September 13, 2001
Reading: Chapter Four
Homework: 4.1,4.2,4.3,4.4
Determining the most probable distribution:
The basic postulate of statistical mechanics is that each microstate is equally
possible. Thus, the number of microstates in a given distribution defines the
probability of the distribution. And the most probable distribution of microstates
is the one for which j is a maximum.
Note: the section title 4.4 Determination of the most probable microstate is a
BAD statement, because all the microstates are equally probable. It should be the
most probable distribution of microstates or the most probable macrostate that
corresponds to the most probable distribution of microstates.
Therefore, our objective is to obtain the most probable distribution by maximizing
given by (for convenience, we drop the superscript j) :
N!
N!
=
=
n0 !n1!n 2 !n3 !
ni !
i=0
subject to the following two constraints:
n
i =0
and
=N
n
i
i =0
=U .
Maximizing is equivalent to maximizing ln; it is then easier to deal with ln
rather than because of the use of the Stirlings relation.
Using the Stirlings approximation (see appendix):
ln N != N ln N N ,
when N>>1,
So,
i=0
i =0
i =0
ln = N ln N N ni ln ni + ni = N ln N ni ln ni
is to be maximized subject to
n
i =0
N =0
n
i =0
U = 0
Using the Lagranges method for maximizing (or minimizing) a function (see
appendix), we define a master function F,
i =0
i =0
F = ln + ( ni N ) + ( ni i U )
where and are called Lagranges multipliers.
Note that nis are not all independent variables for function ln because of the
constraint relations between the nis. Consequently, d(ln) would not be exact
differential for determining the extremum of ln. Thus, the purpose of the
Lagranges method is to introduce a new mater function F with additional two
unknown variables ( and ) that effectively converts all the nis into independent
variables for function F. Consequently, dF will be the exact differential for
determining the extremum of F.
For extremum, dF = 0,
i =0
i =0
i =0
dF = d (ln ) + dni + d ( ni N ) + i dni + d ( ni i U )
i =0
d (ln ) = ( dni ln ni + dni )
i =0
i=0
i =0
i =0
i =0
i =0
dF = ln ni dni + dni + dni + i dni
F
dni = (ln ni + 1 + + i )dni = 0
i = 0 ni
i =0
F
Since for extremum,
= 0 for each i, we must have
ni
ln ni + 1 + + i = 0
So,
dF =
i.e.,
ni = e 1 e i .
Now,
N = ni = e 1 e i , and e 1 =
i =0
i =0
i =0
So,
ni =
Ne i
i =0
Thus, these values of ni extremizing F.
F
Furthermore,
= ln ni + 1 + + i ;
ni
2F
1
> 0 , assuming is a constant independent of ni.
ni
ni
This means F must be a minimum. Because
So,
i =0
i =0
F = ln + ( ni N ) + ( ni i U ) ,
then ln must be a maximum. So, these values of ni give rise to the most probable
distribution of microstates.
Partition Function:
We have shown that the number of microstates is maximized leading to a most
probable distribution with
Ne i
N
= exp( i )
ni =
Z
e i
i =0
Where Z = exp( i ) is know as partition function, which is a constant for a
i =0
given system with fixed energy levels as is constant (see below). Thus, ni, the
occupation number of the energy level i, is proportional to exp( i ) , i.e.
uniquely defined by the energy value of this level. The higher the energy, the less
the number of particles.
Appendix
1. Stirlings relation: ln N != N ln N N ,
when N>>1.
N != N ( N 1)( N 2) 3 2 1
ln N != ln N + ln( N 1) + ln( N 2) + + ln 3 + ln 2 + ln 1
N
ln N != ln x = ln x dx, (dx = 1)
N
ln N ! ln x dx [ x ln x x]1N N ln N N + 1 N ln N N
1
2. Lagranges method for functional extremization under constraints:
The object is to extremize a function of many variables subject to a set of
constraints. It is understood that the number of constraints is much less than the
number of variables.
Suppose there is a function f ( x1 , x 2 ,..., x n ) to be extremized under m sets of
constraints (m < n), given by
1 ( x1 , x 2 ,..., x n ) = 0
2 ( x1 , x 2 ,..., x n ) = 0
m ( x1 , x 2 ,..., x n ) = 0
Lets define a master function F,
F ( x1 , x 2 ,..., x n ) = f ( x1 , x 2 ,..., x n ) + 1 ( x1 , x 2 ,..., x n )1 ( x1 , x 2 ,..., x n )
+ 2 ( x1 , x 2 ,..., x n )2 ( x1 , x 2 ,..., x n ) + ... + m ( x1 , x 2 ,..., x n )m ( x1 , x 2 ,..., x n )
so that xis become independent variables for F. Where ( x1 , x 2 ,..., x n ) are
unknown functions called Lagrange multipliers.
Since all the = 0 , it is easy to see when F is extremized, so does f.
n
F
Now, dF =
dxi ,
i = 0 x i
m
m
j
j
F f
m
f
Where
( j j ) =
=
+
+ j
+ j
xi xi xi j =1
xi j =1
xi
xi
j =1
m
j
Since j = 0 for all j, j
= 0,
xi
j =1
m
j
F f
So,
.
=
+ j
xi xi j =1
xi
F
= 0 for all xi.
Thus, F will be extremized, if
xi
And we will have n independent equations, namely,
m
j
f
+ j
=0
x1 j =1
x1
m
j
f
+ j
=0
x 2 j =1
x 2
m
j
f
+ j
= 0.
x n j =1
x n
Simultaneous solutions of the above equations will then yield m multipliers
m
( 1 , 2 ,..., m ) that extremize function F = f + j j , and hence f, as j = 0 .
j =1