Topological modular forms with level structure (Q5963142)

In number theory, there are many different versions of modular forms: with and without level structure, meromorphic or holomorphic at the cusp etc. Meromorphic modular forms can be seen as sections of certain line bundles on the moduli stack of (smooth) elliptic curves \(\mathcal{M}_{ell}\); holomorphic modular forms are sections of these line bundles on its compactification \(\overline{\mathcal{M}}_{ell}\), i.e. the moduli stack of generalized elliptic curves. Introducing level structures both substantially increases the number of examples of modular forms (e.g. allowing many more examples coming from \(\theta\)-functions) and has the advantage that in many cases the corresponding moduli stacks are actually moduli \textit{schemes}, both simplifying computations and theory. The idea of topological modular forms is to construct \(E_\infty\)-ring spectra that are topological versions of these rings of modular forms. To that purpose, Goerss, Hopkins and Miller have defined a sheaf \(\mathcal{O}^{top}\) of \(E_\infty\)-ring spectra on the étale site of \(\overline{\mathcal{M}}_{ell}\). We can then define periodic \(\mathrm{TMF}=\mathcal{O}^{top}(\mathcal{M}_{ell})\), ``compactified'' \(\mathrm{Tmf}=\mathcal{O}^{top}(\overline{\mathcal{M}}_{ell})\) and connective tmf as its connective cover \(\tau_{\geq 0}\mathrm{Tmf}\). This approach has its limits when it comes to defining topological modular forms with level structure: The corresponding maps of moduli stacks are only étale over \(\mathcal{M}_{ell}\) and not over \(\overline{\mathcal{M}}_{ell}\). Thus, using this approach we could only define periodic versions of topological modular forms with level structure, but not compactified and connective ones. The present paper solves this problem. The authors extend the sheaf \(\mathcal{O}^{top}\) to the \textit{log-étale} site of \(\overline{\mathcal{M}}_{ell}\). The precise notion of a log-étale map is a bit technical, but the rough idea is the following: An étale map is by definition unramified. Instead a log-étale map could, for example, allow ramification of order \(m\) on a specified smooth divisor if \(m\) is invertible. The archetypical example is \[ \mathrm{Spec}\, \mathbb{Z}[\frac1m][[q^{1/m}]] \to \mathrm{Spec}\, \mathbb{Z}[\frac1m][[q]] \] with specified divisor cut out by \(q\) and \(q^{1/m}\) respectively. Note that in the context of log-geometry this divisor (an example of a log-structure) has to be specified as data. In our situation, we choose the cusp \(\overline{\mathcal{M}}_{ell} - \mathcal{M}_{ell}\) as our divisor. If \(n\) is inverted, the compactified moduli stacks with level structure \(\overline{\mathcal{M}}(n)\), \(\overline{\mathcal{M}}_1(n)\) and \(\overline{\mathcal{M}}_0(n)\) are all log-étale over \(\overline{\mathcal{M}}_{ell}\) if we take the log-structure specified by the cusp. (Note here the sublety not mentioned in the article that there are two different definitions of \(\overline{\mathcal{M}}_0(n)\) in the literature. See Remark 4.15 of [\textit{B. Conrad}, J. Inst. Math. Jussieu 6, No. 2, 209--278 (2007; Zbl 1140.14018)].) This allows to define \(\mathrm{Tmf}(n) = \mathcal{O}^{top}(\overline{\mathcal{M}}(n))\) etc. Note though that defining connective versions from this is not always straightforward as in general the connective cover of \(\mathrm{Tmf}(n)\) or \(\mathrm{Tmf}_1(n)\) might not have the desired properties (i.e.\ its homotopy groups need not even rationally coincide with the corresponding ring of modular forms). The authors also define \(E_\infty\)-maps from these topological modular forms with level structures to forms of Tate K-theory (topological \(q\)-expansion) and forms of K-theory (evaluation at the cusp). The basic idea of the construction of the log-étale \(\mathcal{O}^{top}\) is the following: Over \(\mathcal{M}_{ell}\) it is already defined; indeed log-étale maps over \(\mathcal{M}_{ell}\) are already étale. Thus, one only has to define it at the completion at the cusp (and construct an equivalence on the ``overlap''). For example, the completion at the cusp for \(\overline{\mathcal{M}}_{ell}\) itself is the stack quotient of \(\mathrm{Spf}\, \mathbb{Z}[[q]] \) by \(C_2\) and one takes the real Tate K-theory \(KO[[q]]\) as the value of \(\mathcal{O}^{top}\) on it. In general, one gets over the cusp itself just a form of K-theory (as a non-smooth elliptic curve defines a form of the multiplicative formal grop) and one can then define a corresponding version of Tate K-theory by topologically mimicking the definition of a power series ring. The present paper is carefully written and provides also proofs of the folklore results it uses. It is a significant advance in the TMF-literature and has already been used in a number of preprints.