I was up in Boston for a few days, and managed to attend a few of the talks at the conference in honor of George Lusztig’s 60th birthday. Lusztig started out his career in geometry and topology; his thesis was in the area of index theory, working with Michael Atiyah and using the families version of the index theorem. He soon turned his attention to representation theory, which is the field that he has worked in for most of his career, often from a quite algebraic point of view. His papers are dense and can be difficult to read, especially for someone like me who is not so algebraically inclined, but many speakers at the conference remarked on how their work had drawn important inspiration from one or another of these papers.
Among the things he is famous for are his work on quantum groups, on the representation theory of reductive groups over finite fields (called Deligne-Lusztig theory, for an introduction, see here), on a whole new field in Lie theory known as Kazhdan-Lusztig theory (for an introduction, see the article by Deodhar in the proceedings of the 1991 AMS summer institute on algebraic groups), and many other things.
Of the few talks I heard at the conference, two were really exceptional. One of these was by Michael Atiyah, with the title “Quaternions in Geometry, Analysis and Physics”. He began by explaining that not only was Lusztig 60, but, if he were alive, the Irish mathematician Hamilton would be 200. There’s a famous story about Hamilton’s discovery of the quaternions: this took place in a flash of insight on October 16, 1843, after which he supposedly engraved the defining relations of the quaternion algebra into a Dublin bridge. Atiyah described a piece of history I didn’t know, showing an extract from a 1846 paper of Hamilton’s in which he takes a square root of the Laplacian and essentially writes down the Dirac equation (in Euclidean signature, this was long before special relativity…).
Hamilton was very taken with quaternions as a generalization of complex numbers, and wanted to develop a “quaternionic analysis” that would be a generalization of complex analysis, a project he thought would take him at least ten years. It turns out that you can’t simply generalize the beautiful subject of complex analysis and algebraic geometry over the complex numbers to the quaternionic case. Because of non-commutativity, polynomials behave very differently. Atiyah explained that in his view the correct generalization of complex analysis to the quaternionic case was Penrose’s twistor theory. Here one considers all possible ways of identifying R4 with C2, forming a 3 complex dimensional “twistor space”. Complex analysis on this twistor space is what Atiyah claimed should be thought of as the quaternionic analog of complex analysis (on the complex plane).
He reviewed the story of how solutions to various linear equations are related to sheaf cohomology groups on the twistor space, then went on to the non-linear case, where solutions of the anti-self-dual Yang-Mills equations correspond to holomorphic bundles on the twistor space. One can generalize twistor theory to what Atiyah claimed should be thought of as quaternionic analogs of Riemann surfaces: 4d Riemannian manifolds with holonomy in Sp(1)=SU(2), these are self-dual Einstein manifolds, what Penrose would call a “non-linear graviton” (although this is the Riemannian, not pseudo-Riemannian case). The twistor space of these 4d manifolds is a 3d complex manifold, and Atiyah considers complex analysis on this to be the quaternionic analog of complex analysis on a Riemann surface.
The quaternionic analog of higher dimensional complex manifolds are manifolds of dimension 4k, with holonomy Sp(k). Unlike in the complex case, there are few compact examples. Atiyah went on to discuss how examples (mostly non-compact) could be generated as quotients using the quaternionic analog of symplectic reduction. He described several different classes of examples, noting that this construction first appeared in work with physicists studying supersymmetric non-linear sigma models. While I was a post-doc at Stony Brook, Nigel Hitchin was visiting there and working with Martin Rocek and others on this, leading to the 1987 paper in CMP by Hitchin, Karlhede, Lindstrom and Rocek. Atiyah said that he wouldn’t try and describe the relation to supersymmetry, since “I don’t know much about supersymmetry, and if I tried to explain it, you would understand even less”. That Atiyah, after many years of working in this area, still finds supersymmetry to be something he can’t quite understand, is an interesting comment, reflecting the way the subject is still very imperfectly integrated into mathematician’s traditional ways of thinking about geometry and algebra.
Atiyah also commented that off and on over the years he had pursued the idea that quantum groups (which aren’t quite groups), are in some sense the quaternionification of a Lie group (which doesn’t quite exist). He said he hadn’t been successful with this idea, but still thought there was something to it, and hoped that someone else would take up the challenge of trying to make sense of it.
The second wonderful talk I heard was that of Igor Frenkel, from Yale, with the title “Quantum deformation, geometrization, categorification: What is next?”. Unlike Atiyah’s talk, which I pretty much completely understood, Frenkel’s covered much too quickly a lot of material I had never understood, but putting it into an intriguing perspective close to the unsolved problems that seem to me the most important ones for mathematicians and physicists to be looking at. Frenkel began by saying that for many years he had been trying to solve the problem of how to generalize the constructions of representations of loop groups that are related to 2d CFT to representations of 3d gauge groups that should be related to 4d QFT. Some of his thoughts about this are in the write up of his talk at the 1986 ICM. He described himself as having for a long time given up on this problem, moving on to simpler things that he could do: quantum groups which are deformations of the affine Lie algebra story. He went on to talk about “Geometrization”, by which he meant the principle that “all structure constants are Euler characteristics of some variety, all vector spaces are cohomologies”, then “Categorification”, to him the principle that “all structure constants are dimensions of vector spaces, all vector spaces the Grothendieck groups of an Abelian category”. Many of the examples he was using to flesh this out are not well-known to me, I need to do some serious work learning about them before I can say that I clearly understand exactly what he has in mind here.
The last part of his talk, the “What Next?”, went by way too fast but sounded fascinating. He claimed to have some new ways of thinking about the problem of what a representation of these higher dimensional analogs of loop groups should me. I hope to learn more from him in the future to get a better idea of what he has in mind here. He and collaborators at Yale have papers forthcoming, which I look forward to reading. When and if I ever better understand this stuff, I may try and write about it again here.