next up previous
Next: About this document ...

CS-274 - Math-284 Combinatorics and Probability
Homework - May 28, 1998
Instructor: László Babai Ryerson 164 e-mail: laci@cs
Student: Nick Russo e-mail: n-russo@uchicago.edu

1.
Prove that the number of $2 \times n$ latin rectangles is asymptotically equivalent to $\frac{\left(n!\right)^2}{e}$.

The numbers in the top row can be ordered n! ways, so the total number of latin rectangles is n! times the number of ways to construct a second row such that no conflict is implied.

There are n! total ways to arrange the elements of the second row, so there are less than n! ways that are ``good''.

I should be able to use Inclusion/Exclusion to count the number of permutations that fix 1, those that fix 2, those that fix 1 and 2, etc... and in doing so, I should be able to find out the size of the set of ``good'' permutations.

When I use Inclusion/Exclusion, I find that:


$\displaystyle \left\vert \left\{ good\ permutations \right\} \right\vert$ = $\displaystyle \sum_{i=0}^{n}{n \choose i}\left(n-i\right)!
\cdot \left(-1\right)^{i}$ (1)
  = $\displaystyle - {n \choose 1}\left(n-1\right)!
+ {n \choose 2}\left(n-2\right)!
- \cdots$  

But each term contains $\left(n-i\right)!$ on the top, and hidden away in the bottom term, so they cancel out and leave $\frac{n!}{i!}$.


  = $\displaystyle \sum_{i=0}^{n} \frac{n!}{i!} \left(-1\right)^i$ (2)
  = $\displaystyle n! \cdot \sum_{i=0}^{n} \frac{1}{i!} \left(-1\right)^i$ (3)

But that sum is just an expansion of $\frac{1}{e}$ 1, as $n \rightarrow
\infty$, so we have the number of ``good permutaions'' $\sim n! \cdot
\frac{1}{e}$. Then, as noted above, the total number of latin rectangles $\sim n! \times \frac{n!}{e}$, or $\frac{n!^2}{e}$.



 
next up previous
Next: About this document ...
nicholas
1998-06-06