Assignment 5

DEPARTMENT OF PHYSICS
INDIAN INSTITUTE OF TECHNOLOGY, MADRAS
PH5820 Classical Physics Assignment 5 23.9.2022
Microstates, macrostates, probabilities, & all that

A good way to understand many of the basic concepts of statistical physics, such as mi-
crostates, macrostates, accessible microstates, probabilities, etc., is to work out a toy model
in some detail. This is the purpose of this Problem Set.
The simplest model we can construct is a system consisting of N independent elemen-
tary constituents, each of which can have only a small number of states – in fact, just
two states, to keep things really simple. The obvious example of such a “two-level” or
“two-state” object is a coin, which can be either in the “heads” state (denoted by h), or
in the “tails” state (denoted by t). Thus we have a much simpler situation than in a gas,
in which each of the molecules has a very large number of possible states in general. But
our example is not totally unphysical — there are numerous physical systems in which the
constituents have just two (or some such small number of) possible states. Examples are
binary alloys, paramagnets, etc. We will discuss some of these in brief here, and in greater
detail later on.
Our system is thus a set of N coins. Each coin is tossed once, and the coins are laid
out in a row from coin 1 to coin N .
A microstate of our system is a detailed specification of the state (h or t) of each of
the N coins: e.g., hhhh . . . hhh, or hthhthtth . . . ht, and so on. In other words, each
“word” of N letters, made up of just the two letters h and t, is a possible microstate
of our system.
A macrostate of the system is specified as follows. Let H and T denote the total
number of heads and tails, respectively, in the collection. Then the pair of values
(H, T ) specifies a macrostate of the system. Clearly, in general, a knowledge of H
and T is not sufficient to tell us whether any particular coin is in the h state or the
t state. It is therefore obvious that a macrostate of the system provides much less
detailed information about the system than a microstate does. Equivalently, much
less information is necessary to specify a macrostate.
A macrostate of our system could also be specified by the pair (N, M ) where N is
the total number of coins, and M is the difference between the total number of heads
and the total number of tails, because N = H + T and M = H − T . Further, if N
is fixed once and for all, then M alone needs to be specified to label a macrostate of
the system.
1
A few remarks on the analogy between our toy model and a paramagnet will be helpful. In
its simplest form, a paramagnetic specimen consists of a large number N of independent
elementary (or atomic) magnetic dipole moments, each of magnitude µ. Suppose each such
moment can align itself either parallel or anti-parallel to an applied magnetic field. Then
N = N↑ + N↓ where N↑ is the total number of moments parallel to the applied magnetic
field, and N↓ is the number antiparallel to it. The net magnetic moment of the system
is µ(N↑ − N↓ ) = M . (This is why we chose the symbol M for H − T in our toy model!)
However, two important points must be noted.
First, each atomic moment may actually point at any angle θ (0 ≤ θ ≤ π) to the field,
contributing µ cos θ (rather than just +µ or −µ) to the magnetization in the direction
of the field. This is true in some cases. But in other cases in which the moments must
be treated quantum mechanically, it turns out that each atomic moment can only have a
discrete set of possible orientations – in particular, just orientations parallel or anti-parallel
to an applied magnetic field, as assumed above.
The second point is that, unlike a “fair” or unbiased coin which can be in a state h
or state t with equal probabilities (= 21 ), an atomic moment may actually have different
probabilities for being oriented parallel or antiparallel to an applied field. In fact, this
is just what happens in general. Later on, we shall deduce the actual values of these
probabilities. It will turn out that at sufficiently high temperatures the two probabilities
(of parallel or antiparallel orientations with respect to the applied field) tend to become
equal to each other. It is this situation, therefore, that our toy model simulates. It is easy
to generalise it to the case of “unfair” or biased coins, to reproduce what happens in a
paramagnet at finite temperatures. For the moment, let’s stick to fair coins. Our system
simply consists of N fair coins.
1. (a) What is the total number of microstates of the system?

(b) Are all microstates equally probable?
(c) What is the à priori probability of occurrence of each microstate? (“A priori”
means before the event happens, i.e., before the coins are tossed.)
(d) What is the number of accessible microstates in each of the following macrostates?
(i) M = N + 1; (ii) M = N ; (iii) M = N − 1; (iv) M = 1; (v) M = 0; (vi)
M = −N ; (vii) M = −N − 1.
(e) Hence determine the physical range of M , for a given value of N .
(f) Find an expression for the number of accessible microstates in a given macrostate
M.
(g) Hence find the most probable macrostate (for a given N ). Observe that all
the macrostates of the system are not equally probable. Compare this with the
situation in (b) above.
2
(h) What is the probability of obtaining the most probable macrostate?
(i) What is the probability PN (M ) of obtaining a general macrostate labelled by
M?
(j) Analyse the variation of PN (M ) with M , for a given N , in the case when N is
very large. In particular, compare the probabilities of the most probable and
the least probable macrostates.
Factorials of large numbers: Stirling’s Formula

It is instructive to put in some actual numbers in the model above. Let’s take N to be
100, a decently large number (but nowhere near 1023 , of course). We find that factorials
occur in the expressions we have to calculate, via n Cr , etc. Now N ! is an extremely rapidly
increasing function of N – so rapidly increasing that standard pocket calculators won’t even
give you 70!, because this number exceeds 1099 . Fortunately, there’s a very interesting and
ubiquitous formula called Stirling’s approximation, that gives an excellent estimate of N !
for large values of N . This formula is
1 1

1
N −N
N! = N e (2πN ) 2 1+ + a term of order 2 + · · · .
12N N
For sufficiently large N , it suffices to retain just the term 1 in the curly brackets. To see
how good this approximation is, try it out for N = 1, 2, 5, 10, 20, . . . , 60 – you can
see how rapidly the approximate formula approaches the actual value. Clearly, even for
N = 103 , you can’t distinguish between the LHS and RHS for all practical purposes – let
alone for the case N = 1023 . A good bit of statistical physics is based on the fantastic
accuracy of this formula – which is an example of what is called a “law of large numbers”.
It turns out that in physical applications we need the logarithm of the factorial, rather
than the factorial itself. Stirling’s formula is then
1 1
ln N ! ' (N + ) ln N − N + ln(2π) + (terms that tend to zero as N → ∞) .
2 2
The point to remember is that ln N ! is of the order of (N ln N − N ) for very large N .
Stirling’s formula is one of the few things that are worth memorizing.
2. Use Stirling’s formula to go over questions (a) to (j) for the case N = 100.
3. When N is very large, we can think of M as essentially varying continuously. (Steps

of unity are neglible compared to N .) Then we could define Ω(M ), the number
of states (microstates) in a small range δM about the value M . What does this
look like? Does it increase monotonically as M increases, the way Ω(E) does for a
3
collection of particles? Explain the reason for the difference, if any. (Going back to
the paramagnet analogy, observe that if H is the applied magnetic field, the total
energy of the system cannot be greater than N µH in magnitude.)
4. Density fluctuations in a gas: Consider an ideal gas of N particles in equilibrium

in a container of volume V . The average number density is N/V , a constant. Call
this ratio ρ. Consider a sub-volume v of the container.
(a) Find the probability p(n) that the sub-volume contains exactly n particles at a
given instant of time. Verify that p(n) is properly normalized, i.e., that
N
X
p(n) = 1 .
n=0
This is called the binomial distribution, for what should be an obvious reason.
(b) Use Stirling’s formula to work out what happens to p(n) in the limit in which
N → ∞ and V → ∞, keeping the number density ρ = N/V constant and finite.
This is called the thermodynamic limit in statistical mechanics. You must
show that p(n) is the Poisson distribution in this limit, i.e.,
p(n) = (ρv)n e−ρv /n! (n = 0, 1, 2, . . .)
in this case. Find the (i) mean value of n, (ii) the mean squared value of n, (iii)
the standard deviation of n, and (iv) the relative fluctuation in n, defined as
the ratio (standard deviation)/(mean).
(c) Suppose now that the system is a mixture of two noninteracting gases, A and
B. Suppose the probability that v contains nA molecules of gas A is given to be
pA (nA ) = (λA )nA e−λA /nA ! (λA = positive constant.)
Similarly, suppose the probability that v contains nB molecules of gas B is given

by
pB (nB ) = (λB )nB e−λB /nB ! (λB = positive constant.)
Assuming the gases to be independent of each other, find the probability that
v contains a total of n molecules at any instant of time.

Assignment 5

Uploaded by

Assignment 5

Uploaded by

DEPARTMENT OF PHYSICS

INDIAN INSTITUTE OF TECHNOLOGY, MADRAS

PH5820 Classical Physics Assignment 5 23.9.2022

Microstates, macrostates, probabilities, & all that

1. (a) What is the total number of microstates of the system?

Factorials of large numbers: Stirling’s Formula

3. When N is very large, we can think of M as essentially varying continuously. (Steps

4. Density fluctuations in a gas: Consider an ideal gas of N particles in equilibrium

p(n) = (ρv)n e−ρv /n! (n = 0, 1, 2, . . .)

pA (nA ) = (λA )nA e−λA /nA ! (λA = positive constant.)

Similarly, suppose the probability that v contains nB molecules of gas B is given

You might also like