80 1 Mathematical Physics
N!→
√
2 πNNNe−N
x!→
√
2 πxxxe−x
(N−x)!→
√
2 π(N−x)(N−x)N−xe−N+x
B(x)→f(x)=
√
N
2 πx(N−x)
pxqN−xNN
xx(N−x)N−x
=
√
N
2 πx(N−x)
(
Np
x
)x(
Nq
(N−x)
)N−x
Letδdenote the deviation ofxfrom the expected valueNp, that is
δ=x−Np
thenN−x=N−Np−δ=Nq−δ
f(x)=
[
2 πNpq
(
1 +
δ
Np
)(
1 −
δ
Nq
)]− 1 / 2
(
1 +
δ
Np
)−(Np+δ)
·
(
1 −
δ
Nq
)−(Nq−δ)
Assume that|δ|Npqso that
∣
∣
∣
∣
δ
Np
∣
∣
∣
∣1 and
∣
∣
∣
∣
δ
Nq
∣
∣
∣
∣^1
The first bracket reduces to (2πNpq)−
(^12)
. Take logeon both sides.
ln
[
f(x)(2πNpq)
12 ]
=−(Np+δ)ln
(
1 +
δ
Np
)
−(Nq−δ)ln
(
1 −
δ
Nq
)
=−(Np+δ)
(
δ
Np
−
δ^2
2 N^2 p^2
+
δ^3
3 N^3 p^3
−···
)
−(Nq−δ)
(
−
δ
Nq
−
δ^2
2 N^2 q^2
−
δ^3
3 N^3 q^3
−···
)
=−
δ^2
2 Npq
−
δ^3 (p^2 −q^2 )
6 N^2 p^2 q^2
Neglect theδ^3 term and puttingσ=
√
Npqandσ=x−Np= ̄x.
f(x)=
1
σ
√
2 π
exp
[
−(x− ̄x)^2
2 σ^2
]
(Normal or Gaussian distribution)