11
$\begingroup$

It's a well-known formula that $$ e = \lim\limits_{n\rightarrow\infty} (1+1/n)^n $$ and I'm wondering if there exists a similar formula for $\pi$. All the formulas I know for $\pi$ involve series or products or special functions (such as inverse trig functions). Or one can use Stirling's approximation if factorials are allowed. I'm wondering if there exists a function $F:\mathbb{Z}^m\rightarrow \mathbb{R}$ such that $F$ can be written in terms of a finite number of additions, subtractions, multiplications, divisions, and exponentiations, and $$ \pi = \lim\limits_{n\rightarrow\infty} F(\underbrace{n,n,\cdots,n}_{r \text{ times}},k_1,k_2,k_3,\dots,k_{m-r}) $$ for some fixed integers $m$, $r$, and $k_1,k_2,\dots,k_{m-r}$. In particular, a formula that is defined by a series, infinite product, recursive formula, or continued fraction would not satisfy these conditions.

To be perfectly clear about what I mean, let me formalize. Define a set of elementary sequences $S\subset \mathbb{R}^\mathbb{Z}$ as follows: Let $S_0=\{n\}\cup\{\text{all constant, integer-valued sequences}\}$ and define $S_i$ recursively by $$ S_i = \left\{ s_n+t_n,\, s_nt_n,\, (s_n)^{t_n},\, s_n-t_n,\, \frac{s_n}{t_n}\, \mid\, (s_n),(t_n)\in S_{i-1} \right\}. $$ Then let $S = \bigcup\limits_{i=0}^\infty S_i$. Does there exist a sequence $(s_n)\in S$ such that $\lim\limits_{n\rightarrow\infty} s_n =\pi$?

Note that $(s_n)\in S$ implies $(s_n)\in S_i$ for some finite $i$, which means that each term in the sequence $s_n$ has at most $i$ operations in it. So for example, the sequence $(1+1/n)^n$ is in $S_3$ because it involves 3 operations (one addition, one division, and one exponentiation). A series, such as $s_n = \sum\limits_{k=1}^n \frac{(-1)^k 4}{2k-1}$ is not in $S$ because the number of operations in the $n$th term is unbounded.

$\endgroup$
15
  • 1
    $\begingroup$ It's not in terms only of the arithmetic operations (including exponentiation) $\endgroup$ Commented Feb 20 at 19:14
  • 1
    $\begingroup$ What about $F : \mathbb{Z} \rightarrow \mathbb{R}$ defined by $F(n)=\pi$ for every $n$ ? Then one has $$\pi = \lim_{n \rightarrow +\infty} F(n)$$ $\endgroup$ Commented Feb 20 at 19:15
  • 4
    $\begingroup$ "No, all of them either involve recursive formulas, infinite sums, or partial fractions, or some other infinite expressions." And what's wrong with that? The formula for $e$ involves a limit with index tending to infinity. Why is that acceptable but the others aren't? I think you need to be more specific just what conditions you are requiring. $\endgroup$
    – fleablood
    Commented Feb 20 at 19:17
  • 1
    $\begingroup$ Another way to define this: is there a term $t(n)$ in one free variable $n$, built out of the operations $+,\cdot,-,\div,\mathit{exp}$ and rational constants only, such that $\lim_{n\rightarrow\infty}t(n)=\pi$? $\endgroup$ Commented Feb 20 at 19:42
  • 6
    $\begingroup$ If one allowed the intermediate calculations to use complex numbers, then $$\pi=\lim_{n\to\infty} (-1)^{1/2} n (1 - (-1)^{1/n})$$ would work. This is a disguised version of $$-i \lim_{n\to\infty} \frac{(-1)^{1/n} - 1}{1/n};$$ the limit is the derivative of $(-1)^x$ evaluated at $x=0$, which is $\log(-1)=i\pi$. $\endgroup$ Commented Feb 20 at 20:05

1 Answer 1

4
$\begingroup$

Community wiki as it does not satisfy the constrains of the OP:

By this article if you consider the sequence $$ a_{n+1}=\left(1+\frac{1}{2n+1}\right) a_n $$ You have $$ \pi= \lim_n \frac{a_n}{n} $$ The sequence admits a closed form as $$ a_n= \frac{\sqrt{\pi} n!}{\Gamma\left(n+\frac{1}{2}\right)} $$ By this post you can rewrite it as $$ a_n=\frac{(n!)^22^{2n}}{(2n)!} $$ So you have $$ \pi= \lim_{n \to \infty} \frac{(n!)^22^{2n}}{n(2n)!} $$

$\endgroup$
3
  • 1
    $\begingroup$ The OP's constraints clearly do not allow factorial; that said, this is a neat fact. I think it should be a comment rather than an answer. $\endgroup$ Commented Feb 20 at 20:00
  • 2
    $\begingroup$ @NoahSchweber It is too long to be a comment I put it as community wiki $\endgroup$
    – Marco
    Commented Feb 20 at 20:02
  • 1
    $\begingroup$ @NoahSchweber Your criticism is right but I find the OP’s requirement weird if he allows taking $n$-th power in $(1+1/n)^n$ but disallows factorial. $\endgroup$
    – user1551
    Commented Feb 24 at 21:59

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .