Question about the recursion relation for the 1d random walk and Pascal's rule

56 Views Asked by At

Reading "Elements of the Random Walk" written by Joseph Rudnick and George Gaspari, i am confused by the recursion relation they gave on page 13. the book says that For 1d discrete random walk with step length $l$ and equal probability of traveling right and left at any site, $$C(N{\rm{ }};{\rm{ }}x,{\rm{ }}y) = C(N - 1{\rm{ }};{\rm{ }}x,{\rm{ }}y - l) + C(N - 1{\rm{ }};{\rm{ }}x,{\rm{ }}y + l) \tag{1}$$ with $C(N{\rm{ }};{\rm{ }}x,{\rm{ }}y)$ is equal to the number of walks that start at x and end up at y.

The book says that the above equation is equivalent to Pascal's rule $$\frac{{N!}}{{n!(N - n)!}} = \frac{{(N - 1)!}}{{(n - 1)!(N - n)!}} + \frac{{(N - 1)!}}{{n!(N - n - 1)!}}$$

I know that $$P(N{\rm{ }};{\rm{ }}x,{\rm{ }}y) = \frac{1}{{{2^N}}}\frac{{N!}}{{n!(N - n)!}}{\rm{ = }}\frac{1}{{{2^N}}}C(N{\rm{ }};{\rm{ }}x,{\rm{ }}y)\tag{2}$$ if $y-x=nl$. The master equation is $$P(N{\rm{ }};{\rm{ }}x,{\rm{ }}y) = \frac{{\rm{1}}}{{\rm{2}}}P(N - 1{\rm{ }};{\rm{ }}x,{\rm{ }}y - l) + \frac{{\rm{1}}}{{\rm{2}}}P(N - 1{\rm{ }};{\rm{ }}x,{\rm{ }}y + l)\tag{3}$$

where $P(N{\rm{ }};{\rm{ }}x,{\rm{ }}y)$ is the probability of the walk that start at x and end up at y. The combination of equation 2 and 3 will give equation 1.

What puzzles me is that as $$C(N - 1{\rm{ }};{\rm{ }}x,{\rm{ }}y + l) = \left( {\begin{array}{*{20}{c}} {N - 1}\\ {n + 1} \end{array}} \right){\rm{ = }}\frac{1}{{{2^{N - 1}}}}\frac{{(N - 1)!}}{{(n + 1)!(N - n - 2)!}}$$

Should equation 1 gives $$\frac{{N!}}{{n!(N - n)!}} = \frac{{(N - 1)!}}{{(n - 1)!(N - n)!}} + \frac{{(N - 1)!}}{{(n + 1)!(N - n - 2)!}}$$

This seems to be different from the Pascal's rule, so where am i wrong. Thanks for your help!


After discussion with @epi163sqrt, I found that my mistake was that I take for granted that $y-x=nl$ for $C(N {\rm{ }};{\rm{ }}x,{\rm{ }}y )$ as i wrote above. Instead it should be $y-x=(2n-N)l$.

I didn't change the origin post in order to help guys who make similar mistake.

1

There are 1 best solutions below

6
On BEST ANSWER

Hint: The right-most term of (1) and the right-most term of Pascal's rule coincide. We therefore have \begin{align*} C(N-1;x;y+l)=\frac{(N-1)!}{n!(N-n-1)!}\color{blue}{=\binom{N-1}{n}}\tag{1} \end{align*}

Comment: The identity (1) is valid, since we have the situation \begin{align*} C(N;x;y)=\frac{N!}{n!(N-n)!} \end{align*} where we consider according to the book $N$-step walks with

  • $n$ steps to the right and

  • $N-n$ steps to the left

When we now look at the right-most term of \begin{align*} C(N;x;y)=C(N-1;x;y-l)+C(N-1;x;y+l) \end{align*} with the interpretation of $n$ given above, we see, that since $y+l$ is to the right of $y$ we have done all $n$ steps to the right. Since the right-most term represents the number of $N-1$-step walks to $y+l$, we have also done $(N-1)-n$ steps to the left, resulting in \begin{align*} \binom{N-1}{n} \end{align*} different walks.