group cohomology, nonabelian group cohomology, Lie group cohomology
cohomology with constant coefficients / with a local system of coefficients
differential cohomology
The generalized (Eilenberg-Steenrod) cohomology theory/spectrum called $tmf$ – for topological modular forms – is in a precise sense the union of all elliptic cohomology theories/elliptic spectra (Hopkins 94).
More precisely, $tmf$ is the homotopy limit in E-∞ rings of the elliptic spectra of all elliptic cohomology theories, parameterized over the moduli stack of elliptic curves $\mathcal{M}_{ell}$. That such a parameterization exists, coherently, in the first place is due to the Goerss-Hopkins-Miller theorem. In the language of derived algebraic geometry this refines the commutative ring-valued structure sheaf $\mathcal{O}$ of the moduli stack of elliptic curves to an E-∞ ring-valued sheaf $\mathcal{O}^{top}$, making $(\mathcal{M}_{ell}, \mathcal{O}^{top})$ a spectral Deligne-Mumford stack, and $tmf$ is the E-∞ ring of global sections of that structure sheaf (Lurie).
The construction of $tmf$ has motivation from physics (string theory) and from chromatic homotopy theory:
from string theory. Associating to a space, roughly, the partition function of the spinning string/superstring sigma-model with that space as target spacetime defines a genus known as the Witten genus, with coefficients in ordinary modular forms. Now, the interesting genera typically appear as the values on homotopy groups (the decategorification) of orientations of multiplicative cohomology theories; for instance the A-hat genus, which is the partition function of the spinning particle/superparticle is a shadow of the Atiyah-Bott-Shapiro Spin structure-orientation of the KO spectrum. Therefore an obvious question is which spectrum lifts this classical statement from point particles to strings. The spectrum $tmf$ solves this: there is a String structure orientation of tmf such that on homotopy groups it reduces to the Witten genus of the superstring (Ando-Hopkins-Rezk 10).
Mathematically this means for instance that $tmf$-cohomology classes help to detect elements in the string cobordism ring. Physically it means that the small aspect of string theory which is captured by the Witten genus is realized more deeply as part of fundamental mathematics (chromatic stable homotopy theory, see the next point) and specifically of elliptic cohomology. Since the full mathematical structure of string theory is still under investigation, this might point the way:
A properly developed theory of elliptic cohomology is likely to shed some light on what string theory really means. (Witten 87, very last sentence)
from chromatic homotopy theory. The symmetric monoidal stable (∞,1)-category of spectra (finite spectra) has its prime spectrum parameterized by prime numbers $p$ and Morava K-theory spectra $K(n)$ at these primes, for natural numbers $n$. The level $n$ here is called the chromatic level. In some sense the part of this prime spectrum at chromatic level 0 is ordinary cohomology and that at level 1 is topological K-theory. Therefore an obvious question is what the part at level 2 would be, and in some sense the answer is $tmf$. (This point of view has been particularly amplified in the review (Mazel-Gee 13) of the writeup of the construction in (Behrens 13), which in turn is based on unpublished results based on (Hopkins 02)). For purposes of stable homotopy theory this means for instance that $tmf$ provides new tools for computing more homotopy groups of spheres via an Adams-Novikov spectral sequence.
Write
$\mathcal{M}_{cub}$ for the moduli stack of curves for cubic curves;
$\mathcal{M}_{ell}$ for the moduli stack of elliptic curves;
$\mathcal{M}_{\overline{ell}}$ for its Deligne-Mumford compactification obtained by adding the nodal cubic curve.
(Here $\mathcal{M}_{cub}$ is obatined by furthermor adding also the cuspidal cubic curve, hence we have canonical maps $\mathcal{M}_{ell}\to \mathcal{M}_{\overline{ell}}\to \mathcal{M}_{cusp} \to \mathcal{M}_{FG}$).
The Goerss-Hopkins-Miller theorem equips these three moduli stacks with E-∞ ring-valued structure sheaves $\mathcal{O}^{top}$ (and by Lurie (Survey) that makes them into spectral Deligne-Mumford stacks which are moduli spaces for derived elliptic curves etc.)
The $tmf$-spectrum is defined to be the $E_\infty$-ring of global sections of $\mathcal{O}^{op}$ (in the sense of derived algebraic geometry, hence the homotopy limit of $\mathcal{O}^{top}$ over the etale site of $\mathcal{M}$). More precisely one sets
$TMF \coloneqq \Gamma(\mathcal{M}_{ell}, \mathcal{O}^{top})$;
$Tmf \coloneqq \Gamma(\mathcal{M}_{\overline{ell}}, \mathcal{O}^{top})$;
$tmf \coloneqq$ the connective cover? of $Tmf$ (also $\simeq \Gamma(\mathcal{M}_{\overline{cub}}, \mathcal{O}^{top})$ (Hill-Lawson 13, p. 2 (?)).
We survey here some aspects of the explicit construction in (Behrens 13), a review is also in (Mazel-Gee 13),
The basic strategy here is to use arithmetic squares in order to decompose the problem into smaller more manageable pieces.
Write $\overline{\mathcal{M}_{ell}}$ for the compactified moduli stack of elliptic curves. In there one finds the pieces
given by rationalization
(hence this is the moduli of elliptic curves over the rational numbers) and by p-completion
for any prime number $p$, where $\mathbb{Z}_p$ denotes the p-adic integers and $Spf(-)$ the formal spectrum. (Hence this is the moduli of elliptic curves over p-adic integers).
This induces the arithmetic square decomposition which realizes $\mathcal{O}^{top}$ as the homotopy fiber product in
Here $\mathcal{O}^{top}_{\mathbb{Q}}$ can be obtained directly, and to obtain $\mathcal{O}^{top}_p$ one uses in turn another fracture square, now decomposing via K(n)-localization into $K(1)$-local and $K(2)$-local pieces.
(…)
There is a way to “construct” the tmf-spectrum as the E-∞ ring of global sections of a structured (∞,1)-topos whose underlying space is essentially the moduli stack of elliptic curves. We sketch some main ideas of this construction.
The discussion happens in the context of derived geometry in the (∞,1)-topos $\mathbf{H}$ over a small version of the (∞,1)-site of formal duals of E-∞ rings (ring spectra). This is equipped with some subcanonical coverage. For $R \in E_\infty Ring$ we write $Spec R$ for its image under the (∞,1)-Yoneda embedding $(E_\infty Ring)^{op} \hookrightarrow \mathbf{H}$.
The terminal object in $\mathbf{H}$ is the formal dual of the sphere spectrum
Because the sphere spectrum is the initial object in $E_\infty Ring$.
The crucial input for the entire construction is the following statement.
The formal dual of the complex cobordism Thom spectrum $M U$ is a well-supported object in $\mathbf{H}$, in that the morphism
to the terminal object in $\mathbf{H}$ is an effective epimorphism.
This means that $Spec M U$ plays the role of a cover of the point. This allows to do some computations with ring spectra locally on the cover $Spec M U$ . Since $M U^*$ is the Lazard ring, this explains why formal group laws show up all over the place.
To see this, first notice that the problem of realizing $R = tmf$ or any other ring spectrum as the ring of global sections on something has a tautological solution : almost by definition (see generalized scheme) there is an $E_\infty$-ring valued structure sheaf $\mathcal{O}Spec(R)$ on $Spec R$ and its global sections is $R$. So we have in particular
In order to get a less tautological and more insightful characterization, the strategy is now to pass on the right to the $Spec M U$-cover by forming the (∞,1)-pullback
The resulting Cech nerve is a groupoid object in an (∞,1)-category given by
which by formal duality is
where the smash product $\wedge$ of ring spectra over the sphere spectrum $\mathbb{S}$ is the tensor product operation on function algebras formally dual to forming products of spaces.
As a groupoid object this is still equivalent to just $Spec(tmf)$.
To simplify this we take a drastic step and apply a lot of decategorification: by applying the homotopy group (∞,1)-functor to all the $E_\infty$-rings involved these are sent to graded ordinary rings $\pi_*(tmf)$, $\pi_*(M U)$ etc. The result is an ordinary simplicial scheme
which remembers the fact that its structure rings are graded by being equipped with an action of the multiplicative group $\mathbb{G} = \mathbb{A}^\times$ (see line object).
This general Ansatz is discussed in (Hopkins).
This simplicial scheme, which is degreewise the formal dual of a graded ring of generalized homology-groups one can show is in fact a groupoid, hence a stack: effectively the moduli stack of elliptic curves. $\mathcal{M}_{ell}$. See (Henriques).
In fact if in this construction one replaced $Spec tmf$ by the point, one obtains the simplicial scheme
which one finds is the moduli stack of formal group laws $\mathcal{M}_{fg}$.
Now, a priori these underived stacks remember little about the original derived schemes $Spec tmf$ etc. They may not even carry any $E_\infty$-ring valued structure sheaf anymore (though some of them do).
If they do carry an $E_\infty$-ring valued structure sheaf $\mathcal{O}$, one can compute the homotopy groups of its global sections by a spectral sequence
But it turns out that even if the derived structure sheaf does not exist, this spectral sequence may still converge and may still compute the homotopy groups of the ring spectrum that one started with. This gives one way to compute the homotopy groups of $tmf$.
For the case of $tmf$ one finds that the homotopy sheaves $\pi_q(\mathcal{O}(\mathcal{M}_{ell}))$ are simple: they vanish in odd degree and are tensor powers $\omega^{\otimes k}$ of the canonical line bundle $\omega$ in even degree $2 k$, where the fiber of $\omega$ over an elliptic curve is the tangent space of that curve at its identity element. A section of $\omega^{\otimes k}$ is a modular form of weight $k$. So the whole problem of computing the homotopy groups of $tmf$ boils down to computing the abelian sheaf cohomology of the moduli stack of elliptic curves with coefficients in these abelian groups of modular forms — and then examining the resulting spectral sequence.
This can be done quite explicitly in terms of a long but fairly elementary computation in ordinary algebra. A detailed discussion of this computation is in (Henriques)
Write $B^2 U(1) \simeq K(\mathbb{Z},3)$ for the abelian ∞-group whose underlying homotopy type is the classifying space for circle 2-bundle. Write $\mathbb{S}[B^2 U(1)]$ for its ∞-group ∞-ring.
See (Ando-Blumberg-Gepner 10, section 8).
This means that every circle 2-bundle ($U(1)$-bundle gerbe) given by a modulating map $\chi \colon X \to B^2 U(1)$ determines a class represented by
in the $tmf$-generalized cohomology of its base space $X$.
The inclusion of the compactification point (representing the nodal curve but being itself the cusp of $\mathcal{M}_{\overline{ell}}$) into the compactified moduli stack of elliptic curves $\mathcal{M}_{\overline{ell}}$ is equivalently the inclusion of the moduli stack of 1-dimensional tori $\mathcal{M}_{1dtori} = \mathcal{M}_{\mathbb{G}_m}$ (Lawson-Naumann 12, Appendix A)
and pullback of global sections of Goerss-Hopkins-Miller-Lurie theorem-wise $E_\infty$-ring valued structure sheaves yields maps
exhibiting KO $= \Gamma(\mathcal{M}_{\mathbb{G}_m}, \mathcal{O}^{top})$.
At least after 2-localization the canonical double cover of the compactification of $\mathcal{M}_{\mathbb{G}_m} \simeq \mathbf{B}\mathbb{Z}_2$ similarly yields under $\Gamma(-,\mathcal{O}^{top})$ the inclusion of $ko$ as the $\mathbb{Z}_2$-homotopy fixed points of $ku$ (see at KR-theory for more on this)
and combined with the above this comes with maps from $tmf$ by restriction along the inclusion of the nodal curve cusp as
(Lawson-Naumann 12, theorem 1.2), where $tmf_1(3)$ denotes topological modular forms with level-3 structure (Mahowald-Rezk 09).
Moreover, including not just the nodal curve cusp but its formal neighbourhood which is the Tate curve, there is analogously a canonical map of $E_\infty$-rings
to Tate K-theory (this is originally asserted in Ando-Hopkins-Strickland 01, details are in Hill-Lawson 13, appendix A).
The $tmf$-spectrum is the codomain of the Witten genus, or rather of its refinements to the string orientation of tmf with value in topological modular forms
The original Witten genus is the value of the composite of this with the map to Tate K-theory on homotopy groups. (Ando-Hopkins-Rezk 10)
chromatic level | complex oriented cohomology theory | E-∞ ring/A-∞ ring | real oriented cohomology theory |
---|---|---|---|
0 | ordinary cohomology | Eilenberg-MacLane spectrum $H \mathbb{Z}$ | HZR-theory |
0th Morava K-theory | $K(0)$ | ||
1 | complex K-theory | complex K-theory spectrum $KU$ | KR-theory |
first Morava K-theory | $K(1)$ | ||
first Morava E-theory | $E(1)$ | ||
2 | elliptic cohomology | $Ell_E$ | |
second Morava K-theory | $K(2)$ | ||
second Morava E-theory | $E(2)$ | ||
algebraic K-theory of KU | $K(KU)$ | ||
$n$ | $n$th Morava K-theory | $K(n)$ | |
$n$th Morava E-theory | $E(n)$ | BPR-theory | |
$n+1$ | algebraic K-theory applied to chrom. level $n$ | $K(E_n)$ (red-shift conjecture) | |
$\infty$ | complex cobordism cohomology | MU | MR-theory |
The spectrum $tmf$ is self-dual under Anderson duality, more precisley $tmf[1/2]$ is Anderson-dual to $\Sigma^{21} tmf[1/2]$ (Stojanoska 11, theorem 13.1)
See at modular equivariant elliptic cohomology and at Tmf(n).
Substructure of the moduli stack of curves and the (equivariant) cohomology theory associated with it via the Goerss-Hopkins-Miller-Lurie theorem:
covering moduli space | of level-n structures (modular curve) | ||||||||
$\ast = Spec(\mathbb{Z})$ | $\to$ | $Spec(\mathbb{Z}[ [q] ])$ | $\to$ | $\mathcal{M}_{\overline{ell}}[n]$ | |||||
structure group | $\downarrow^{\mathbb{Z}/2\mathbb{Z}}$ | $\downarrow^{\mathbb{Z}/2\mathbb{Z}}$ | $\downarrow^{SL_2(\mathbb{Z}/n\mathbb{Z})}$ (modular group) | ||||||
$\mathcal{M}_{1dTori}$ | $\hookrightarrow$ | $\mathcal{M}_{Tate}$ | $\hookrightarrow$ | $\mathcal{M}_{\overline{ell}}$ | $\hookrightarrow$ | $\mathcal{M}_{cub}$ | $\to$ | $\mathcal{M}_{FG}$ | |
moduli stack | of 1d tori | of Tate curves | of elliptic curves | of cubic curves | of formal groups | ||||
$\mathcal{O}^{top}_{\Sigma}$ | KU | $KU[ [q] ]$ | elliptic spectrum | complex oriented cohomology theory | |||||
$\Gamma(-, \mathcal{O}^{top}) =$ | (KO $\hookrightarrow$ KU) = KR-theory | Tate K-theory ($KO[ [q] ] \hookrightarrow KU[ [q] ]$) | (Tmf $\to$ Tmf(n)) (modular equivariant elliptic cohomology) | tmf | $\mathbb{S}$ |
The idea of a generalized cohomology theory with coefficients the ring of topological modular forms providing a home for the refined Witten genus of
and produced as a homotopy limit of elliptic cohomology theories over the moduli stack of elliptic curves was originally announced, as joint work with Mark Mahowald and Haynes Miller, in
(There the spectrum was still called ”$eo_2$” instead of ”$tmf$”.) The details of the definition then appeared in
A central tool that goes into the construction is the Goerss-Hopkins-Miller theorem, see there for references on that.
Expositions include
Aaron Mazel-Gee, You could’ve invented $tmf$, April 2013 (pdf slides, notes pdf)
See also
An actual detailed account of the construction of $tmf$ (via decomposition by arithmetic squares) is spelled out in
A complete account of the computation of the homotopy groups of $tmf$ (following previous unpublished computations) is in
A survey of how this works is in
Akhil Mathew, The homotopy groups of $TMF$ (pdf)
(This presents as an instructive much simpler but analogous case the construction of KO in analogy to the construction of $tmf$, more details on this are in Mathew 13, section 3.)
and course notes that go through the construction of tmf and the computation of its homotopy groups are here:
Talbot workshop on TMF (web)
Mike Hopkins (talk notes by Michael Hill), Stacks and complex oriented cohomology theories (pdf)
André Henriques, The homotopy groups of tmf (pdf)
André Henriques, The moduli stack of elliptic curves (pdf)
The non-connective version of this is discussed in
The $\mathbb{Z}_2$-homology of $tmf$ is discussed in
The refinement of the Witten genus to a morphism of E-∞ rings to $tmf$, hence the string orientation of tmf is due to
Michael Hopkins, Topological modular forms, the Witten Genus, and the theorem of the cube, Proceedings of the International Congress of Mathematics, Zürich 1994 (pdf)
Matthew Ando, Michael Hopkins, Neil Strickland, Elliptic spectra, the Witten genus and the theorem of the cube, Invent. Math. 146 (2001) 595–687 MR1869850
Michael Hopkins, Algebraic topology and modular forms, Proceedings of the ICM, Beijing 2002, vol. 1, 283–309 (arXiv:math/0212397)
Matthew Ando, Michael Hopkins, Charles Rezk, Multiplicative orientations of KO-theory and the spectrum of topological modular forms, 2010 (pdf)
see also remark 1.4 of
and for more on the sigma-orientation see
Discussion of twisted cohomology with coefficients in $tmf$ is in section 8 of
Topological modular forms with level N-structure – $tmf(N)$ – is discussed in
Mark Mahowald Charles Rezk, Topological modular forms of level 3, Pure Appl. Math. Quar. 5 (2009) 853-872 (pdf)
Donald Davis, Mark Mahowald, Connective versions of $TMF(3)$ (arXiv:1005.3752)
Vesna Stojanoska, Duality for Topological Modular Forms (arXiv:1105.3968)
Tyler Lawson, Niko Naumann, Strictly commutative realizations of diagrams over the Steenrod algebra and topological modular forms at the prime 2, Int. Math. Res. Not. (2013) (arXiv:1203.1696)
Michael Hill, Tyler Lawson, Topological modular forms with level structure (arXiv:1312.7394)
The self-Anderson duality of $tmf$ is discussed in (Stojanoska 11).