5

In this post, the problem was given integer/rational $N$, to solve for algebraic number $z$ in the equation, $$\begin{aligned}\frac{1}{N} &=I\left(z^2;\ a,b\right)\\[1.5mm] &= \frac{B\left(z^2;\ a,b\right)}{B\left(a,b\right)} \end{aligned} $$ using the beta function $B(a,b)$, incomplete beta $B(z;a,b)$ and regularized beta $I(z;a,b)$. It seems for some $a,b,$ there can be a pattern for the solutions.

Given the equation for $n>1$,

$$I\left(x_1^2;\ \tfrac14,\tfrac12\right)=\frac{1}{2^n}\tag1$$

First define $\color{blue}{\gamma = u+\sqrt{-1+u^2}},$ with fundamental unit $u=1+\sqrt{2}.\ $ Then,

For $n=2$: $$x_1=x_2-\sqrt{1+x_2^2},\quad \color{blue}{x_2 =\gamma}$$ For $n=3$: $$x_1=x_2-\sqrt{1+x_2^2},\quad \color{green}{x_2 = x_3+\sqrt{-1+x_3^2},\quad x_3=x_4+\sqrt{1+x_4^2},}\quad \color{blue}{x_4 = \gamma}$$ For $n=4$: $$x_1=x_2-\sqrt{1+x_2^2},\quad x_2 = x_3+\sqrt{-1+x_3^2},\quad x_3=x_4+\sqrt{1+x_4^2},\\ \quad \color{green}{x_4 = x_5+\sqrt{-1+x_5^2},\quad x_5=x_6+\sqrt{1+x_6^2},}\quad \color{blue}{x_6 = \gamma}$$

and so on, where we add the same two nested layers (in green) each time and so end with even index $x_m$.

Questions:

  1. Does this pattern really hold for all $2^n$ and $n>1?$ Why the regularity?
  2. What is the integral associated with $(1)$ similar to the ones in the post cited above?

1 Answers1

2

We have $$ B\left(z;\ \tfrac14,\tfrac12\right)=4\sqrt[4]{z}\, _2F_1\left(\frac{1}{2},\frac{1}{4};\frac{5}{4};z\right). $$ By formula 2.1.15 from Erdelyi, "Higher transcendental functions", vol.I $$ _2F_1\left(\frac{1}{2},\frac{1}{4};\frac{5}{4};z\right)=\sqrt{\frac{1}{1-z}} \, _2F_1\left(\frac{1}{2},\frac{1}{4};\frac{5}{4};-\frac{4 z}{(1-z)^2}\right). $$ Since $z$ and $-\frac{4 z}{(1-z)^2}$ have different signs when $z$ is real we need to apply this formula one more time $$ \, _2F_1\left(\frac{1}{2},\frac{1}{4};\frac{5}{4};z\right)=\frac{1}{2}\sqrt[4]{\frac{16 (1-z)^2}{(z+1)^4}} \, _2F_1\left(\frac{1}{2},\frac{1}{4};\frac{5}{4};\frac{16 z (1-z)^2}{(z+1)^4}\right). $$ In terms of incomplete Beta function one gets $$ B\left(z;\ \tfrac14,\tfrac12\right)=\frac12 B\left(\tfrac{16 z (1-z)^2}{(z+1)^4};\ \tfrac14,\tfrac12\right). $$ I think this formula answers the question 1. From this formula one can work out the recursion for the argument.

  • +1 Ah, clever. Some experimentation shows there are corresponding relations for $\frac{1}{2}I\big(\alpha;\frac{1}{2},\frac{1}{3}\big)=I\big(\beta;\frac{1}{2},\frac{1}{3}\big)$ and $\frac{1}{2}I\big(\alpha;\frac{1}{2},\frac{1}{4}\big)=I\big(\beta;\frac{1}{2},\frac{1}{4}\big)$ which should yield, for example $I\big(\beta;\frac{1}{2},\frac{1}{3}\big)=\frac{1}{6}$ and infinitely more as discussed in this post. – Tito Piezas III Dec 03 '16 at 08:31
  • A favor. Can you include the integral for question 2? I'm having trouble getting its explicit form. – Tito Piezas III Dec 04 '16 at 01:12
  • @TitoPiezasIII Are you sure it is $\frac{1}{2}I\big(\alpha;\frac{1}{2},\frac{1}{3}\big)=I\big(\beta;\frac{1}{2},\frac{1}{3}\big)$ with prefactor $\frac12$, not $\frac13$? – Martin Nicholson Dec 04 '16 at 06:23
  • Yes, its a duplication formula for the other integral. Kindly see this new answer. I used it to find the missing $I\big(z^2;\frac{1}{2},\frac{1}{3}\big)=\frac{1}{6}$ in my list. – Tito Piezas III Dec 04 '16 at 06:57
  • I just saw your edit. Er, what I was actually looking for is the integral associated with $I\left(z^2;\ \tfrac14,\tfrac12\right)$. It's missing in this list as it should be the $4$th. And I've moved your triplication formula here :) – Tito Piezas III Dec 04 '16 at 15:34
  • These are identities (12) and (13) from Vidunas' paper https://arxiv.org/abs/0811.4641 – Martin Nicholson Jan 22 '18 at 13:46