Strings and knots: the answer

Just in case you missed it, here's the question.

Let an denote the average number of loops when we start with n strings. We can make a couple key observations to find a recursive equation for an, which will give us the answer.

First, let's do some small-n checking to help gain some understanding. If n=1, then clearly there must be exactly 1 loop at the end. This means that

a1 = 1.

What if n=2? We take one end, and with probability 1/3 (since there are 3 ends to choose from), we tie it to itself. In that case, we must end up with 2 loops. Otherwise, with probability 2/3, we tie it to the other string and we must end up with 1 loop. We get

a2 = (1/3) * 2 + (2/3) * 1 = 4/3.

Now for the general case. We take one end, and we choose randomly from the remaining 2n-1 ends. With probability 1/(2n-1), we tie this string to itself, and from there the situation is exactly as if we had started with 1 loop + n-1 strings. Otherwise, with probability (2n-2)/(2n-1), we tie this end to another string, and from there the situation is exactly as if we had started with n-1 strings.

This breakdown can allow us to express an in terms of an-1 as follows:

an = 1/(2n-1) * (1 + an-1) + (2n-2)/(2n-1) * (an-1) = 1/(2n-1) + an-1.

We can initiate this recursive equation with a1=1, or even a0=0 if we wish (they both give the same answers for all greater n). And we can summarize the end result nicely by writing it as:

an = 1 + 1/3 + 1/5 + ... + 1/(2n-1).

That is, if we start with n strings, then we can find the average number of loops by adding the reciprocals of the odds up to 2n-1.

If we're interested in going a little further, we can notice that

an = H2n - 1/2Hn,

where Hn is the nth harmonic number, which is the sum 1 + 1/2 + 1/3 + ... + 1/n. Hn grows like log(n), the natural logarithm. Thus

an    ~    log(2n) - 1/2 log(n)    ~    1/2 log(n),

where the symbol ~ here means asymptotic equivalence.

blog front page

home