In "On elliptic modular surfaces", Shioda proves some interesting theorems on smooth elliptic surfaces (admitting a section); he then focuses on "modular elliptic surfaces" and proves some more theorems, e.g. he shows that there is a one to one correspondence between modular forms vanishing at cusps of a modular curve, and holomorphic two forms on the corresponding modular surfaces.

After this short introduction, my question is: Can you give me an explicit example of a smooth elliptic surface (with section) with at least 3 singular fibers, which is not modular? What if the base curve has genus zero?

More explanation: Let $S\to B$ be an elliptic surface with section and at least 3 singular fibers. Let $\Sigma\subset B$ be the complement of singular values. Then $\Sigma=\mathbb{H}/G$, is a quotient of upper half plane. The $j$ function of this elliptic fibration $j:\Sigma \to \mathbb{H}/PSL(2,\mathbb{Z})$ (is it always holomorphic?) gives a map $$ \iota: G \to SL(2,\mathbb{Z}).$$ I then feel that $S$ should be the modular surface corresponding to $\iota(G)\subset SL(2,\mathbb{Z})$ and I can't see what might go wrong?!

Regarding the answer of Remke below: Assuming that the map $\iota$ above has finite dimensional kernel, it seems that $S$ should be the pull-back of the modular surface corresponding to $\iota(G)$ via the ramified covering map $$ \mathbb{H}/G \to \mathbb{H}/\iota(G);$$ is this true?


1 Answer 1


The modular elliptic surfaces are quite rare. E.g., the Mordell-Weil group is finite and the Picard number of the surface equals $h^{1,1}$ (see Shioda's paper). Such elliptic surfaces are called extremal and for a fixed Euler number of the surface and fixed genus of the base curve there are only finitely many extremal elliptic surfaces with non-constant $j$-map.

The $j$-map of an extremal elliptic surfaces is unramified away from the preimages of $0$, $1728$ and $\infty$. So any elliptic surface with $j$-map which has ramification over a point different from $0,1728$ and $\infty$ is not a modular elliptic surface.

  • $\begingroup$ I have added a comment to the end of my question regarding your answer. What do you think of it? $\endgroup$ Oct 17, 2013 at 23:39
  • $\begingroup$ To be honest, I do not really understand the second part of your question. For example, why is the map H/G to H/PSL(2,Z) induced by a self map of H? You need this to have the map i:G->PSL(2,Z). $\endgroup$ Oct 18, 2013 at 17:34
  • $\begingroup$ We can lift the j-map to $j:\mathbb{H}\to \mathbb{H}/PSL(2,\mathbb{Z})$; since $\mathbb{H}$ is simply connected, we can then lift this to a map $j:\mathbb{H}\to\mathbb{H}$. Then for $\gamma \in G$, we have to have $j(\gamma\cdot z)=\iota(\gamma)j(z)$, for some representation $\iota:G \to PSL(2,\mathbb{Z})$ (I am assuming this lifts to $SL(2,\mathbb{Z})$ itself, which should not be a big deal for now) $\endgroup$ Oct 18, 2013 at 19:49
  • $\begingroup$ @MohammadF.Tehrani How does the j-map lift to a self-map of $\mathbb H$? The action of PSL$_2(\mathbb Z)$ is not free, and moreover, the complex upper half-plane $\mathbb H$ is not the universal cover of the affine line PSL$_2(\mathbb Z) \backslash \mathbb H$. $\endgroup$ Mar 9, 2014 at 19:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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