2
$\begingroup$

So I've been trying to work out if it is possible to write:

$\large \Pi_{i=1}^n (3i-1)$ as an expression involving the quotient or product of two factorials, or really any expression involving factorials that isn't something like $\large (\Pi_{i=1}^n (3i-1))\frac{n!}{n!}$.

I actually started with another expression, namely $\large \Pi_{i=1}^n (2i-1)$

I managed to work with this expression by noticing we can consider $(2i-1)!$ and then divide by $(2^{n-1})(n-1)!$ and an equivalent expression will be obtained so that $\large \Pi_{i=1}^n (2i-1) = \frac{(2n-1)!}{2^{n-1}(n-1)!}$

When attempting to reason similarly with $\large \Pi_{i=1}^n (3i-1)$ I did not seem to get anywhere, which made me beg the question, is it never possible to do this for expressions of the form $\large \Pi_{i=1}^n (ki-1)$ where $k$ is an integer larger than $2$ ?

$\endgroup$
  • $\begingroup$ One clarification about your claimed identity. Take $n=6$. Mathematica gives $\prod_{i=1}^n (2i)-1=46079$, while $\frac{(2n-1)!}{2(n-1)!}=166320$... $\endgroup$ – Pierpaolo Vivo Jan 9 '16 at 11:11
  • $\begingroup$ I think he means it to be the product of $(2i-1)$ $\endgroup$ – πr8 Jan 9 '16 at 11:12
  • $\begingroup$ It doesn't work either, does it? $\prod_{i=1}^n (2i-1)=10395$, for $n=6$... $\endgroup$ – Pierpaolo Vivo Jan 9 '16 at 11:15
  • $\begingroup$ You are absolutely correct. I will make an edit to my post to amend this. $\endgroup$ – Aneesh Jan 9 '16 at 11:30
3
$\begingroup$

Short answer is no, there's nothing quite so concise involving factorials of integers. There are some other tools which allow you to write these expressions in shorthand - for example, your product would be $3^n(2/3)_n$ in the Pochhammer notation.

There are some combinations of the type of product that you mention which can be written in the factorial form. Consider:

$$\prod_{i=1}^n(3i-1)(3i-2)=\frac{(3n)!}{3^nn!}$$

$$\prod_{i=1}^n(4i-1)(4i-3)=\frac{(4n)!}{2^{2n}(2n)!}$$

$$\prod_{i=1}^n(5i-1)(5i-2)(5i-3)(5i-4)=\frac{(5n)!}{5^nn!}$$

$$\prod_{i=1}^n(6i-1)(6i-5)=\frac{(6n)!(n)!}{2^{2n}3^n(2n)!(3n)!}$$

and how you could "fill in the gaps" to get an expression involving pure factorials. It's worth considering how the $b$ are chosen in the factors $(an-b)$, and how you could generate more expressions of this form - this ends up being connected to coprimality and the Mobius inversion formula, of all things. Fun to play with.

For reference: if we define

$$U_n(x):=\prod_{0<a<n, (a,n)=1}(nx-a)$$

then

$$\prod_{x=1}^kU_n(x)=\prod_{d\vert n}[(dk)!(\frac{n}{d})^{dk}]^{\mu(\frac{n}{d})}$$

Furthermore, if you have any interest in the asymptotics of your product, we can compare it to $3^nn!$ quite easily, viz:

$\prod_{i=1}^n(3i-1)=\prod_{i=1}^n(\frac{3i-1}{3i}\times(3i))=(3^nn!)\prod_{i=1}^n(1-\frac{1}{3i})$

If you write the terms of the remaining product as $(1-\frac{1}{3i})=[(1-\frac{1}{3i})e^{1/{3i}}]e^{-1/3i}$, we see that the product is equal to:

$$(3^nn!)e^{-\frac{1}{3}(1/1+1/2+...+1/n)}\prod_{i=1}^n(1-\frac{1}{3i})e^{\frac{1}{3i}}$$

We now note the following:

  • The term in the exponential outside the product is the $n^{th}$ Harmonic number $H_n$, which satisfies $H_n=\log(n)+\gamma+o(1)$
    • The term inside the product is $(1-\frac{1}{18i^2}+o(\frac{1}{i^2}))$, which means that the product will converge to a nonzero constant (compare with convergence of $\sum\frac{1}{i^2}$ to convince yourself of this). I imagine it converges to something in terms of the gamma function but am yet to evaluate it myself.
    • The leading factorial can be asymptotically approximated by Stirling's approximation, $n!=\sqrt{2\pi}n^{n+1/2}e^{-n}(1+o(1))$

So, the product can be asymptotically approximated as:

$$(3^n)(\sqrt{2\pi}n^{n+1/2}e^{-n}(1+o(1)))(e^{-\frac{1}{3}(\log(n)+\gamma+o(1))})(constant)$$

$$=C(\frac{3}{e})^nn^{n+\frac{1}{6}}(1+o(1))$$

where $C$ is a constant.

$\endgroup$
  • $\begingroup$ Great answer, very detailed! $\endgroup$ – Aneesh Jan 9 '16 at 11:35
  • $\begingroup$ You inspired me ! Thanks for the fun. $\endgroup$ – Claude Leibovici Jan 9 '16 at 15:22
1
$\begingroup$

This is not an answer but it is too long for a comment.

Inspired by πr8's answer, I looked at the more general problem of $$P_n=\prod_{i=1}^n(a\,i+b)=a^n \left(\frac{a+b}{a}\right)_n$$ using Pochhammer notation. Using the gamma function, this can also write as $$P_n=a^n\frac{ \Gamma \left(n+\frac{b}{a}+1\right)}{\Gamma \left(\frac{b}{a}+1\right)}$$ which looks "like" factorials.

Taking logarithms and using Stirling approximation for the gamma function, the asymptotics write $$\log(P_n)=n \big(\log (n)+\log (a)-1\big)+\left(\frac{b}{a}+\frac{1}{2}\right) \log (n)+\log \left(\frac{\sqrt{2 \pi }}{\Gamma \left(\frac{a+b}{a}\right)}\right)+$$ $$\left(\frac{a^2+6 a b+6 b^2}{12 a^2 }\right)\frac 1n+O\left(\frac{1}{n^{3/2}}\right)$$ which, at least to me, looks quite nice.

$\endgroup$

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.