Geometry of Formal Varieties in Algebraic Topology

This talk has since been given to a live audience and typeset at 1 and 2.

 all right, it's noon by my clock :) the seminar topic is on one role of algebraic geometry in algebraic topology. i've tried to make the vocabulary in the seminar fairly unassuming; we'll see if i've succeeded -!- ChanServ changed the topic of #mathematics to: If you have a question, type '!' and wait for me to cede the floor. the impetus of algebraic geometry is to try to assign geometric meaning to concepts coming from ring theory.  the classical example is that, given a set X of n-variate polynomials over a(n algebraically closed) field k, we can associate the simultaneous vanishing locus of all elements of the set, Z(X), in n-dimensional k-space.  Z(X) is called an algebraic set, and irreducible (in a sense) algebraic sets are called varieties a key thing to notice here is that if you have a larger X, you have more polynomials, and so it gets harder for all of them to have zeroes in the same spots --- and so the associated vanishing locus gets smaller. this means that, if we expect any kind of correspondence of this type between algebra and geometry, it has to be 'direction reversing' a lot of energy was invested in figuring out what setting we can build that 1) supports all rings and 2) has nice geometric properties. varieties, for instance, are insufficient; they turn out only to give a geometric description of finitely generated k-algebras with no nilpotent elements eventually, things called schemes were invented and people presented them as locally ringed spaces --- this is what everyone sees in a first or second semester algebraic geometry course. since i come from topology, i find a different presentation of schemes much easier to think about: their functors of points from a categorical perspective, this presentation of the opposite category of rings is almost obvious; to every ring R we can build an object Hom(R, -), which is functor that assigns rings S to the set of ring homomorphisms R --> S. the assignment spec: R |-> Hom(R, -) is contravariant (i.e., inclusion-reversing), and so provides an embedding of Rings^op into some larger category of functors off the category of rings, which I will call the category of schemes hmm have to keep an eye on long lines while not very obviously geometric, this idea of studying rings by studying the maps off them turns out to be a very good idea. in a sense, rather than asking "what is the ring R?" it suggests that we instead ask "what does the ring **DO**?" to illustrate what i mean, take the polynomial ring Z[x], and associate to it the scheme spec Z[x]. we can apply spec Z[x] to a test ring S and see what we get: (spec Z[x])(S) = hom(Z[x], S). a map Z[x] --> S is determined completely by where x is sent and x can be sent to any element --- so spec Z[x] models the forgetful functor Rings --> Sets. this is kinda neat to get a sense of what else rings can do, recall that each ring R comes with an additive group (R, +) on its underlying set of points, and maps between rings give rise to maps between these additive groups the Yoneda lemma is a fundamental result of category theory, which in one language says that structure visible on the image of a representable functor actually lives on the representing object --- so we should expect a group structure on the scheme spec Z[x] or, since spec is direction-reversing, we could call this a cogroup (or 'Hopf algebra') structure on the ring Z[x]. to illustrate, we should get a map Z[t] --> Z[x] (x) Z[y] corresponding to the 'opposite' of addition in terms of the action of spec, to any two maps Z[x] --> S and Z[y] --> S (called S-valued points of the schemes spec Z[x] and spec Z[y]) we need to associate a map spec Z[t] --> S corresponding to their sum the content of Yoneda's lemma is that it's enough to do this "in the universal case": the identity maps Z[x] --> Z[x] and Z[y] --> Z[y] classify the elements x and y, and hence Z[t] --> Z[x] (x) Z[y] should classify the element x + y --- and this is the opposite addition map we wanted this gives rise to a comultiplication map on Z[x], and so a 'multiplication' map on spec Z[x]. spec Z[x] together with this group structure we call the 'additive group (scheme)', written Ga this happens *all the time* with all kinds of interesting functors on the category of rings another functor everyone knows is the group of units: there's a functor Gm with Gm(R) equal to the invertible points of R. to get a point of Gm, we need to select an element x in R and another element y playing the role of its inverse, so xy = 1. using this logic, we can model Gm as spec Z[x, y] / . this scheme is called the 'multiplicative group' shifting gears slightly, one of the classical presentations of infinitessimal elements are elements that 'square to zero'. this turns out to be a fairly difficult thing to make both precise and analytically useful, which is why most analysts spend their time thinking about other things --- but it's a very useful notion in algebraic geometry. we think of the nilpotent elements of a ring R as being infinitessimal there's a very useful category related to our category of schemes. let R be a topologized augmented k-algebra whose augmentation ideal I is topologically nilpotent --- the category of such objects is called the category of adic k-algebras (with maps continuous homomorphisms) we can do the same functor-of-points construction here, but the resulting schemes are called 'formal schemes' --- they're contravariant functors from adic k-algebras to Sets. < Kasadkad> ! yeah? < Kasadkad> what's topologically nilpotent? it means that for any n there's an m so that x^m lies in I^n, where R has the I-adic topology so rather than requiring actual nilpotence, you mean the limit of the powers of the elements tends to zero in this sense that make sense? < Kasadkad> yes cool < Kasadkad> and the augmented algebra bit means R/I =~ k? yeah < Kasadkad> ok so, as an example, the ring of power series kx is an augmented adic k-algebra with the -adic topology, and so gives a formal scheme spf kx. this example is particularly relevant --- there's a subcategory of formal schemes called formal varieties, consisting of those functors which are (noncanonically) isomorphic to spf kx1, ..., xn for some n the name, for those interested, comes from taking the sheaf of functions of a smooth algebraic variety and completing at a point. what you get is a k-algebra of this type, where the number of indeterminates is equal to the dimension of the variety and this is how you should think of a formal variety. just like a smooth manifold has about every point a neighborhood where it looks like euclidean space, a smooth variety has about every point an infinitessimal neighborhood where it looks like 'formal n-space', and we're interested in that tiny neighborhood some more vocabulary: a formal variety V is something noncanonically isomorphic to spf kx1, ..., xn = A^n. a selected isomorphism between V and A^n is called a 'coordinate' on V, corresponding to picking charts on a manifold. the space of maps A^n --> A^m corresponds to m-tuples of n-variate power series, just like maps between charts on analytic manifolds and now here i thought i'd take a break to let you catch your breath and ask questions if you have any < Kasadkad> "spf kx1, ..., xn = A^n" is just introducing the notation "A^n"? yeah and 'spf', which i think i forgot to define, is the analogue of 'spec' for formal schemes < Kasadkad> so i gathered ok, let's get back to it: let's apply some of these words to algebraic topology. in first or second semester algebraic topology, one thing everyone learns about singular cohomology is that it supports a ring structure --- there's a notion of 'cup product' H^n M x H^m M --> H^(n+m) M for a space M as budding algebraic geometers, rather than trying to write down what these rings are, we know that we should try to write down what these rings 'do'. this might not sit well with those of you that haven't seen lots of algebraic topology; the examples given in an introductory course have very sparse information, but basically all spaces of lasting interest to a topologist have extremely complicated cohomological information embedded the cohomology ring associated to S^n, for instance, isn't so gripping --- but K(Z/p, n), for instance, is at any rate, this suggests that rather than thinking of H^* M as a cohomology ring, we should think of M_H = spf H^* M as a formal scheme (where the ideal inducing the I-adic topology corresponds to cohomological information above degree 0) over the formal scheme S_H, detecting the coefficient ring of H (in fact, H can be replaced by any ring-valued cohomology theory to produce a similar theory of schemes associated to spaces) (one thing the reader should note (and object to!) is that this construction is extremely insensitive to issues with grading. namely, we've thrown away the natural grading given to us by cohomological dimension, and we've even assumed that the odd-dimensional parts vanish, so that we get a strictly commutative ring.  this is a lot to ask, and introduces a few problems that we'll ignore :) ) so, what are some examples where this ideology is a good idea in topology?  one calculation that everyone should know is the singular cohomology of infinite dimensional complex projective space: H^* CP^infty = Zx, with x a cohomology class in degree 2 (those of you who are especially picky might also object here to the use of power series rather than a polynomial ring; this is totally a matter of taste in taking a limit or a colimit when building the cohomology ring from its graded pieces, and i'm going to take a limit because it makes my theory much more interesting) so, CP^infty_H = spf H^* CP^infty is isomorphic formal affine line, A^1 = spf Zx, where our choice of x in the cohomology of CP^infty gave us a coordinate on CP^infty_H. that's neat, albeit not too interesting; it's the most basic formal scheme we know of, pretty much to help make it interesting, there is a group structure on CP^infty, as the classifying space for line bundles, corresponding to the universal tensor product of line bundles let's study what structure this induces on the formal scheme associated to CP^infty by computing the product map CP^infty_H x CP^infty_H --> CP^infty_H using the coordinate we have we effectively need to ask what x pulls back to under the multiplication map.  well, a priori, we have that it pulls back to some bivariate power series F in H^* (CP^infty x CP^infty) = Zx, y the niceness of tensor product tells us a few things about F: the tensor product is commutative, so F(x, y) = F(y, x). the tensor product satisfies L (x) 1 = L for any bundle L, so F(x, 0) = x. lastly, the tensor product is associative, (L (x) L) (x) L = L (x) (L (x) L), so we in ordinary cohomology, we can actually calculate F. the Kunneth formula says that in the calculation H^*(CP^infty x CP^infty) = Zx, y with x and y both of degree 2, so the only possible thing F could be is F(x, y) = x + y, due to degree restrictions --- but this is already interesting. :) another interpretation of this calculation is that the induced map CP^infty_H x CP^infty_H --> CP^infty_H corresponding to the product on formal schemes is given by the power series F(x, y) = x + y, which, as we calculated earlier, is exactly the function corresponding to the addition in the additive group and so CP^infty_H is another name for Ga let's do the same calculation for K-theory, another kind of cohomology theory, where K^* X corresponds basically to vector bundles over X. for some topological reasons, writing R for the coefficient ring of K-theory, K^* CP^infty also turns out to be Rx, and so CP^infty_K is also an affine line CP^infty_K comes with a natural choice of coordinate: namely, to the universal line bundle L over CP^infty, we associate the K-theoretic class [L]-1 corresponding to L (the -1 is to fix some dimensional issues; you should not think too hard about it) that tensor product corresponds to multiplication in K-theory means that we can calculate the action of the tensor product manually: this is kind of ugly over IRC :) but: write c(L) for the image of x under pullback along some map M --> CP^infty classifying a line bundle L over M. then c(L (x) L') = [L][L'] - 1 = [L][L'] - [L] - [L'] + 1 + [L] - 1 + [L'] - 1 = [L]-1)([L']-1) + [L]-1 + [L']-1 = c(L)c(L') + c(L) + c(L'), and so F(x, y) = xy + x + y here's the real punchline, so you can skip the previous line: if you work it out, this is exactly the product structure associated to the multiplicative group Gm with the coordinate x - 1 ! so CP^infty_K is isomorphic to Gm hmm looks like i sped up a bit again < Kasadkad> i'm with you cool :) briefly: some other less pivotal examples include BU(n)_H, which is isomorphic to n-dimensional affine space, with group structure corresponding to divisors of degree n on S_H. in the limit, BU_H corresponds to the formal group of divisors on S_H at a prime p, the even part of spec H_* K(Z, 3) turns out to corepresent the functor of Weil pairings Ga^2 --> Gm if you're familiar with Morava K-theory, K(Z, *)_K(n) is the free ring object on BZ/p^infty_K(n), which is also tied in with Weil pairings on p-divisible groups i don't really want to explain either of these examples, i just want to point out that this geometric language organizes huge calculations elsewhere in topology too < Kasadkad> ! sure < Kasadkad> what is S_H, at least? oh, did i not explain that earlier? S_H is H^* of a sphere, so spf of the coefficient ring of H < Kasadkad> don't think so, ok it corresponds to the 'k' in 'category of adic k-algebras' since every based space X comes with maps pt --> X --> pt, you get an augmented H^* pt-algebra structure on H^* X by H^* pt <-- H^* X <-- H^* pt gonna take another couple minute break all right, back to it: let's stick to analyzing CP^infty, since we recovered our two earlier examples with it and there seemed to be a lot left, as those axioms we imposed on F were much laxer than "F is x + y OR F is x + y + xy." the cases where our analysis of the situation thus far goes through is when CP^infty_E is isomorphic to the affine line over S_E for our cohomology theory E of interest. these cases will be called 'complex orientable', and a choice of coordinate on CP^infty_E is a 'complex orientation'. the resulting bivariate power series F is called a formal group law the absolute most important theorem in the study of complex oriented spectra has to do with the structure of a cohomology theory called complex cobordism, MU. the theorem, due to Quillen, has two parts, i'll talk about both in turn, but not about their proofs, which involve a complicated piece of topology called H_infty ring spectra and generalized power operations the first thing we should notice is that the space of formal group laws itself constitutes a scheme there is a scheme representing all bivariate power series: spec Z[a_00, a_10, a_01, a_20, a_11, a_02, ...] = spec P, which supports the power series f = sum_{i, j} a_ij x^i y^j. any other power series over any other ring R occurs uniquely as a pullback of f along a map spec R --> spec P (equiv, a map P --> R) selecting the coefficients of the power series then, we can impose the conditions on the representing ring specified by the axioms of formal group laws through a quotient, since those are all algebraic --- just like we did for Gm = spec Z[x, y] /  symmetry, for instance, asserts that a_{ij} - a_{ji} = 0, and rigidity states that a_{i0} = 0 for all i >= 1, and the associativity condition is much harder to write down. the resulting ring is called L, since Lazard first studied it and found that it was (noncanonically) isomorphic to the infinite polynomial ring Z[c2, c3, ...] --- a very nontrivial result! the first part of Quillen's first theorem states that S_MU, the coefficient ring of MU, is (noncanonically) isomorphic to spec L the second part says that this is really a topological statement: if E is a complex oriented ring spectrum and we've chosen an orientation for MU, then there is a unique map multiplicative map MU --> E of cohomology theories (in fact, of spectra) so that S_E --> S_MU = spec L classifies the formal group law associated to the complex orientation of E this is pretty nuts on its own --- this means that the entire theory of commutative one-dimensional formal group laws embeds faithfully into algebraic topology through CP^infty and the tensor product of complex line bundles. this has a *lot* to say about the algebraic geometry associated to cohomology theories! but it's not over yet! the product spectrum MU ^ MU comes with two maps from MU: there's inclusion on the left and inclusion on the right correspondingly, we have two complex orientations of CP^infty_{MU ^ MU}, but we the scheme CP^infty_{MU ^ MU} exists totally in isolation of the words 'complex orientation' --- so what we're really saying is that we have two coordinates of CP^infty_{MU ^ MU}, and they are necessarily related by an invertible change of coordinates for almost formal reasons, this turns out to mean that, just like the coefficient ring of MU carried the universal example of a formal group law, the coefficient ring MU ^ MU carries the universal example of a formal group law in topological terms, this means that the homology cooperations associated to MU correspond to reparametrizations of its formal group. to explain, to all cohomology theories E we get a map E^* E (x) E^* X --> E^* X corresponding to an action of 'cohomology operations', and in nice cases we get a reverse map in homology E_* X --> E_* X (x) E_* E corresponding to a coaction of 'homology cooperations' geometrically, what does this mean? Lazard's ring L corresponds to the 'moduli space of formal group laws', in the sense that to any formal group law F over any ring R, we get a map spec R --> spec L pulling back the universal formal group law to F. but, using our language of formal varieties, we see that in doing so we've made an arbitrary choice --- what we'd really like is a formal group, without a choice of coordinate gosh what a sea of text correspondingly, the group of automorphisms of A^1 as a formal variety act on spec L, corresponding to the action of S_{MU ^ MU} on S_MU. over an arbitrary formal group spec R with a coordinate and formal group law F, we have a slightly smaller action: Aut F acts on spec R this data assembles into what is called a(n affine, algebraic) stack. in the language of functors of points, what we're saying is that we're associating more than just a set of formal group laws to R: we're also associating a bunch of isomorphisms between them, so that taking connected components of the resulting groupoid gives you the space of formal group structures on spec R in fact, the interpretation of the coefficient ring of MU ^ MU as reparametrizations of the universal formal group law and as the homology operations associated to complex bordism means that the homology cooperation coaction on MU_* X means that MU_* X gives rise to a automorphism-equivariant sheaf on S_MU, or equivalently a sheaf on this stack --- well, before that means anything, i guess i should say what a (quasicoherent) sheaf of modules is over a formal scheme X to each ring R and map spec x: R --> X (called an R-valued point of X), we assign an R-module M_x in a functorial way (i.e., maps spec S --> spec R --> spec X push modules forward up to isomorphism and so forth). whoops x: spec R --> X, and spec S --> spec R --> X in our setting, this is really not an exciting statement --- MU_* X is indeed an MU_*-module, and any map spec R --> S_MU coming from a map f: MU_* --> R gives rise to an R-module: R (x)_{MU_*} MU_* X. this is all we meant. :) what *is* exciting is the existence of the Adams spectral sequence, which is some functorially available spectral sequence which converges to the homotopy of a space X. the E_2 page in topologist's language starts with Ext_{*, *}^{MU_* MU}(MU_*, MU_* X), which in geometer's language is the cohomology of the sheaf associated to MU_* X over the stack of formal groups what this is indicating is that the category of quasicoherent sheaves of modules over the moduli stack of formal groups is **very** similar to the stable homotopy category, and the geometry of sheaves can tell us a whole lot about the global structure of the entire field of stable homotopy < Kasadkad> ! go ahead < Kasadkad> what does MU_* MU (or E_* E) mean E_* E is shorthand for the homotopy groups of the product E ^ E < Kasadkad> hmm, equivalently, the coefficient ring of the theory E ^ E < Kasadkad> not sure what the product E ^ E is let me paste two more things, then i have another break written in before we make everything concrete, so i'll go back and sketch out what you should expect these things to look like. this question really boils down to 'what is a generalized cohomology theory' < Kasadkad> k here are my two things: probably the most striking example of this relation after complex orientations themselves is gross-hopkins duality, which i don't fully understand yet. the core of it is to use projective geometry to make strong statements about what the moduli stack of formal groups looks like, out of which comes some really striking topology. maybe someday i can say something useful :) and also one way that this picture could be improved is to stop pushing down to homotopy/homology/cohomology groups everywhere, which is what we've had to do in order to say 'ah, look, here's a stack!' derived algebraic geometry is what you get when you try to build this picture without performing this lossy step from homotopy theory to algebra ok, so, regarding what E ^ E is and so forth: so another thing you should know from basic cohomology theory is that the cohomology groups H^n(M; G) of a space are representable as homotopy classes of maps M --> K(G, n) for some special target spaces K(G, n), called Eilenberg-Mac Lane spaces the key relation among these spaces is that if you build the loopspace of K(G, n), you get something homotopy equivalent to K(G, n-1) this is the picture that generalizes well.  a generalized cohomology theory E is made up of a sequence of spaces (E_0, E_1, ...) with Loops E_n homotopy equivalent to E_{n-1} and so an E-cohomology class of a space X corresponds to some homotopy class X --> E_n the language of spectra is invented so that you can think of these constituent spaces as one big collective --- rather than having a homotopy class X --> E_n, you instead build a homotopy class of maps of spectra X --> E < Kasadkad> mm which raises the question: what should a generalized homology theory be? well, it turns out that these available maps E --> X themselves form a spectrum, called the function spectrum F(X, E) and the function spectrum has that nice adjoint property that you've come to expect from elsewhere in math --- there's some notion of a product on the category of spectra, written ^, so that maps(X ^ Y, Z) is the same as maps(X, F(Y, Z)) < Kasadkad> hmm is that coming from smash product or something and so as '^' is dual to 'F(-, -)', the right interpretation of 'E-cohomlogy is dual to E-homology' turns out to be to define E-homology groups of X as the path components of the product spectrum E ^ X yes, it is very much a smash product, but it's tricker than just taking smash products on the individual spaces E_n or whatever < Kasadkad> ok it was a huge source of headaches in the 70s and 80s --- but that's how you should think of it < Kasadkad> yes i was thinking of it as a huge source of headaches already anyway, this smash product on the level of spectra also lets you ask what the E-homology of a spectrum F is haha E_* F is defined to be pi_* E ^ F does that answer your original question? i forget what it even was < Kasadkad> about E_* F and E ^ E < Kasadkad> E ^ F w/e < Kasadkad> and yes cool so there's one more leg of the seminar to go, where we actually take these lofty ideas and compute some thangs let's restrict to thinking about k a field of positive characteristic to a formal group law F over k, we associate an integer n, called its height, as follows: let [m]_F(x) denote x +_F ... m times ... +_F x. then it turns out that [p]_F(x) is always of the form g(x^(p^n)) for some n >= 0 with g'(0) nonzero, and the integer n is called the height of the formal group it is basically measuring the size of the kernel of the multiplication-by-p map on the formal group a theorem of Cartier says that this invariant is exceptionally strong: when k is a perfect field of positive characteristic, a formal group over k is uniquely characterized by its height. (moreover, there is a canonical coordinate, called the Cartier coordinate, associated to each formal group, that is particularly 'nice' in certain senses. we won't need this part)
 * isomorphism*. this is quillen's theorem pt 2

so, what this means is that the quotient of the moduli stack of formal group laws by the coordinate change group (i.e., the orbit space) turns into a bunch of points: N u {infty}. we'd really like to understand the isotropy groups of these individual pieces for instance, let's restrict to the substack of additive formal groups over the prime 2. there's an obvious formal group law, which we've already encountered: F(x, y) = x + y. we can ask: what's the isotropy group of F? well, that's the space of maps A^1 --> A^1 commuting with the group structure on both sides, i.e., a power series f(x) with f(x + y) = f(x) + f(y) in characteristic 2, these are easy to classify: they're the power series of the form f(x) = x + sum_{i>0} a_i x^{2^i} for some sequence of coefficients a_i in the base field --- which is to say that f can be picked out by the scheme spf Z[a_1, a_2, ...] = Aut Ga the group structure on the isotropy group comes from composition of these power series: let g(x) = x + sum_{j>0} b_j x^{2^j}, then g(f(x)) = sum_j b_j (sum_i a_i x^{2^i})^{2^j} = sum_n x^{2^n} (sum_{i+j = n} b_j a_i^{2^j}) the corresponding map of formal schemes Aut Ga x Aut Ga --> Aut Ga sends c_n to sum_{i+j = n} b_j a_i^{2^j} --- and this is exactly the milnor diagonal on the dual of the steenrod algebra, which took topologists decades and much more than 2 lines to unearth so the isotropy group of the formal group law associated to a complex oriented cohomology theory E is somehow telling us about the homology cooperations associated to E this shouldn't be a huge surprise --- since a complex orientation of E corresponds to a map MU --> E, we also get an induced map MU ^ MU --> E ^ E, and hence of homology cooperations. it is a very useful guide for guessing what E ^ E looks like in many, many cases (another note for the careful reader: the cooperation dual to the bockstein for p > 2 is missing from this picture as i've sketched it. it can be accounted for, but it takes some work.) what about the automorphisms above the multiplicative group law, F = x + y + xy --- the one associated to K-theory? well, first we need to work with p-adically completed K-theory, since we've made assumptions about how the base field looks and so forth. then, multiplication by an integer in the formal group structure induces a map n |-> [n]_Gm(x) = (1+x)^n - 1, Z --> Aut_k(Gm) what's neat is that this map extends to a map from the p-adic integers to Aut_k(Gm) by alpha |-> sum_{k >= 1} (alpha choose k) x^k. the units of this ring turn out to exhaust the automorphisms of p-adically completed K-theory, corresponding to the Adams operations i don't actually know much about this interpretation of the Adams operations, so i won't say more there is an obvious question in another direction though: what if rather than considering the map MU ^ MU --> E ^ E, we instead have two complex oriented spectra, MU --> E and MU --> F? then we get a map MU ^ MU --> E ^ F. what does this buy us? in many good cases, the coefficient ring of E ^ F looks like you'd expect --- it's a ring that corepresents the isomorphisms from the formal group law associated to E-theory to the one associated to F-theory, i.e., the space of power series f with f(x +_E y) = f(x) +_F f(y) and f invertible another thing that's worth noting is that this actually sometimes gives unstable information too! if we destabilize the spectra E and F into a sequence of infinite loop spaces E_* and F_*, then the E-homology of the sequence of spaces F_* forms what's called a Hopf ring, which is supposed to corepresent the space of formal group **homomorphisms** CP^infty_E --> CP^infty_F this is also a very fruitful idea that lets you compute, for instance, the Hopf ring BP_* BP_*, or K(n)_* K(n)_*, or K(n)_* K(Z, *) (<-- this was mentioned earlier). steve wilson, doug ravenel, and neil strickland spent a good chunk of their lives computing huge, amazing things using this as a guideline at any rate, all of this should be very exciting. cohomology theories are immensely complicated objects, but it frequently turns out that their rings of (co)operations --- also immensely complicated objects --- are controlled by isotropies of formal group laws, which are really pretty tractable as these things go stumbling across the milnor diagonal was really supposed to be the punchline :) serre's work in the 60s spanned several papers, trying to compute the steenrod algebra using exceptionally complicated (for the time) spectral sequence techniques and i think i'm about done. anything leftover that i can clarify or expand on? < Kasadkad> I'm not sure what [m]_F(x) meant < Kasadkad> or rather what +_F meant x +_F y is another notation for F(x, y).  applying the group law F to the elements x and y < Kasadkad> ok that's what I figured so x +_F x is like 'two times x', sort of, and we write [2](x) gonna switch back to irssi now that i'm not pasting things it might cut off phew < Kasadkad> wb so, putting aside questions about the material for a moment, i will probably have to give a similar talk in front of a Real Live Audience inside of this coming month.  they know some topology, but the smartest one among them asked me whether it was worth his time to learn any algebraic geometry at all --- so i figured i'd do something like this to tell him: **YES** are there things that you think ought to be changed for a second run of this? < Kasadkad> hmm the ending could be modified, so that i at least don't have to say 'hey look you might have missed the punchline but that happened a couple minutes ago' but beyond that, idk < Kasadkad> heh < Kasadkad> i don't really know the material well enough to have suggestions for presenting anything differently but as it was presented, it didn't feel like too much was omitted? you could actually follow point to point, and things were somewhat justified < Kasadkad> yeah well that's something < Kasadkad> i kept up for longer than i expected < Kasadkad> and when my eyes were glazing over, i think it was on account of not really knowing topology < Kasadkad> so if they know some that's what they tell me they spent last week constructing the steenrod squares and intend to continue this week on tuesday, so i guess this punchline will be especially potent since it doesn't take 3 hours to state it's a little encouraging about how much topology they actually know though. idk < Kasadkad> who are "they", anyway, some topology classes? < Kasadkad> *-es some people attending this student-run seminar http://math.berkeley.edu/~aaron/xkcd/ grad students from berkeley and one guy from stanford why would they call it that? < Kasadkad> ah burned: idk, bad decisions at any rate, they're interested in topology and competent, but they're young and maybe not experienced or learned eeww and nerds, so they are a triple threat < jimi_> Where is the log going to be? good question yourwiki is still down, looks like