Fermi-Dirac Statistics

Simon Saunders
Fermi-Dirac statistics are one of two kinds of statistics exhibited by identical
quantum particles, the other being Bose-Einstein statistics. Such particles
are called fermions and bosons respectively (the terminology is due to Dirac
[1902-1984] [1]). In the light of the spin-statistics theorem, and consistent
with observation, fermions are invariably spinors (of half-integral spin), whilst
bosons are invariably scalar or vector particles (of integral spin). See spin.
In general, in quantum mechanics, the available states of a homogeneous
many-particle system in thermal equilibrium, for given total energy, are counted
as equiprobable. For systems of exactly similar (identical) fermions or bosons,
states which dier only in the permutation of two or more particles are not only
counted as equiprobable they are identied (call this permutivity). Fermions
dier from bosons in that no two fermions can be in exactly the same 1-particle
state. This further restriction follows from the Pauli exclusion principle. The
thermodynamic properties of gases of such particles were rst worked out by
Fermi [1901-1954] in 1925 [2], and, independently, by Dirac in 1926 [3].
To understand the consequences of these two restrictions, consider a system
of weakly-interacting identical particles, with states given by the various 1-
particle energies c
s
together with their degeneraciesthe number C
s
of distinct
1-particle states of each energy c
s
. From permutivity, the total state of a gas
is fully specied by giving the number of particles with energy c
s
in each of
the C
s
possible states, i.e. by giving the occupation numbers :
s
|
for each :,
/ = 0, 1, , , , C
s
. We suppose all possible states of the same total energy 1 and,
supposing particle number is conserved, of the same total number , are avail-
able to the particles when in thermal equilibrium, i.e. all sets of occupations
numbers that satisfy:
|=cs
P
s,|=1
:
s
|
=
P
s

s
= ;
P
s

s
c
s
= 1. (1)
Since this is quantum mechanics, we suppose that superpositions of such states
are available to the system as well.
Imposing Paulis restriction that no two particles can be in the same 1-
particle state, it follows that the occupation numbers are all zeros and ones and
that C
s
_
s
. The number of distinct sets of occupation numbers :
s
0
, :
s
1
, .., :
s
cs
that sum to
s
satisfying this condition is:
C
s
!

s
!(C
s

s
)!
.
Since the occupation number states span the subspace of the total Hilbert
space to which the
s
particles are conned, this is the dimensionality the
volume of the available spate space for fermions of energy :.
1
For comparison, if the exclusion principle is not obeyed, the number of dis-
tinct sets of :
s
|
that sum to
s
is rather:
(C
s
+
s
1)!

s
!(C
s
1)!
the state-space measure that applies to bosons of energy :. The total number of
distinct sets of occupation numbers for =
P
s

s
particles is then for fermions:
1

=
Q
s
C
s
!

s
!(C
s

s
)!
and for bosons:
1
+
=
Q
s
(C
s
+
s
1)!

s
!(C
s
1)!
.
By conventional reasoning, the equilibrium coarse-grained distribution is that
for which 1

is a maximum. The equilibrium entropy is proportional to the

= / log 1

, where / is Boltzmanns constant.

= / log 1

- /
X
s
[(C
s
log C
s

s
log
s
((C
s

s
) log(C
s

s
)].
If this is to be stationary under independent variation of the numbers
s

s
+c
s
, subject to the constraints (1), then
0 = c log 1

=
X
s
[c
s
log
s
+c
s
log(C
s

s
)].
Were the variations c
s
completely independent each term in this summand
would have to vanish. Introducing undetermined Lagrange multipliers a, ,, for
each of the constraints (1), conclude rather that for each ::
log
s
+ log(C
s

s
) c ,c
s
= 0.
Rearranging:

s
= C
s
(c
o+ots
(1)
1
. (2)
In the case of light quanta, there is no constraint on particle number and the
multiplier c does not occur. The multiplier , meanwhile has its usual meaning,
, = 1,/T, where T is the absolute temperature. C
s
is the number of distinct
1-quanta states in the energy range [c
s
, c
s
+ dc
s
] , where c
s
= /i
s
. It is given
by:
C
s
= 8\ i
2
s
di
s
,c
3
(3)
(obtained either classically, from the wave theory, or by Boses method). From
(2) and (3) the Planck black-body radiation law follows immediately. The num-
bers
s
of (2) are proportional to the radiation energy density in the frequency
range [i
s
, i
s
+di
s
], which can be directly measured.
2
The contrast with the statistics of non-identical particles is that in the lat-
ter case (failing permutivity) there is the further question of which of the
s
particles is in which of the C
s
one-particle states (C
s
s
possible distributions
in all). There is also the question of how the particles are partitioned into
the occupation numbers
1
,
2
, ...
s
, ... . Taking both into account, the total
number of distinct states 1
0
with occupation numbers
1
, ...,
s
, .. is:
!

1
!..
s
!....
Q
s
C
s
s
. (4)
By a similar calculation as before, this yields:

s
= C
s
c
oots
(5)
Evidently (2) (for either sign) and (5) are approximately the same for C
s

s
(equivalently, when c+,c
s
1), and the dierence in the statistics for identical
and non-identical particles disappears.
At the other extreme, for bosons for which C
s

s
, from (2) it follows:

s
= C
s
(c +,c
s
)
1
. (6)
For c = 0, and C
s
as given by (3), (6) is the Rayliegh-Jeans black-body distri-
bution; (5) is the Wien distribution. The discovery of Plancks constant began
with the puzzle of how to understand these distributions, which yielded the
observed long (C
s

s
) and short (C
s

s
) wavelength behaviour respec-
tively, and with Plancks black body formula (2) (with negative sign), obtained
by interpolating between them [10]. The method of counting (4) is associated
with Maxwell-Boltzmann or classical statistics. It was derived, using specically
quantum-mechanical methods, by Ehrenfest [1880-1933] and Uhlenbeck [1900-
1988] immediately after the discovery of Fermis statistics. They concluded
that wave mechanics does not yet per se imply the refutation of Boltzmanns
method [4 p.24]. The dierence, in quantum mechanics, resides solely in the
assumption of permutivity. It is an easy slide to think, since classical statistical
mechanics delivers the same statistics as quantum mechanics for non-identical
particles, that classical particles likewise are non-identical (and do not satisfy
permutivity), i.e. that the correct classical count of states 1
0
is (4). But Gibbs
[1839-1903] had argued for the permutivity of classical particles long before [6],
and for a non-quantized classical phase space, permutivity makes no dierence
to the statistics [11]. That is, computing the volume of classical phase space,
subject to permutivity, rather than a count of equiprobable states, one should
use:
1
0
=
Q
s
C
s
s

s
!
(7)
rather than (4). The logarithm of 1
0
as given by (7) yields an extensive entropy
function, as required [12].
Fermi in 1924 was led to assume that no two electrons could occupy the
same elementary volume in phase space, because only thereby could he obtain
3
agreement with the Sarkur-Stern expressions for the chemical potential and
absolute entropy [5]. That was enough, the following year, to get out a new
equation of state, but little more. Dirac, a few months later, had many more
fragments of the nascent theory of quantum mechanics to hand. He considered
the question of how to formulate permutivity in terms of matrix mechanics
directly. He was led to the question by Heisenbergs dictum: the new mechanics
was to be restricted to observable quantities. In matrix mechanics the observable
quantities were the matrix elements, corresponding to the intensities of the
various transition processes giving rise to line spectra. In the still unresolved
problem of the helium atom, the question arose of how to treat a transition
involving both electrons in one-particle states c
n
, c
n
, of the form (::)
(:
0
:
0
), and its relation to the transition (::) (:
0
:
0
). Only the sum of
the two, Dirac noted, was observable. Hence, in order to keep the essential
characteristic of the theory that it shall enable one to calculate only observable
quantities, one must adopt the second alternative that (::) and (::) count as
only one state. [3 p.667].
Incorporating this into the matrix mechanics (and in particular in terms
of his theory of uniformizing variables) presented certain technical diculties,
whereas in wave mechanics the way forward was much easier (an early indicator
for Dirac that Schrodingers wave theory may have denite advantages over the
matrix mechanics). In the two particle case the state (::) of the composite
system of electrons, labelled 1 and 2, must be of the form
c
nn
= a
nn
c
n
(1)c
n
(2) +/
nn
c
n
(1)c
n
(2) (7)
where a
nn
= /
nn
(and superpositions of such). Dirac observed that the
antisymmetric case (a
nn
= /
nn
) leads to Paulis principle and the symmetric
case to the Bose-Einstein statistical mechanics. He went on to deduce the theory
just sketched; he thought, as had Fermi, that the new statistics, applying as it
did to electrons in the atom, was likely to apply to material gases as well.
Dirac shortly after remarked on the possibility of alternative (more com-
plicated) representations of the permutation group, other than the completely
symmetrized (boson) and antisymmetrized (fermion) representations (in 1930,
in the rst edition of his Principles). These alternatives lead to a variety of
dierent statistics - parastatistics - that are not realized in nature (or not in
3+1 dimensions; special considerations apply to particles eectively restricted
to two spatial dimensions). It was thought, for a time, that they might oer an
alternative to the quark model of deep inelastic scattering, but without success
[13].
Heisenberg [1901-1976] as well as Dirac had been preoccupied with the he-
lium problem. His earlier papers in 1926 on the helium and related 2-electron
spectra had made use of the Pauli exclusion principle and, for the rst time, the
Schrdinger wave mechanics (albeit only as a calculational tool). He too arrived
at the two classes of states (7), but under a somewhat dierent interpretation
from Diracs and with no understanding of the fact that they gave rise to dier-
ent statistics. He was led, rather, to an idea absent from Diracs paper that
4
a two-electron system, each with identical allowed energies 1
n
(1) = 1
n
(2),
1
n
(1) = 1
n
(2) (with 1
n
1
n
), would in wave-theoretic terms be subject to
resonance, with energy 1
n
1
n
passing from one electron to the other under
the transition (::) (::) (states that Dirac had identied). Likewise the
perturbation due to the electron charge will in general contain terms corre-
sponding to transitions in which the systems 1 and 2 switch places (den Platz
tauschen) [7 p.417].
Thus did the idea of exchange forces rst arise. A similar interpretation
was advanced by Heitler [1904-1981] and London [1900-1954] the following year
in their treatment of the homopolar bond [8]. But by this time, as Heitler
went on to remark, this question of interpretation had become closely wed to
disputes over other interpretative issues in quantum mechanics, notably over
Schrdingers continuous beat picture of emission and absorption processes as
compared to Borns statistical interpretation [14]. What was being exchanged,
Heitler concluded, remained completely unclear. ([9 p.48]).
What was clear was that in any of the symmetric, triplet states of spin,
for which the spatial wave-function must be antisymmetric, the norm of the
wave-function for electron coordinates close together is extremely small (and
for coinciding coordinates, vanishes). In this sense electrons in bound states
with correlated spins eectively repel one another. Those with anticorrelated
spins, in the antisymmetric singlet state, have greater amplitudes for small
relative distances, for their spatial wavefunction must then be symmetric - the
amplitude is much greater than if there were no overall symmetry requirement
on the state (the case of non-identical fermions). This eect is independent of
the Coulomb force altogether, and plays a key role in ferromagnetism as well in
the chemical bond, as Heisenberg was shortly to show, again with reference to
electron exchange, and exchange forces.
Whether interpreted as an exchange force involving the identity of quan-
tum particles over time, or as a consequence of permutivity and the Pauli ex-
clusion principle, Fermi-Dirac statistics is fundamental to the whole of quantum
chemistry and throughout the physics of the solid state.
Literature
Primary
[1] P.A.M. Dirac: The Principles of Quantum Mechanics, 3rd edition, Oxford:
Oxford University Press (1945)
[2] E. Fermi: Aur Quantelung des idealen einatomigen Gases, Zeitschrift fr
Physik 36 902-912 (1926)
[3] P.A.M. Dirac: On the theory of quantum mechanics, Proceedings of the Royal
Society (London) A112, 281-305 (1926)
[4] P. Ehrenfest and G. Uhlenbeck: Die wellenmechanische Interpretation der
Botzmannschen Statistik neben der neueren Statistiken, Zeitschrift fr Physik,
41, 24-26 (1927)
5
[5] E. Fermi: Considerazioni sulla quantizzazione dei sistemi che contegno degli
elementi identici, Nouvo Cimento 1, 145-152 (1924)
[6] J. Gibbs: Elementary Principles in Statistical Mechanics, Woodbridge, Con-
necticut: Ox Bow Press (1981)
[7] W. Heisenberg: Merhrkorperproblem und Resonanz in der Quantnmechanik,
Zeitschrift fr Physik, 38, 411-426 (1926)
[8] W. Heitler and F. London: Wechselwirkung neutraler Atome und homopo-
lare Bindung nach der Quantenmechanik, Zeitschrift fr Physik 44, 455-472
(1927)
[9] W. Heitler: Strungsenergie und Austausch beim Merhkrprobleme, Zeitschrift
fr Physik 46, 47-72 (1927)
Secondary
[ 10] F. Hund: Geschichte der Quantentheorie, Manheim: Bibliographisches
Institut. Translated as The History of Quantum Theory, G. Reece (trans.),
London: Harrap (1974)
[11] N. Huggett: Atomic metaphysics, Journal of Philosophy, 96, 5-24 (1999)
[12] S. Saunders: On the explanation for quantum statistics, Studies in the
History and Philosophy of Modern Physics 37 192-211 (2006)
[13] S. French: The esperable uberty of quantum chromodynamics, Studies in
the History and Philosophy of Modern Physics 26 87-105 (1995)
[14] C. Carson: The peculiar notion of exchange forcesI: origins in quantum
mechanics, 1926-28, Studies in the History and Philosophy of Modern Physics
27, 23-45 (1996)
6