The homology of \(\mathrm{tmf}\) (Q505341)

The spectrum of topological modular forms comes in many variants, in particular there is the periodic version \(TMF\) and the connective version \(tmf\). While \(H^*(TMF; \mathbb{F}_p) = 0\), the cohomology of \(tmf\) is interesting and the present paper shows that \(H^*(tmf; \mathbb{F}_2) \cong \mathcal{A}/\mathcal{A}(2)\). Here \(\mathcal{A}\) denotes the mod-\(2\) Steenrod algebra and \(\mathcal{A}(i)\) is the subalgebra generated by \(Sq^1, \dots, Sq^{2^i}\). The result was already announced by Hopkins and Mahowald in the late nineties, but they never published the details. Interestingly, Davis and Mahowald had shown in 1982 that no spectrum with cohomology \(\mathcal{A}/\mathcal{A}(2)\) exists, but their proof had a mistake. The cohomology of \(tmf\) can be seen as analogous to that of real connective K-theory \(ko\), whose \(\mathbb{F}_2\)-cohomology is \(\mathcal{A}/\mathcal{A}(1)\). While this can be shown from the known cohomology of \(BO\), an analogous approach for \(tmf\) fails because no concrete geometric model for \(tmf\) is known. Instead, the argument proceeds by first computing the cohomology of \(tmf_1(3)\), which is a quotient of \(BP\), and then using that \(tmf_1(3) \simeq tmf \wedge DA(1)\) for a certain \(8\)-cell complex \(DA(1)\). Using similar techniques, the paper under review computes also \(MU_*tmf\). The result implies that the stack associated to the graded Hopf algebroid \((MU_*tmf, (MU\wedge MU)_*tmf)\) is the moduli stack of cubic curves. This implies that the \(E^2\)-term of the Adams-Novikov spectral sequence for \(tmf\) can be computed as the cohomology of the so-called Weierstrass Hopf algebroid, which was already used in [\textit{T. Bauer}, Geometry and Topology Monographs 13, 11--40 (2008; Zbl 1147.55005)].