Our lecturer today mentioned how a piston's head being at equal pressure maximised the multiplicity of states.

enter image description here

He mentioned the following:

If I have a fixed number of particles $N_A$ on left and $N_B$ on the right, and the whole system has a fixed total volume of $M$, we can say the movable piston partitions the total volume into $M_A$ and $M_B$ lattice sites respectively on each side, and therefore

$$M_{total} = M_{A}+M_{B} = constant$$

Hence if we want the volumes of the two sides that maximise the multiplicity function,

$$\Omega(N, M)=\frac{M !}{(M-N) !} \frac{1}{N !}$$

This all made perfect sense. The next step is where I have confusion.

As we are dealing with an ideal gas, the densities are so low that $\frac{N}{M} << 1$ hence

$$\frac{M !}{(M-N)!} \approx M^{N}$$

I understand the approximation as there are fewer molecules N than there are sites M for an ideal gas, but I fail to understand how he managed to reach the conclusion above from the previous equation above that.

I tried to solve this using the approximation:

$$x! = \left(\frac{x}{e}\right)^x$$ but this failed to reach me to the correct working in order to prove the $\approx M^{N}$ approximation.

I managed to get:

$$\frac{M^M}{(M-N)^{M-N}}\times \frac{1}{N^N}$$

but I suspect this is deviating from the main way of reaching the approximation.

How is the approximation achieved?

  • $\begingroup$ Your multiplicity expression $\Omega$ has a factor $1/N!$ which is missing from the approximation in your title, and in the line you quote after "densities are so low." Is that intentional? $\endgroup$ – rob May 18 '19 at 0:04
  • $\begingroup$ @rob yes the line given in the notes was "We are dealing with gases so low densities N/M<<1" and the approximation was also given as above without the $\frac{1}{N!}$ factor as part of the notes! $\endgroup$ – vik1245 May 18 '19 at 0:10
  • $\begingroup$ @BobSmith By the way: If you accidentally created two accounts, please use the "contact" link at the bottom of any page to request a merge. $\endgroup$ – rob May 18 '19 at 0:50
  • $\begingroup$ @rob have contacted now and just realised! Many thanks. $\endgroup$ – David Smith May 18 '19 at 9:33

$$ \frac{M !}{\left(M-N\right)!} ~=~ \begin{alignat}{10} M & \times & \left(M - 1\right) & \times & \left(M - 2\right) & \times & ~\cdots~ & \times & \left(M - N + 1\right) & \times & \left(M - N\right) & \times & \left(M - N - 1\right) % & \times & \left(M - N - 2\right) & \times & ~\cdots~ % & \times & 2 & \times & 1 \\[-25px] \hline & & & & & & & & & & \left(M - N\right) & \times & \left(M - N - 1\right) % & \times & \left(M - N - 2\right) & \times & ~\cdots~ % & \times & 2 & \times & 1 \end{alignat} $$

$$ \begin{align} &~=~ M \times \left(M - 1\right) \times \left(M - 2\right) \times ~\cdots~ \times \left(M - N + 1\right) \\[10px] % &~=~ \prod_{i=M-N+1}^{M}{i} \end{align} $$

Since $\frac{N}{M} \ll 1 ,$ then $M \gg N,$ and $M \gg 1,$ so $$ \begin{align} \frac{M !}{\left(M-N\right)!} &~=~ M \times \underbrace{\left(M - 1\right)}_{\approx M} \times \underbrace{\left(M - 2\right)}_{\approx M} \times ~\cdots~ \times \underbrace{\left(M - N + 1\right)}_{\approx M} \\[10px] &~\approx~ \underbrace{M \times \cdots \times M}_{N~\text{times}} \\[10px] &~=~ M^N \,, \end{align}$$ so $$ \frac{M !}{\left(M-N\right)!} ~\approx~ M^N \,.$$

| cite | improve this answer | |
  • $\begingroup$ Apparently MathJax complains about nesting align blocks? $\endgroup$ – Nat May 18 '19 at 0:18
  • $\begingroup$ More like the line width got overflow. Nice typesetting by the way. $\endgroup$ – acarturk May 18 '19 at 0:28

$$M! = M \times (M-1) \cdots \times (M-N+1) \times (M-N)!$$ $$\frac{M!}{(M-N)!}=\frac{M \times (M-1) \cdots \times (M-N+1) \times (M-N)!}{(M-N)!}$$ The terms $(M-N)!$ cancel, $$\frac{M!}{(M-N)!}=\prod\limits_{i=0}^{N-1}\,(M-i)$$ We know $i < N$. Assuming $N \ll M$, we get $i \ll M$ and we can write $(M-i)$ as $M$, then $$\frac{M!}{(M-N)!}=\prod\limits_{i=0}^{N-1}\,M = M^N.$$

Sorry, my initial response was wrong.

| cite | improve this answer | |
  • $\begingroup$ But surely if you approximate $(M-N) \approx M$ then $\frac{M!}{(M-N)!} \ approx \frac{M!}{M!} = 1$ how can that work? or am I thinking about this matter incorrectly? $\endgroup$ – vik1245 May 18 '19 at 0:00
  • $\begingroup$ @BobSmith I realised that I did some mistakes. Can you look again?. $\endgroup$ – acarturk May 18 '19 at 0:02
  • $\begingroup$ I understand the first part correctly but my only confusion comes from how you managed to go from $\frac{M!}{(M-N)!}$ to $\prod\limits_{i=0}^{N-1}\,M$ although I understand why you write $(M-i)$ as $M$ using the approximation - once I understand that then I can accept your answer! The final step makes sense too - just the middle part of the final step! $\endgroup$ – vik1245 May 18 '19 at 0:05
  • $\begingroup$ @BobSmith The assumption $N<<M$ also means that the dummy variable $\forall i: i<<M$. The trick is to open the biggest $N$ terms of $M!$, and see that the rest, i.e. $(M-N)!$ cancel with the denominator. After that? $M>>i$ therefore the first degree terms should consist only of $M$. Since this assumption gives us the correct result, I need not factor in $i$. $\endgroup$ – acarturk May 18 '19 at 0:09
  • $\begingroup$ Minor note: the spacing is nicer if you use \ll and \gg rather than << and >>. Compare $\ll$ to $<<$. $\endgroup$ – rob May 18 '19 at 0:15

Just to see how to do it using Stirling's formula: \begin{align} \frac{M!}{(M- N)!} &= \frac{M^M e^{-M}}{(M-N)^{M-N} e^{-(M-N)}} \\ &= \frac{M^N}{e^N} \left( \frac{M}{M- N} \right)^{M-N} \\ &= \frac{M^N}{e^N} \left( 1 - \frac{N}{M-N} \right)^{-(M-N)} \end{align} But since $N \gg M$, we have $x \equiv (M-N)/N \gg 1$ and so
$$ \left( 1 - \frac{1}{x} \right)^{-Nx} = \left[ \left( 1 - \frac{1}{x} \right)^{x}\right]^{-N} \approx e^{N} $$ using the approximation $(1 - 1/x)^x \approx e^{-1}$ for $x \gg 1$. Thus, $$ \frac{M!}{(M- N)!} \approx M^N. $$

More generally, you can use Stirling's formula in the form $\ln n! \approx n \ln n - n$ to calculate corrections to this approximation. If I've done my derivation correctly, for example, it turns out that $$ \ln \left[ \frac{M!}{(M- N)!} \right] \approx N \ln M + \mathcal{O} \left(\frac{N}{M}\right)^2. $$

| cite | improve this answer | |

When density is low one could in principle follow the particles for a while. Assume the particles are distinguishable. Suppose there is one particle and that there are six sites: Ω = 6. Now two particles, a red one and a blue one - Ω = 36. Etcetera.

Interesting to note though that this approximation fails if there is a small hole in the partition. What final distribution would maximize Ω?

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.