Given the Rogers-Ramanujan continued fraction

$R(q)= \cfrac{1}{1+\cfrac{q}{1+\cfrac{q^2}{1+\cfrac{q^3}{1+\ddots}}}}$

where $q=\exp(2\pi i \tau)$, $|q|\lt1$

for the sake of brevity, let us introduce the following notation





up to $R_{n}$, $\frac{1}{R_{n}}=R(q^n,q)$ for natural number $n$, where $R(a,q)$ is the Generalized Rogers-Ramanujan continued fraction as pointed out by @ccorn

It is then conjectured that the following infinite series holds


It has the ascending continued fraction equivalent

$-\frac{R'(q)}{R(q)}= \frac{\frac{1}{R_1}+\large{\frac{\frac{-2q^2}{R_2}+\large{\frac{\frac{3q^5}{R_3}+...}{R^2_2}}}{R^2_1}}}{R_0} \tag2$

After considering the Rogers-Ramanujan continued fraction with the factor $q^{1/5}$ and applying the identity $\frac{R'(q)}{R(q)}=\frac{1}{5q}\frac{(q;q)^5_{\infty}}{(q^5;q^5)_{\infty}}$ due to Ramanujan,we are led to the following beautiful identity


$\frac{(q)^5_{\infty}}{(q^5)_{\infty}}=1+5\sum_{n=1}^{\infty}\frac{(-1)^n nq^{\frac{n(n+1)}{2}}}{\prod_{k=1}^{n-1}\Big(R_{k}+q^k\Big)}$

Question:How do we prove that the conjecture is true?

  • $\begingroup$ Can you have a close look at your definition for $R_2$ ... $R_{2}=\frac{1}{\Big(\frac{1}{q^2}\Big(\frac{1}{\Big(\frac{1}{q}\Big(\frac{1}{R(q)}-1\Big)\Big)}\color{red}{-1}\Big))}$ $\endgroup$ – Donald Splutterwit Jul 22 '17 at 20:28
  • $\begingroup$ If your goal is to calculate $R'(q)$, you might consider using the fact that $$R(q) = q^{1/5}\prod_{k=0}^{\infty} \frac{(1-q^{5k+1})(1-q^{5k+4})}{(1-q^{5k+2})(1-q^{5k+3})}$$ $\endgroup$ – Michael Lee Jul 22 '17 at 21:09
  • $\begingroup$ @MichaelLee How did you get that? If I may ask. $\endgroup$ – Simply Beautiful Art Jul 22 '17 at 21:32
  • 1
    $\begingroup$ L. J.. Rogers first and then S. Ramanujan later both proved it over a hundred years ago. See the Wikipedia article Rogers-Ramanujan identities. $\endgroup$ – Somos Jul 22 '17 at 22:19
  • $\begingroup$ Note that $\frac{1}{R_n} = R(q^n, q)$ where $R(a,q)$ is the generalized Rogers-Ramanujan continued fraction mentioned at Mathworld (37). It is often easier to prove statements about $R(a,q)$ because of the recurrence relation $R(a,q)=\frac{1}{1+aq\,R(aq,q)}$. $\endgroup$ – ccorn Jul 23 '17 at 7:55

A little study of $\,R_n\,$ turns up the result $\,R_n = 1 + q^{n+1} / R_{n+1}\,$ for $n\ge 0$. Taking a derivative gives $$R_n' = \frac{(n+1)q^n R_{n+1} - q^{n+1}R_{n+1}'}{R_{n+1}^2}$$ and then $$\frac{R_n'}{R_n} = \frac{ (n+1)q^n }{(R_nR_{n+1})} - \frac{q^{n+1}}{(R_nR_{n+1})}\frac{R_{n+1}'}{R_{n+1}}.$$

For brevity, define $\, P_n := \frac{1}{R_nR_{n+1}},\: Q_n := \frac{R_n'}{R_n} \,$ and rewrite our result as $$ Q_n = (n+1)q^nP_n - q^{n+1}P_nQ_{n+1}.$$ By iteration we get $$ Q_0 = 1q^0P_0 - q^1P_0Q_1 = 1q^1P_0 -q^1P_0(2q^1P_1 - q^2P_1Q_2) $$ and taking the limit $$ Q_0 = 1q^0P_0 - 2q^2P_0P_1 + 3q^5P_0P_1P_2 - 4q^9P_0P_1P_2P_3 + \dots$$ which is our result, but notice since $\, R(q) := 1/R_0, \,$ the left side is also $\, -R'(q)/R(q). \,$

By the way, if we define $\, F(x,q) = 1 + x/F(xq,q),\,$ then $\, R_n = F(q^{n+1},q). \,$

  • $\begingroup$ @Nicco Thanks, I like it myself for getting it. You may want to fix the sign error in your equation (1) since, as I mention, the left side should be $-R'(q)/R(q)$. $\endgroup$ – Somos Jul 23 '17 at 17:43

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.