Skip to the Main Content

Note:These pages make extensive use of the latest XHTML and CSS Standards. They ought to look great in any standards-compliant modern browser. Unfortunately, they will probably look horrible in older browsers, like Netscape 4.x and IE 4.x. Moreover, many posts use MathML, which is, currently only supported in Mozilla. My best suggestion (and you will thank me when surfing an ever-increasing number of sites on the web which have been crafted to use the new standards) is to upgrade to the latest version of your browser. If that's not possible, consider moving to the Standards-compliant and open-source Mozilla browser.

July 13, 2006

Seminar on 2-Vector Bundles and Elliptic Cohomology, VI

Posted by Urs Schreiber

In the 4th (and probably last) session of our seminar Birgit Richter talked in more detail about

\;\; 0) elliptic curves and formal groups

\;\;1) “classical” elliptic cohomology (according to Landweber, Ochanine and Stong)

and a tiny bit about

\;\;2) topological modular forms (due mainly to Hopkins)

and ran out of time before talking about

\;\;3) other forms of elliptic cohomology (e.g. Kriz-Sati) ,

complementing my rough outline last time with more technical details.

0) elliptic curves and formal groups

The Weierstrass form of an elliptic curve is an equation in two variables xx and yy of the form

(1)E:y 2+a 1xy+a 3y=x 3+a 2x+a 4y+a 6. E : \;\;\; y^2 + a_1 xy + a_3y = x^3 + a_2 x + a_4 y + a_6 \,.

There is something called the discriminant Δ\Delta of EE and if it is nonvanishing we have a smooth curve.

Thinking of the above equation as living over the real numbers, such smooth curves are certain smooth curves in 2\mathbb{R}^2. Straight lines in 2\mathbb{R}^2 which coincide with this curve in three point PP, QQ and RR define an abelian group structure on points by setting

(2)P+Q+R=0. P + Q + R = 0 \,.

For many applications it is convenient to perform a coordinate transformation from (x,y)(x,y) to (w,z)(w,z) with

(3)w =1y z =xy. \begin{aligned} w &= - \frac{1}{y} \\ z &= -\frac{x}{y} \,. \end{aligned}

Then there is an ff such that the above equation for the ellitptic curves reads equivalently

(4)w=f(z,w). w = f(z,w) \,.

By iteratively re-inserting ff into itself according to this equation, we find that

(5)f(z,w)=z 3(1+A 1z+A 2z 2+)[a 1,a 2,a 3,a 4,a 6][[z]], f(z,w) = z^3( 1 + A_1 z + A_2 z^2 + \cdots ) \in \mathbb{Z}[a_1,a_2,a_3,a_4,a_6][[z]] \,,

which is a power series in zz starting in degree 3, with coefficients being polynomials in the a ia_i over the integers.

Using this, we can understand the above mentioned addition on the elliptic curve as given by a power series in two variables. Namely, if (w 1,z 1)(w_1,z_1) and (w 2,z 2)(w_2,z_2) are two points on the smooth elliptic curve EE (which means that the zz coordinate is determined by ww), then the result of adding them has a zz-coordinate which is given by a power series

(6)F E(z 1,z 2)[a 1,,a 6][[z 1,z 2]]. F_E(z_1,z_2) \in \mathbb{Z}[a_1,\cdots, a_6][[z_1,z_2]] \,.

This F EF_E is a formal group law, which implies (as I mentioned last time) that it satisfies equations

1) F(z 1,0)=z 1F(z_1,0) = z_1

2) F(z 1,z 2)=F(z 2,z 1)F(z_1,z_2) = F(z_2,z_1)

3) F(F(z 1,z 2),z 3)=F(z 1,F(z 2,z 3))F(F(z_1,z_2),z_3) = F(z_1,F(z_2,z_3))

for all z iz_i.

Form this one can show that inverses of all elements exist.

The prototypical example of such a formal group law is obtained by taking a 1-dimensional real Lie group, looking at the tangent space T eT_e at a given point ee, using the exponential map to identify a neighbourhood of ee in the group with the tangent space and expanding for x,yT ex,y \in T_e the multiplication in the group as

(7)μ(x,y) = n,ma nmx ny m =x+y+higher terms[[x,y]]. \begin{aligned} \mu(x,y) &= \sum_{n,m} a_{nm} x^n y ^m \\ &= x + y + \text{higher terms} \in \mathbb{R}[[x,y]] \end{aligned} \,.

In general, formal groups are local expansions of group laws. The power series F EF_E associated to a smooth elliptic curve as described above is similarly the expansion of the additve group law defined by the elliptic curve.

Quillen explained that formal groups are related to complex cobordisms.

Let MU *MU_* be the complex bordsim ring, which is the ring whose elements are cobordism classes of (stably) complex manifolds with multiplication being cartesian product and additon being disjoint union. We write

(8)Ω * U=MU *. \Omega_*^U = MU_* \,.

This ring is universal for formal group laws in the sense that there is a formal group law

(9)F MU F_{MU}

over MU *MU_* such that for every formal group law FF over any ring RR there is a unique ring homomorphism

(10)θ:MU *R \theta : MU_* \to R

such that

(11)F=θ *F MU F = \theta_* F_{MU}

which means that if F MU(x,y)= n,ma nmx ny mF_{MU}(x,y) = \sum_{n,m} a_{nm}x^n y^m then

(12)F(x,y)= n,mθ(a nm)x ny m. F(x,y) = \sum_{n,m} \theta(a_{nm})x^n y^m \,.

Recalling from above that every elliptic curve EE gives rise to a formal group law F EF_E over the ring [a 1 ,,a 6]\mathbb{Z}[a_1^,\cdots,a_6], we find that for every elliptic curve there is a unique ring homomorphism

(13)θ E:MU *[a 1,,a 6]. \theta_E : MU_* \to \mathbb{Z}[a_1,\dots,a_6] \,.

Using this homomorphism we get an action of MU *MU_* on [a 1,,a 6]\mathbb{Z}[a_1,\cdots,a_6]. We want to use this to form a generalized cohomology theory (\to) by tensoring [a 1,,a 6]\mathbb{Z}[a_1,\dots,a_6] with the universal cohomology MU *MU_* theory defined by complex cobordisms.

Instead of describing the ring spectrum which represents MU *MU_*, we here just say how the MUMU cohomology MU *(X)MU_*(X) of any space XX looks like.

We set MU n(X)MU_n(X) to be the ring of maps

(14)ϕ:M nX \phi : M^n \to X

from stably complex nn-manifolds M nM^n to XX, where we identitfy two maps if their domain manifolds are cobounded by a stably complex (n+1)(n+1)-manifold.

Here stably complex means that we can embed M nM^n in some R N\mathrm{R}^N for NN sufficiently large, such that the normal bundle of M nM^n in N\mathbb{R}^N is a \mathbb{C}-vector bundle.

The entire ring MU *(X)MU_*(X) is just the direct sum

(15)MU *(X)= nMU n(X) MU_*(X) = \oplus_n MU_n(X)

and in particular the bare MU *MU_* from above is shorthand for the MUMU-cohomology of a point

(16)MU *:=MU *(pt). MU_* := MU_*(\text{pt}) \,.

It is important for the following construction that there is a natural graded action of MU *(pt)MU_*(\text{pt}) on any MU *(X)MU_*(X)

(17)MU n(pt)×MU m(X)MU n+m(X) MU_n(\text{pt}) \times MU_m(X) \to MU_{n+m}(X)

simply given by taking a map

(18)M npt M^n \to \text{pt}


(19)M mX M^m \to X

and forming the obvious map

(20)M n×M mpt×XX. M^n \times M^m \to \text{pt}\times X \simeq X \,.

In summary, we have an action of the ring MU *(pt)MU_*(\text{pt}) both on the ring [a 1,,a 6]\mathbb{Z}[a_1,\cdots,a_6] and the ring MU *(X)MU_*(X), for all XX.

Hence, for each elliptic curve EE and each space XX, we can form the graded ring

(21)E *(X):=MU *(X) MU *(pt)[a 1,,a 6]. E_*(X) := MU_*(X) \otimes_{MU_*(\text{pt})} \mathbb{Z}[a_1,\cdots,a_6] \,.

This is the elliptic cohomology ring of XX with respect to the elliptic curve EE.

As an example for this we recover ordinary integral cohomology and K-theory as degenerate cases of elliptic cohomology.

Namely, if our elliptic curve happens to be

(22)y 2=x 3 y^2 = x^3

with a bad singularity at (0,0)(0,0), the corresponding group law is simply (this is not supposed to be obvious)

(23)F(x,y)=x+y. F(x,y) = x + y \,.

As I reviewed last time, this is the group law which corresponds to ordinary integral cohomology.

The elliptic curve

(24)y 2=x 3+x 2 y^2 = x^3 + x^2

has a singularity which is not quite as bad. It gives rise to the group law

(25)F(x,y) =1(1x)(1y) =x+y+xy. \begin{aligned} F(x,y) &= 1 - (1-x)(1-y) \\ &= x + y + xy \end{aligned} \,.

As you can see from the table given last time, this is the group law which identifies complex K-theory.

1) classical elliptic cohomology

A special case of elliptic curves are the Jacobi curves, which are of the form

(26)y 2=12δx 2+ϵx 4, y^2 = 1 - 2\delta x^2 + \epsilon x^4 \,,

depending on two parameters δ\delta and ϵ\epsilon.

The discriminant of these is

(27)Δ=ϵ(δ 2ϵ) 2. \Delta = \epsilon(\delta^2 - \epsilon)^2 \,.


(28)g 2 :=(δ 23ϵ)/3 g 3 =δ(δ 29ϵ)/27 \begin{aligned} g_2 &:= (\delta^2 - 3\epsilon)/3 \\ g_3 &= \delta(\delta^2 - 9 \epsilon)/27 \end{aligned}

we can alternatively write

(29)y 2=4x 3g 2xg 3, y^2 = 4 x^3 - g_2 x - g_3 \,,

which however works only in characteristic >3\gt 3, which is problematic in particular when applied to speher spectra, cause homotopy classes there have lots of 2- and 3-torsion.

Anyway, the formal group law corresponding to these curves is

(30)F(x,y)=xR(y)+yR(x)R(x), F(x,y) = \frac{x \sqrt{R(y)} + y \sqrt{R(x)}}{R(x)} \,,


(31)R(x):=12δx 2+ϵx 4. R(x) := 1 - 2\delta x^2 + \epsilon x^4 \,.

This formula was originally found by Euler, even though he did not call it a formal group law.

We can rewrite F(x,y)F(x,y) as

(32)F(x,y)=g 1(g(x)+g(y)), F(x,y) = g^{-1}(g(x)+g(y)) \,,


(33)g(x)= 0 xR(t) 1/2dt. g(x) = \int_0^x R(t)^{-1/2}\, dt \,.

Now let M˜ *\tilde M_* be the ring of modular forms under the subgroup of SL(2,)SL(2,\mathbb{Z}) generated by ττ+2\tau \mapsto \tau + 2 and τ1τ\tau \mapsto -\frac{1}{\tau}.

There is a theorem due to Landweber, Ravenal and Stong which says that for δ\delta and ϵ\epsilon algebraically independent over \mathbb{Q} we have

(34)M˜ *=[12][δ,ϵ] \tilde M_* = \mathbb{Z}[\frac{1}{2}][\delta,\epsilon]

and for all of the rings in the diagram

(35)M˜ * M˜ *[(δ 2ϵ) 1] M˜ *[ϵ 1] M˜ *[Δ 1] \array{ \tilde M_* &\to& \tilde M_*[(\delta^2 - \epsilon)^{-1}] \\ \downarrow && \downarrow \\ \tilde M_*[\epsilon^{-1}] &\to& \tilde M_*[\Delta^{-1}] }

there exists a generalized homology theory h *h_* such that h *(pt)h_*(\text{pt}) is that given ring.

This is constructed by noticing that the formal group law defes an action of the oriented cobordism ring on the given ring R *R_* from above, which allows us to form the homology ring of some space XX as

(36)XΩ * SO(X) Ω * SOR *. X \mapsto \Omega^\mathrm{SO}_*(X) \otimes_{\Omega_*^\mathrm{SO}} R_* \,.

This is the homology which does the job.

The relation to genera is as follows.

A genus is a ring homomorphism

(37)ϕ:Ω * SOΛ \phi : \Omega_*^\mathrm{SO} \to \Lambda

from oriented cobordisms to any other ring Λ\Lambda that is also a \mathbb{Q}-algebra.


(38)Ω * SO[[P 2],[P 4],] \Omega_*^\mathrm{SO} \otimes \mathbb{Q} \simeq \mathbb{Q} \left[ \left[\mathbb{C}P^2\right], \left[\mathbb{C}P^4\right], \cdots \right]

it suffices to specify ϕ\phi on [P 2n][\mathbb{C}P^{2n}].

One calls the expression

(39)g ϕ(x):= 0 x n0ϕ[P 2n]t 2ndt= n0ϕ[P 2n]x 2n+12n+1 g_\phi(x) := \int_0^x \sum_{n \geq 0} \phi[\mathbb{C}P^{2n}]t^{2n}\, dt = \sum_{n\geq 0} \phi[\mathbb{C}P^{2n}] \frac{x^{2n+1}}{2n+1}

the logarithm of the genus.

Ochanine defined an genus to be elliptic if this logarithm is of the form

(40)g ϕ(x)= 0 x(12δt 2+ϵt 4) 1/2dt g_\phi(x) = \int_0^x (1 - 2\delta t^2 + \epsilon t^4)^{-1/2} \; dt

for suitable ring elements δ\delta and ϵ\epsilon, algebraically independent over \mathbb{Q} and Δ0\Delta \neq 0.

There is a theorem by Landweber, Ochanine and Stong which says that if a genus ϕ\phi is elliptic, then its image is

(41)ϕ(Ω * SO)=[δ,2γ,2γ 2,,2γ 2 n] \phi(\Omega_*^\mathrm{SO}) = \mathbb{Z}[\delta, 2\gamma, 2\gamma^2,\cdots, 2\gamma^{2^n}]


(42)γ=δ 2ϵ4. \gamma = \frac{\delta^2 - \epsilon}{4} \,.

Furthermore, the image of spin cobordisms is

(43)ϕ(Ω * Spin)=[16δ,(8δ) 2,ϵ]. \phi(\Omega_*^\mathrm{Spin}) = \mathbb{Z}[16 \delta, (8\delta)^2, \epsilon] \,.

Again, looking at degenerate cases we find famliar examples.

1) In the case that ϵ=δ\epsilon = \delta we get

(44)g ϵ=1,δ=1(x) = 0 x(12t 2+t 4) 1/2dt = 0 x11t 2dt =tanh 1(x). \begin{aligned} g_{\epsilon=1,\delta=1}(x) &= \int_0^x (1-2t^2 + t^4)^{-1/2} \, dt \\ &= \int_0^x \frac{1}{1-t^2}\,dt \\ &= \mathrm{tanh}^{-1}(x) \,. \end{aligned}

This corresponds to the signature genus of L-genus (which I also mentioned last time).

Here, too, the corresponding Jacobi-curve is singular.

2) Another example is ϵ=0\epsilon = 0 and δ=18\delta = -\frac{1}{8}. Here one gets

(45)g 18,0(x)= 0 x(114t 2) 1/2dt g_{-\frac{1}{8},0}(x) = \int_0^x (1-\frac{1}{4}t^2)^{-1/2}\, dt

and this corresponds to the A^\hat A-genus.

As Atiyah and Singer found with their famous index theorem, for MM compact and spin and dimM=2n\mathrm{dim} M = 2n the A^\hat A-genus

(46)A^(M)=ind(D) \hat A(M) = \mathrm{ind}(D) \in \mathbb{Z}

is the index of a Dirac operator on MM, taking values in the integers.

We want some lifting of this statement to the loop space of MM.

It turns out that this is possible if MM is string (\to), which means, according to a theorem by Laughlin, that it sfirst two Stieffel-Whitney classes and one half of the first Pontryagin class vanishes.

In this case there is something like a Dirac operator on loop space LMLM and a theorem due to Witten and Zagier says that its index is a genus which is the qq-series of a modular form - the Witten genus (partition function of the heterotic string).

What we are after is the homology theory which corresponds to this genus.

tmf - topological modular forms

According to Birgit Richer, in her experience it takes a group of experts a full week to discuss the construction of tmf. At that point 4 minutes time were left.

Apart from that, what is important about tmf is that, as Jacob Lurie describes on pp. 9-10 of his “survey” (\to), tmf is something like the universal elliptic cohomology.

We can get a glimpse of what this means by realizing that the way elliptic cohomology was defined above depended on a choice of coordinates (in the Weierstrass form) for an elliptic curve. In a vague sense tmf is the coordinate-free version of alliptic cohomology. Or something like that.

The point is that the Weierstrass form

(47)y 2+a 1xy+a 3y=x 3+a 2x 2+a 4x+a 6 y^2 + a_1 xy + a_3 y = x^3 + a_2 x^2 + a_4 x + a_6

is invariant under the coordinate transformations of the form

(48)x λ 2+r y λ 3y+λ 2sx+t, \begin{aligned} x &\mapsto \lambda^2 + r \\ y &\mapsto \lambda^3 y + \lambda^2 sx + t \end{aligned} \,,

where λ\lambda is a “unit” (invertible). Call the group of these transformations

(49)G. G \,.


(50)A=[a 1,,a 6][u ±1] A = \mathbb{Z}[a_1,\cdots,a_6][u^{\pm 1}]

and form the cohomology theory

(51)X(E A) *(X)=MU *(X) MU *A. X \mapsto (E_A)_*(X) = MU_*(X) \otimes_{MU_*}A \,.

Then, according to a theorem by Hopkins, Miller, Goerss which has been given a more conceptual proof by Jacob Lurie, the GG-invariant part (E A) G(E_A)^G of E AE_A is a model for tmf.

(I am only 30 per cent convinced that this statement makes good sense as stated. Need to check that.)

There is a map from the tmf cohomology of a point to modular forms

(52)tmf *(pt)M * \mathrm{tmf}_*(\text{pt}) \to M_*

whose kernel and cokernel are annihilated by 24.

Moreover, it is known that

(53)tmf *[16][pt]=[1/6,c 4,c 6]=M *[1/6]. \mathrm{tmf}_*[\frac{1}{6}][\text{pt}] = \mathbb{Z}[1/6,c_4,c_6] = M_*[1/6] \,.

Finally, the Witten genus is the composition map

(54)MStringtmfKU[[q]] M\mathrm{String} \to \mathrm{tmf} \to KU[[q]]

from string cobordisms over tmf\mathrm{tmf} to power series in qq with coefficients in K-theory, restricted to the 00-dimensional string manifold.

Oh dear, you can tell that this last part was transmitted and received in mere 4 minutes.

Posted at July 13, 2006 8:04 PM UTC

TrackBack URL for this Entry:

1 Comment & 0 Trackbacks

Re: Seminar on 2-Vector Bundles and Elliptic Cohomology, VI

We invite you to submit your papers to our journals ( with advanced referring system and participate in our open forums at

Advanced publications

Posted by: Jonathan on May 26, 2007 9:12 AM | Permalink | Reply to this

Post a New Comment