Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I was trying to find a closed-form for $0<x<1$ in,

$$\frac{\,_2F_1(\frac{1}{m},\,1-\frac{1}{m},\,1,\,1-x)}{\,_2F_1(\frac{1}{m},\,1-\frac{1}{m},\,1,\,x)} = \sqrt{n}$$

where $\,_2F_1(a,b,c,z)$ is the hypergeometric function. There are formulas for $m = 2,3,4,6$, so I was wondering if there are for other m as well. However, one thing I observed was that, let,

$$q = \exp\left(\frac{-\,\pi\sqrt{n}}{\sin(\pi/m)}\right)$$

Conjecture:

$$\lim_{n\to \infty}\frac{x}{q} = \text{constant}$$

namely,

$$\begin{array}{cc} m&\lim_{n\to \infty}\frac{x}{q}\\\\ 2&16\\ 3&27\\ 4&64\\ 5&25\sqrt{5}\left(\frac{1+\sqrt{5}}{2}\right)^\sqrt{5}=163.95\dots\\ 6&432\\ 7&1152.795095384373\dots\\ 8&2^8\left(1+\sqrt{2}\right)^{2\sqrt{2}}=3096.65\dots\\ \end{array}$$

and so on. This implies a good approximation to x in,

$$\frac{\,_2F_1(\frac{1}{5},\frac{4}{5},\,1,\,1-x)}{\,_2F_1(\frac{1}{5},\frac{4}{5},\,1,\,x)} = \sqrt{n}$$

is given by,

$$x \approx 25\sqrt{5}\left(\frac{1+\sqrt{5}}{2}\right)^\sqrt{5} \exp\left(\frac{-\,\pi\sqrt{n}}{\sin(\pi/5)}\right)$$

(One can numerically solve for x for a given n using Mathematica's FindRoot command.)

Question: Is the conjecture true? And what is the closed-form for the constant $1152.79509\dots$ when $m=7$?

EDIT:

Courtesy of Sasha’s answer below, then the closed-form for m = 7, as radicals raised to radical powers is,

$$ (14)^2 \prod_{k=1}^{3} \frac{1}{\sin(\pi k/7)^{4\cos(2\pi k/7)}} = 1152.79509\dots$$

In general,

$$\lim_{n\to\infty}\frac{x}{q} = (2m)^2 \prod_{k=1}^{\lfloor (m-1)/2 \rfloor} \frac{1}{\sin(\pi k/m)^{4\cos(2\pi k/m)}} $$

share|improve this question
    
How did you obtain the formula for $m=5,8$ if closed form solutions exist only for $m=2,3,4,6$? –  Eckhard Dec 18 '12 at 18:12
    
For $m=5$, I used large enough $n$ and Plouffe's Inverse Symbolic Calculator (ISC) was able to recognize $y=\frac{x}{\sqrt{125}\,q} = 2.93298991\dots$. Further tests with larger $n$ gave more digits which agreed with ISC. (Similarly for $m=8$.) –  Tito Piezas III Dec 18 '12 at 18:43

1 Answer 1

up vote 4 down vote accepted

The hypergeometric function ${}_2F_1\left(\frac{1}{m}, 1-\frac{1}{m} ; 1; y\right)$ is an increasing from of $y$, starting at 1 for $y=0$ and increasing boundlessly as $y$ approaches one.

Thus, for large $n$ we should expect $x$ to be small. Taking the first to terms of the series expansion at unity: $$ {}_2F_1\left(\frac{1}{m}, 1-\frac{1}{m} ; 1; 1 - x \right) = -\frac{1}{\pi} \sin\left(\frac{\pi}{m} \right) \left( \log x + 2 \left( \gamma + \psi\left(\frac{1}{m} \right) \right) \right) - \cos\left(\frac{\pi}{m} \right) + \mathcal{O}(x) $$ which coincides with the expansion of the ratio of hypergeometric functions. $$ \frac{ {}_2F_1\left(\frac{1}{m}, 1-\frac{1}{m} ; 1; 1 - x \right) }{ {}_2F_1\left(\frac{1}{m}, 1-\frac{1}{m} ; 1; x \right)} = -\frac{1}{\pi} \sin\left(\frac{\pi}{m} \right) \left( \log x + 2 \left( \gamma + \psi\left(\frac{1}{m} \right) \right) \right) - \cos\left(\frac{\pi}{m} \right) + \mathcal{O}(x) $$ This gives a large $n$ approximation of the root: $$ x_n = \exp\left( -\frac{\pi}{\sin\left(\frac{\pi}{m}\right)} \left( \sqrt{n} + \cos\left(\frac{\pi}{m}\right) \right) - 2 \left( \gamma + \psi\left(\frac{1}{m}\right) \right) \right) $$ where $\gamma$ denotes the Euler-Mascheroni constant. From here, denoting $q_n = \exp\left( -\frac{\pi \sqrt{n} }{\sin\left(\frac{\pi}{m}\right)} \right)$ $$ \lim_{n \to \infty} \frac{x_n}{q_n} = \exp\left( -\frac{\pi}{\tan\left(\frac{\pi}{m}\right)} - 2 \left( \gamma + \psi\left(\frac{1}{m}\right) \right) \right) $$

The expression for $m=5$ agrees numerically with your expression, implying $$ \psi\left(\frac{1}{5}\right) = -\gamma - \frac{1}{2} \left( \pi \sqrt{1 + \frac{2}{\sqrt{5}}} + \ln\left( 25 \sqrt{5} \left(\frac{1+\sqrt{5}}{2} \right)^{\sqrt{5}} \right) \right) \tag{1} $$ I checked this identity using Mathematica, and found it to agree to 50,000 significant decimal points:

In[136]:= N[(-(Log[25*Sqrt[5]*((1 + Sqrt[5])/2)^Sqrt[5]] + 
               Sqrt[(1/5)*(5 + 2*Sqrt[5])]*Pi))*(1/2) - EulerGamma, 
  50000] - N[PolyGamma[1/5], 50000]

Out[136]= 0``49998.97559316131

I wonder if this identity is a known one. This identity is the Gauss's digamma theorem for $k=5$ and $m=1$ (thanks Edgar): $$ \psi\left(\frac{m}{k}\right) = -\gamma -\ln(2k) -\frac{\pi}{2}\cot\left(\frac{m\pi}{k}\right) +2\sum_{n=1}^{\lfloor (k-1)/2\rfloor} \cos\left(\frac{2\pi nm}{k} \right) \ln\left(\sin\left(\frac{n\pi}{k}\right)\right) $$

share|improve this answer
1  
Is it known? Evaluate $\psi(1/5)$ here: en.wikipedia.org/wiki/… –  GEdgar Jan 3 '13 at 16:09
    
Thanks, Sasha! It's good to finally know the closed-form of the limit, for any m, as radicals raised to radicals. –  Tito Piezas III Jan 3 '13 at 18:39

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.