relative probabilities will be approximately equal. We can represent this condition as
N N N N NAB NAB ZA A +1 ZB B +1 ZAB 1 ZAB ZA A ZB B = NAB ! NA ! NB ! (NAB 1)! (NA + 1)! (NB + 1)!

(4)

This is easily rewritten as ZB ZA ZAB = NAB NA + 1 NB + 1 (5)

For large numbers of particles there is little dierence between NA + 1 and NA , so we can neglect the +1s in the denominator on the right hand side. Now substituting the expressions for the three individual partition functions ZAB , ZA and ZB into this expression, we nd V (2MAB kT /h2 )3/2 Zab (int) V (2MA kT /h2 )3/2 Za (int) V (2MB kT /h2 )3/2 Zb (int) = NAB NA NB (6) In each of these three factors there is a ratio of V /N = 1/(N /V ) = 1/[n ]. Here [n ] is the density of particles of type . We measure this in number of particles per cm3 . By unwinding these factors carefully, we nd [nA ][nB ] = [nAB ] 2(MA MB /MAB )kT h2
3/2

Z(int)a Z(int)b Z(int)ab

(7)

We now turn this into a computationally useful expression. We assume that AB is bound by an energy E which is negative (13.6 for the hydrogen atom and 4.476 for the hydrogen molecule). We assume the molecule AB has no internal structure and only the single bound state at E, so Z(int)ab = eE . We further assume that neither A nor B has any internal structure, so that Z(int)a = 1, Z(int)b = 1. Next, we introduce a reaction coordinate x that describes quantitatively how far the reaction proceeds in some direction (by 2

convention, to the right). In terms of this coordinate, if the initial concentration of molecules AB at low temperature is n0 , at nite temperature T , [nAB ] = (1 x)n0 while nA = nB = xn0 . Taking logs (to base e) of both sides of Eq.(7) we nd x2 1x 3 3 log(E)+ log 2 2 2(MA MB /MAB )(E) n0 h2
2/3

log

= E

(8)

If the masses are measured in units of the proton mass, energies are measured in electron volts, and densities are measured in Avogardos per cm3 , this expression becomes x2 1x 3 3 3 = E log(E)+5.9769+ log(M/Mp )+ log(|E|/(1eV))log(n/n0 ) 2 2 2 (9)

log

Application 1: Recombination
As the universe expanded and cooled after the Big Bang, the temperature and the particle density decreased to a point where it was entropically feasible for charged free protons and oppositely charged free electrons to combine to form electrically neutral hydrogen atoms: p+ + e H (10)

This occurred about 380,000Y after the Big Bang. At this time the temperature was about 3000 K and the baryon density was about 1.2 102 protons/cm3 , so that n/n0 = 0.2 1021 . We use Mp Me /MH = Me = 1/1836, measured in units of proton mass. The reaction coordinate x, as a function of temperature, is derived from Eq.(9) and plotted in Fig. (1). Technical Details: These details can be skipped on rst reading. Both the electron and the proton have spin 1 , so each momentum state is two-fold 2 degenerate. As a result, both the electron and the proton partition functions should be multiplied by a factor of 2. Countering that, the hydrogen atom 1s ground state is four-fold degenerate (both the proton and electron can have spin up or spin down), so its partition function should be multiplied by a factor of 4. The ratio 22 /4 leaves Eq. (9) unchanged. A more serious problem concerns the internal structure of the hydrogen atom. It would seem that the appropriate internal partition function, consisting of the sum of the Boltzmann factors over all possible bound states, is

ZH (int) =
Bound States

eEn
n=1

4 n2 eEg /n

where Eg = 13.58 eV. This sum diverges. We have kept only the rst term in this sum in Eq. (9), which is very small except for very high temperatures. 3

Dissociation Curve at Recombination Densities

Degree of Dissociation, H leftrightarrow p^+ + e^H leftrightarrow p^+ + e^- at 1.2 x 10^(-21) moles

0.8

0.6

0.4

0.2

0 3000

3200

3400 3600 Temperature (K)

3800

4000

Figure 1: Reaction coordinate x as a function of temperature at about 380,000 ABB when the baryon density is about 1.2 102 protons per cm3 .

What justies this gross approximation? We claim that the Boltzmann factors for the excited hydrogenic states should not be considered on the grounds that these states are not in thermal equilibrium with the surroundings. To justify this claim we point out that population inversion of the higher states, leading to observed maser radiation, is not possible in thermal equilibrium.

Application 2: Molecular Dissociation

At the surface of some stars the molecular to atomic dissociation H2 2H
E

(11)

takes place. Molecular hydrogen is bound by 4.476 eV at 0 K. We introduce a reaction coordinate x as above. If the concentration of H2 at very low temperatures is n0 , then at higher temperatures [nH2 ] = (1 x)n0 and [nH ] = 2xn0 . 2 The left-hand side of Eq.(9) is log( (2x) ) while the right hand side diers in 1x the values of the parameters E and n0 . If we choose n0 = 1 Avogardo, the 2 dissociation curve is as shown in Fig. 2. The molecular dissociation shown in Fig. (2) occurs at a slightly higher temperature (by a factor of 4) than the recombination, or atomic ionization that is shown in Fig. (1). This is due to the vastly dierent densities that have been assumed for the two cases. In the rst case the density is about 102 protons / cm3 while for the second case it is larger by 21 orders of magnitude. 4

Dissociation of Molecular to Atomic Hydrogen

Density of Protons: Avogadros Number / cm^3

Reaction Coordinate: H_2 leftrightarrow 2H

0.8

0.6

0.4

0.2

0 5000

10000

15000

20000

Temperature (K)

Figure 2: Reaction coordinate x as a function of temperature for dissociation of molecular into atomic hydrogen at the surface of a star where the density is 6.024 1023 protons per cm3 .

Application 3: Dissociation - Ionization

A gas of molecular hydrogen at fairly low temperature will dissociate into atomic hydrogen as the temperature increases, and the atomic hydrogen will dissociate (ionize) as the temperature continues to increase. The two processes can be described by the Saha equation. We formulate this double dissociation process in terms of Boltzmann factors and partition functions. The partition function describing all possible states is Z=
Ne p H N ZH2 2 ZH H Zp+ Ze NH2 ! NH ! Np+ ! Ne !
+

(12)

The contribution to this sum with the largest weight is searched for in two dierent directions, corresponding to the two independent reactions. These are described by the stoichiometric coecients i : Reaction Coord. x y The equilibrium conditions are i=1 Zi Ni
i

H2 H 1 +2 0 1

p+ e 0 0 +1 +1

=1 5

for each reaction

(13)

When we unwind the two expressions represented by Eq. (13) in terms of the stoichiometric coecients i , the factors (2mkT /h)3/2 , and the internal partition functrions, we nd [nH ]2 [nH2 ] [np+ ][ne ] [nH ] 1 eE1 1 eE2
2 2(MH /MH2 )kT h2 3/2

Rx 1 : Rx 2 :

= =

(14)
3/2

2(Mp+ Me /MH )kT h2

(15)

There are two reactions and four reactants. The four concentrations can be expressed in terms of 2 = 4 2 reaction coordinates. We begin with an electrically neutral gas of hydrogen molecules at low temperature, with a concentration of n0 molecules/cm3 . at any nite temperature, [nH2 ] = (1 x)n0 , [nH ] = 2(x y)n0 , and [np+ ] = [ne ] = 2yn0 . Two two equations that dene the values of the two reaction coordinates are now [2(x y)]2 1x (2y)2 2(x y) 3 3 log(|E1 |) + log 2 2 3 3 log(|E2 |) + log 2 2
1 2( 2 MH )|E1 |

1 : log 2:

= =

E1 E2

n0 h2 2Me |E2 | n0 h2
2/3

2/3

(16) (17) (18)

log

This pair of coupled nonlinear equations is certainly formidable looking. However, as long as the two dissociation energies E1 and E2 are very dierent the results are easy to describe. Very dierent means the transition ranges from x 0 x 1 and y 0 y 1 are both narrower than the energy dierence |E1 E2 |. In this case, as a function of increasing temperature x rises from 0 to 1 while y remains 0. After a further increase of temperature y rises from 0 to 1 while x remains 1. The temperatures at which the two transitions take place can be computed as if only a single process were taking place. Note that we cannot use Figs. (2) and (1) in series because they were computed under vastly dierent density conditions. Note also that a density change of 1010 corresponds to a temperature change by a factor of only about 23 101.37 .