Biomedical Engineering Reference
In-Depth Information
A standard method for tackling constrained variational problems of this kind is
Lagrange's method of undetermined multipliers , which is discussed in detail in all the
advanced texts. I will just outline the solution without justification.
The first step in our derivation is to take logarithms of either side of expression (7.6):
p
ln W
=
ln N
!−
ln N i !
(7.8)
i = 1
For large values of n it is appropriate to make use of the Stirling formula for factorials
n
ln n
O 1
n 2
1
2 ln 2π
1
2
1
12 n +
ln n
!=
+
+
n
+
+···
(7.9)
In practice all terms except n ln n are usually omitted and we have
p
ln W
=
ln N
!−
( N i ln N i !−
N i )
(7.10)
i
=
1
We therefore look for a maximum of ln W . We know from elementary differential
calculus that
p
∂ ln W
N i
dln W
=
d N i =
0
(7.11)
i = 1
at a stationary point. We cater for the constraints by differentiating Equations (7.7)
0
=
d N 1 +
d N 2 +···+
d N p
d N p ε p
multiplying by arbitrary undetermined multipliers that I will call α and
0
=
d N 1 ε 1 +
d N 2 ε 2 +···+
β and then adding:
p
p
p
∂ ln W
N i
dln W
=
d N i +
α
d N i
β
ε i d N i
(7.12)
i = 1
i = 1
i = 1
All the d N i can now be treated as independent variables and at a stationary point
∂ ln W
N i
+
α
βε i =
0
(7.13)
All that remains is to use the Stirling formula, and the final result turns out as
N i =
exp (α) exp (
βε i )
(7.14)
Next we have to evaluate the undetermined multipliers α and β; if we sum over all the
allowed quantum states then we have
p
p
N
=
N i =
exp (α)
exp (
βε i )
i = 1
i = 1
which gives
N
exp (α)
=
(7.15)
i = 1 exp (
p
βε i )
 
Search WWH ::




Custom Search