I want to Evaluate $1-5(\frac{1}{2})^3+9(\frac{(1)(3)}{(2)(4)})^3-13(\frac{(1)(3)(5)}{(2)(4)(6)})^3+...$ ,I tried from arcsin(x) series and got $\frac{1-z^4}{(1+z^4)^{\frac{2}{3}}}= 1-5(\frac{1}{2})z^4+9(\frac{(1)(3)}{(2)(4)})z^8-13(\frac{(1)(3)(5)}{(2)(4)(6)})z^{12}+...$ ,but i got stuck with $\frac{(2n-1)!!}{2n!!}$ terms and I don't know how to get $(\frac{(2n-1)!!}{2n!!})^3 $ Can anyone help me to Evaluate this infinite sum? You can share your own way.
-
In other words you want to evalutate $$\sum_{n\ge 0}(-1)^n\binom{2n}n^3\frac{4n+1}{64^n}$$ when I am not mistaken. – mrtaurho Feb 10 '19 at 11:28
-
Related: https://math.stackexchange.com/questions/338623/how-to-evaluate-a-sum-of-powers-of-central-binomial-coefficients – mrtaurho Feb 10 '19 at 11:49
-
Does it have other simple way that doesn't use Hypergeometric. – Unik Sillavich Feb 10 '19 at 12:26
-
This is a famous series given by Ramanujan and its value is $2/\pi$. See https://paramanands.blogspot.com/2012/03/modular-equations-and-approximations-to-pi-part-2.html?m=0 – Paramanand Singh Sep 14 '19 at 02:31
-
@ParamanandSingh I have written a answer for this question not based on Elliptic Integrals, I was hoping you could verify it once as I am not sure of it. Thank you. – Miracle Invoker Mar 19 '24 at 12:59
1 Answers
$$\frac{2}{\pi}=\sum_{n=0}^{\infty}(-1)^n\binom{2n}{n}^3\left(\frac{1+4n}{2^{6n}}\right)$$
One can prove this using Fourier-Legendre Expansions thereby completely bypassing the Elliptic Integral Pathway.
First denote,
$$P_s:=P_s(k)={_2F_1}\left[\begin{array}c-s,1+s\\1\end{array}\middle|\,k^2\right]$$
and $P'_s:=P_s(\sqrt{1-k^2})$
which represents the Alternate Elliptic Integrals for $$-s\in\left\{\frac{1}{2},\frac{1}{3},\frac{1}{4},\frac{1}{6}\right\}$$
But that won't be the focus here, instead we will be using this when $s$ is an integer.
Using Differential Equations, one may prove that
$$\int_0^1 kP_a'P_b\ dk=\frac{\sin \pi b-\sin \pi a}{2\pi[b(b+1)-a(a+1)]}$$
and we can also prove that if $n$ is an integer then $P_n'=(-1)^nP_n$
And we also have the evaluation of $P_s$ at $k=1/\sqrt{2}$
$$P_s\left(\frac{1}{\sqrt{2}}\right)=\frac{\cos(\pi s/2)}{2^{s}}\binom{s}{s/2}$$
This results in for $a$ and $b$ integer,
$$\int_0^1kP_aP_b\ dk=0,\quad a\neq b$$
$$\int_0^1kP^2_n\ dk=\frac{1}{2(2n+1)}$$
which is basically an orthogonality property.
So one may write for a function $f(k)$
$$f(k) \sim2\sum_{n=0}^{\infty}(2n+1)\left[\int_0^1kf(k)P_n\ dk\right]P_n $$
Those familiar with Fourier Legendre Expansions will realize it's the same thing.
Using Clausen's Formula we can prove that,
$$\int_0^{\pi/2}P_s(2kk'\sin t)\ dt=\frac{\pi P^2_s}{2}$$
or evaluating it at $k=1/\sqrt{2}$ gives
$$\int_0^{\pi/2}P_s(\sin t)\ dt=\frac{\pi}{2}\frac{\cos(\pi s/2)^2}{2^{2s}}\binom{s}{s/2}^2$$
but this integral can be rewritten as,
$$\int_0^1k\left[\frac{1}{k\sqrt{1-k^2}}\right]P_{2n}\ dk=\frac{\pi}{2}\frac{1}{2^{4n}}\binom{2n}{n}^2$$
which gives us the Fourier Legendre Expansion,
$$\frac{1}{k\sqrt{1-k^2}}\sim\pi\sum_{n=0}^{\infty}\binom{2n}{n}^2\left(\frac{1+4n}{2^{4n}}\right)P_{2n}$$
This doesn't converge so I am not sure of this last step but if we put $k=1/\sqrt{2}$ and use
$$P_{2n}\left(\frac{1}{\sqrt{2}}\right)=\frac{(-1)^n}{2^{2n}}\binom{2n}{n}$$
we will recover back the series for $1/\pi$.
I have come up with this proof myself no I am not sure of its validity.
I hope someone can verify.
EDIT
It seems that if the following series converges absolutely
$$\sum_{n=0}^{\infty}c_n$$
then the series of functions $$\sum_{n=0}^{\infty}c_nP_n$$ converges pointwise.
So in our case as $$\sum_{n=0}^{\infty}\binom{2n}{n}^2\left(\frac{1+4n}{2^{4n}}\right)$$
does not converge then the series of functions does not converge to $\frac{1}{k\sqrt{1-k^2}}$.
But it is still interesting so I will leave this answer here as reference for future readers.
- 3,200
- 1
- 3
- 22
-
1You need to check the conditions under which the Fourier Legendre expansion converges to the function $f(k) $ and see if it applies here in the last step. BTW, I haven't given any vote (up/down) so far. – Paramanand Singh Mar 20 '24 at 01:41
-
@ParamanandSingh Thank you for your input, it seems the expansion does not converge pointwise but I am not sure why it seems to work for $1/\sqrt{2}$, I will look into it. My last three answers have gotten downvoted too (I can understand this one, but the other two were perfectly fine) – Miracle Invoker Mar 20 '24 at 04:17
-
1If you think someone is targeting you via down votes you can raise a flag and we will look into it. – Paramanand Singh Mar 20 '24 at 05:35