I have heard that Legendre functions are important in number theory. Can any one tell me how?

The Legendre function of the first kind $P_s$ is defined by \begin{eqnarray*}P_s(x) =& \frac{1}{2\pi}\int_{-\pi}^{\pi}\left(x+\sqrt{x^2-1}\cos\theta\right)^s d\theta
\newline =&\frac{1}{\pi}\int_0^1\left(x+\sqrt{x^2-1}(2t-1)\right)^s\frac{dt}{\sqrt{t(1-t)}},& s\in\mathbb{C}, x\ge 1. \end{eqnarray*}

  • 1
    $\begingroup$ I've never been fond of that integral representation TBH; I had always preferred the generating function or Rodrigues or even the hypergeometric definition. $\endgroup$ – J. M. is a poor mathematician Sep 20 '10 at 12:06
  • $\begingroup$ @J. M. Sure, but, Rodrigues formula is for Legendre polynomials and Hypergeometric representation is more adequate on a disc. We can also define $P_s$ as the solution to $\frac{d}{dx}\left((x^2-1)u'(x)\right) = s(s+1)u$, such that $u(1)=1$. $\endgroup$ – AD. Sep 20 '10 at 12:22
  • $\begingroup$ Well, you can take derivatives to arbitrary order, remember? ;) Also ${}_2 F_1$ admits an analytic continuation... anyway this is way off what you want; I'll look with interest at what the experts may come up with. $\endgroup$ – J. M. is a poor mathematician Sep 20 '10 at 13:25
  • $\begingroup$ Are you sure there is no mixup between Legendre functions and the Legendre symbol? $\endgroup$ – Accipitridae Sep 20 '10 at 19:17
  • $\begingroup$ @Accipitridae No it was a connection between Legendre functions and prime numbers.. (maybe it was associated Legendre function, that I don't remember..) $\endgroup$ – AD. Sep 20 '10 at 20:53

Perhaps this has something for you: http://wis.kuleuven.be/analyse/walter/coimbra.pdf

This uses the Legendre polynomials in proving irrationality of certain numbers.


If you accept enumerative combinatorics as part of number theory, here is a nice application of the Legendre polynomials.

Count the number of ways to insert paired, grammatical and non-repeating parentheses enclosing at least two letters in a word of length $n > 1$. Define $a_1 = 1$ and ignore outmost paratheses. For $n = 2$, there is only one way $(ab) = ab$. For $n = 3$, we have $abc, (ab)c, a(bc)$. For $n = 4$, we have $abcd$, $(ab)cd$, $a(bc)d$, $ab(cd)$, $(ab)(cd)$, $a(bcd)$, $a(b(cd))$, $a((bc)d)$, $(abc)d$, $(a(bc))d$, and $((ab)c)d$. And, so on.

The sequence continues {$1$, $1$, $3$, $11$, $45$, $197$, $903$, $4279 \dots$ } (A001003) and the terms are called Super Catalan numbers or Little Schroeder numbers. In general, \begin{eqnarray} a_n = \tfrac{1}{n} \sum_{k =0}^{n-2} \binom{2n-k-2}{n-1} \binom{n-2}{k} = \tfrac{1}{4n}(3P_{n-1}(3) - P_{n-2}(3)) + \tfrac{1}{2} \epsilon(n), \end{eqnarray} where $\epsilon$ is the unit arithmetic function, which is $1$ when the argument is $1$ and $0$ otherwise.

In fact, there are at least $10$ different combinatorial problems enumerated by this sequence, involving restricted lattice paths, integer compositions and even matrix row-sums. The Online Encyclopedia of Integer Sequences has the present state of knowledge about these fascinating numbers; see there reference section of http://oeis.org/A001003 for relevant papers and books.

  • $\begingroup$ In the formula, what is the "unit arithmetic function" epsilon(n)? $\endgroup$ – T.. Nov 15 '10 at 18:08

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.