I. Klein

In "On the Order-Seven Transformation of Elliptic Functions" (pp. 287-331), he discusses in p. 298 what we now call the Klein quartic,

$$\lambda^3\mu+\mu^3\nu+\nu^3\lambda= 0\tag1$$

and in p. 313 introduces what we can call the Klein septic resolvent,

$$z^7-2^2\cdot7^2\big(7\mp\sqrt{-7}\big)z^4+2^5\cdot 7^4\big(5\mp\sqrt{-7}\big)\color{red}z\\ \mp 2^9\cdot3\cdot7^3\sqrt{-7}\frac{g_2}{\sqrt[3]\Delta}=0\tag2$$

(where the red linear $\color{red}z$ is missing in the paper and I assume is a typo). In the same page, he says the roots of $(2)$ in terms of $\lambda,\mu, \nu$ are,

$$z =\frac{\pm 2\sqrt{-7}\Big(P_1+\frac{-1\mp\sqrt{-7}}{2}P_2\Big)}{\sqrt[3]\nabla}\tag3$$


$$P_1 = \gamma^{2x}\lambda^2+\gamma^{x}\mu^2+\gamma^{4x}\nu^2\\ P_2 =\gamma^{6x}\mu\nu+\gamma^{3x}\nu\lambda+\gamma^{5x}\lambda\mu$$

with $\gamma= e^{2\pi i/7}$ as first mentioned in p. 313.

II. Ramanujan

Unbeknownst to Klein (d. 1925), it turns out Ramanujan (d. 1920) found an elegant parameterization to $(1)$. Define the Ramanujan theta function,

$$f(a,b) = \sum_{n=-\infty}^\infty a^{n(n+1)/2}\, b^{n(n-1)/2}$$


$$\lambda = q^{1/56}f(-q^3,-q^4)\\ \mu= \color{red}{-}q^{25/56}f(-q,-q^6)\\ \nu= q^{9/56}f(-q^2,-q^5)\\$$

where $q=e^{2\pi i\tau}$.

III. Question

I tried to implement this in Mathematica. Unfortunately, I couldn't get $(3)$ to be a root of the septic $(2)$.

Q: Was I wrong in assuming that any parametrization to $(1)$ would do? Or is there some typo or confusion of variables in the paper that screwed up my implementation?

  • $\begingroup$ Isn't the upside-down Delta the Hessian of (20)? Also, $x$ ranges from 0 to 6, cf. (18). $\endgroup$ Mar 21 '19 at 22:43

There seems to be a problem with Klein's septic equation $(2)$ combined with the purported roots in $(3)$. Let $\,k\,$ be any integer. Define $$ P_1(k) := \gamma^{k}\mu^2 + \gamma^{2k}\lambda^2 + \gamma^{4k}\nu^2 \tag{1} $$ and $$ P_2(k) := \gamma^{3k}\lambda\nu + \gamma^{6k}\mu\nu + \gamma^{5k}\lambda\mu. \tag{2} $$ Let $$ \,a := 2\sqrt{-7}/\eta(\tau)^2 \; \textrm{ and } \; b := -(7+\sqrt{-7})/\eta(\tau)^2. \tag{3} $$ Define the roots of $(2)$ as $$ r_k := a\,P_1(k) + b\,P_2(k). \tag{4} $$ The polynomial $\, P(z) := (z - r_1)(z - r_2)\cdots (z - r_7)\,$ expands to $$ P(z) = z^7 \!-\! 2^2\! \cdot\! 7^2\, (7+\sqrt{-7})\, z^4 \!+\! 2^5\!\cdot \!7^4\, (5+\sqrt{-7})\,z \!+\! 2^7\! \cdot\! 7^3 \sqrt{-7} \frac{g_2(\tau)}{\eta(\tau)^8}. \tag{5}$$ This was for values of $\,\lambda,\mu,\nu\,$ as suggested by Tito Piezas III.

P.S. Note that $\, \sqrt[3]{\Delta} = \eta(\tau)^8$ is the denominator of the constant term of $P(z).$ Also note that $\, 1728 J(\tau) = j(\tau) = g_2(\tau)^3/\eta(q)^{24}\,$ where $\,J(\tau)\,$ is Klein's invariant. Thus using cube roots $\, 12\sqrt[3]{J(\tau)} = g_2(\tau)/\eta(\tau)^8\,$ and the constant term can be written as $\, 2^9\!\cdot 3 \cdot\! 7^3 \sqrt{-7} \sqrt[3]{J(\tau)} \,$ which is closer to Klein's version. In fact, just before Klein's equation $(49)$ (Tito's equation $(2)$) Klein writes $\ J = g_2^3/\Delta.$

P.P.S. I see that Tito has used a set of roots $\,y_k\,$ differing by a common factor from $\,z_k\,$ to simplify the septic. Perhaps Klein would have used that version, but he preferred $\,J(\tau)\,$ instead of $\,j(\tau).$

  • $\begingroup$ Beautiful! I verified it with my own revised Mathematica implementation. I knew there may have been a 2nd typo in the paper (after I found the 1st one). And to complicate things, I used the modern definition of the modular discriminant as $$\Delta' = (2\pi)^{12} \eta^{24}(\tau)$$ though Mathworld did point out that older literature might just use, $$\Delta = \eta^{24}(\tau)$$ which is the case in this post. $\endgroup$ Mar 22 '19 at 2:53

(This summarizes the accepted answer of Somos, and uses just the j-function and Dedekind eta function.)

Given the j-function $j(\tau)$,

$$j(\tau) = 1728J(\tau)$$

implemented in Mathematica as j(t) = 1728KleinInvariantJ[t], then Klein's septic resolvent reduces to the elegant formula,


where the seven roots are,

$$y_k = \frac{P_1(k)+\alpha P_2(k)}{\eta^2(\tau)}\\ \alpha=\frac{-1+\sqrt{-7}}2$$

for $k=1,2\dots7,$ with $P_1(k)$ and $P_2(k)$ as defined by Klein (and Somos).

  • 2
    $\begingroup$ Note that, $$\frac{5+\sqrt{-7}}2=-\left(\frac{1+\sqrt{-7}}2\right)^3$$ $\endgroup$ Mar 22 '19 at 3:57

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.