(Inspired by this post.) Given the tribonacci constant $\Phi_3$, the tetranacci constant $\Phi_4$, etc. How do prove that,

$$\sum_{n=0}^\infty \binom{4n}{n}\frac{1}{(3n+1)\,2^{4n+1}}=\frac{1}{2}{\;}_3F_2 \left(\frac14,\frac24,\frac34; \color{blue}{\frac23,\frac43};\frac{4^4}{\color{red}{2^4}\,3^3} \right)=\frac1{\Phi_3}$$

$$\sum_{n=0}^\infty \binom{5n}{n}\frac{1}{(4n+1)\,2^{5n+1}}=\frac{1}{2}{\;}_4F_3 \left(\frac15,\frac25,\frac35,\frac45; \color{blue}{\frac24,\frac34,\frac54};\frac{5^5}{\color{red}{2^5}\,4^4} \right)=\frac1{\Phi_4}$$

$$\sum_{n=0}^\infty \binom{6n}{n}\frac{1}{(5n+1)\,2^{6n+1}}=\frac{1}{2}{\;}_5F_4 \left(\frac16,\frac26,\frac36,\frac46,\frac56; \color{blue}{\frac25,\frac35,\frac45,\frac65};\frac{6^6}{\color{red}{2^6}\,5^5} \right)=\frac1{\Phi_5}$$

and so on? And what is the corresponding hypergeometric formula for $\displaystyle \frac1{\Phi_2}$ with golden ratio $\Phi_2$?

(Courtesy of Jack D'Aurizio) By the Lagrange inversion theorem, a solution to, $$x^k-x+1=0$$ is given by, $$x=\sum_{n=0}^\infty \binom{kn}{n}\frac{1}{(k-1)n+1}$$

Similarly, given the general polynomial of the $k$-nacci constants as, $$x^k(2-x)-1=0$$ how do we show that, $$\frac1{x}=\frac1{\Phi_k}=\sum_{n=0}^\infty \binom{(k+1)n}{n}\frac{1}{(kn+1)\,2^{(k+1)n+1}}$$

  • $\begingroup$ The pattern using the binomial coefficients does yield $\displaystyle \frac1{\Phi_2}$, but the pattern from the generalized hypergeometric function does not. $\endgroup$ – Tito Piezas III Nov 23 '17 at 14:46
  • $\begingroup$ Isn't this just Lagrange inversion theorem? $\endgroup$ – Jack D'Aurizio Nov 23 '17 at 18:36
  • $\begingroup$ @JackD'Aurizio: Do you mean invert these series to give series for $\Phi_n$ that are easer to prove? $\endgroup$ – James Arathoon Nov 23 '17 at 22:16
  • $\begingroup$ @JackD'Aurizio: I've edited the post re your comment. $\endgroup$ – Tito Piezas III Nov 24 '17 at 3:30

What about this for golden ratio $\Phi_2$?

$$\sum_{n=0}^\infty \binom{3n}{n}\frac{1}{(2n+1)\,2^{3n+1}}=\frac{1}{2}{\;}_2F_1 \left(\frac13,\frac23; \color{blue}{\frac32};\frac{3^3}{\color{red}{2^3}\,2^2} \right)=\frac1{\Phi_2}$$

Further Observations added

The first constant in this sequence is $1$ which is its own inverse. $$\sum_{n=0}^\infty \binom{2n}{n}\frac{1}{(n+1)\,2^{2n+1}}=?=\frac1{\Phi_1}=1$$

If $\frac1{\Phi_1}$,$\frac1{\Phi_2}$,$\frac1{\Phi_3}$ and $\frac1{\Phi_4}$ are used as the roots of the successive polynomials instead of $\Phi_1$,$\Phi_2$,$\Phi_3$ and $\Phi_4$ then the polynomials are instead respectively

$$x-1=0$$ $$x^2+x-1=0$$ $$x^3+x^2+x-1=0$$ $$x^4+x^3+x^2+x-1=0$$

If you look at the relationship I found for finding roots Rational Series for Positive Integer Root of $x>1$ then it seems the number multiplying n to produce the binomial coefficients should be one less; that is for $\Phi_2$ it should be 2 instead of 3.

However if the above polynomials are each multiplied by $(x-1)$ then by rearrangement we can arrive at a remarkable sequence of self-similar formulae of the form


I am not sure this is the right way to proceed, but it is an interesting pattern nevertheless.

The Plot Thickens

Contrary to what I was saying above the sequence below seems to hold numerically as well

$$\sum_{n=0}^\infty \binom{rn}{n}\frac{1}{((r-1)n+1)\,2^{rn+1}(\Phi_r)^n}=\frac1{\Phi_r}$$

For the first formula in the sequence we have numerically

$$\sum_{n=0}^\infty \binom{2n}{n}\frac{1}{(n+1)\,2^{2n+1}x^n}=\frac{1}{1+\sqrt{\frac{x-1}{x}}}$$

  • $\begingroup$ Ah, I missed and used instead $$\frac{1}{2}{\;}_2F_1 \left(\frac13,\frac23; \color{blue}{\frac22};\frac{3^3}{\color{red}{2^3}\,2^2} \right)$$ $\endgroup$ – Tito Piezas III Nov 23 '17 at 15:25
  • $\begingroup$ As an aside the only formula I can find for the inverse of the silver ratio is $\sum_{n=0}^\infty \binom{2n}{n}\frac{(2n+1)}{(n+1)\,2^{3n+2}}$ $\endgroup$ – James Arathoon Nov 23 '17 at 21:51
  • $\begingroup$ I've edited the post. $\endgroup$ – Tito Piezas III Nov 24 '17 at 3:32
  • $\begingroup$ Let $C(x)$ be the Catalan generating function. Then the identity for $\Phi_1$ simply says that $(1/2)C(1/4)=1$, which is easy to check. $\endgroup$ – Alexander Burstein Nov 25 '17 at 2:22
  • $\begingroup$ @AlexanderBurstein: Updated my post - the identity for $\Phi_1$ also has an interesting connection to $\Phi_2$ $\endgroup$ – James Arathoon Nov 26 '17 at 2:10

Note that $\frac{1}{kn+1}\binom{(k+1)n}{n}$ is the number of $(k+1)$-ary trees on $n$ vertices. Let $T_{k+1}$ be the ordinary generating function for the class of $(k+1)$-ary trees. Then $T_{k+1}$ satisfies the equation $$ T_{k+1}(z)=1+zT_{k+1}^{k+1}(z). $$ Let $a=\frac{1}{2}T_{k+1}\left(\frac{1}{2^{k+1}}\right)$, then we need to prove that $\frac{1}{a}=\Phi_k$. From the functional equation, we have $$ 2a=1+\frac{1}{2^{k+1}}(2a)^{k+1}=1+a^{k+1}, $$ so $a$ satisfies the equation $$ 1-2x+x^{k+1}=0, $$ and thus $\frac{1}{a}$ satisfies $$ x^{k+1}-2x^k+1=(x-1)(x^k-x^{k-1}-\dots-x-1)=0. $$ We will leave the argument that $0<a<1$ for $k\ge 2$ to the reader. Then $\frac{1}{a}$ is a root of the equation $\phi_k(x)=x^k-x^{k-1}-\dots-x-1=0$. From Proposition 1 on p. 6 of this paper, we see that $\phi_k(x)$ has a unique root greater than $1$, so $a$ is uniquely determined by this polynomial. However, $\Phi_k$ is also the unique root greater than $1$ of $\phi_k(x)$, so $\frac{1}{a}=\Phi_k$.

Perhaps, it would be interesting to interpret both sides probabilistically. A few years ago, I think I have actually heard of that problem for $k=1$, where the sum is $1$, but I'm not sure it's still open or not.


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.