Uncategorized


This is the second post of my thoughts inspired mainly by reading Fernando Zalamea’s “Synthetic Philosophy of Contemporary Mathematics” (and also a few other sources). The first part is here.

I do have a few issues with the Zalamea book: mainly, as a reader, pinning down what a lot of the sentences really mean can be hard. This might be a combination perfectly reasonable things: the fact that it’s doing philosophy  – and it’s not analytic philosophy, which aspires to math-like rigour. (Indeed, one of the many ideas the book throws around is that of “synthetic philosophy”, modelled not after formal logic, but sheaf theory, with many local points of view and ways of relating them in areas where they overlap. Intuitively appealing, though it’s hard to see how to make it rigorous in the same way.)

So, many of the ideas are still formative, and the terms used to describe them are sometimes new coinages. Then, too, the combination of philosophical jargon and the fact that it’s translated from Spanish probably contribute. So I give the author the benefit of the doubt on this point and interpret the best I can. Even so, it’s still difficult for me to say exactly what some of it is saying. In any case, here I wanted to break down my understanding of some themes it is pointing out. There is more there than I have space to deal with here, but these are some major ones.

I had a somewhat similar response to David Corfield’s book, “Toward a Philosophy of Real Mathematics” (which Zalamea mentions in a chapter where he acknowledges some authors who have engaged the kind of issues he’s interested in). That is, both of them do well at pointing out topics which haven’t received much attention, but the main strength is by pointing out areas of actual mathematical activity, and describing what they’re like (for example, Corfield’s chapter on higher category theory, and Zalamea’s account of Grothendieck’s work). They both feel sort of preliminary, though, in that they’re pointing out areas where a lot more people need to study, argue, and generally thrash out various positions on the issues before (at least as far as I can see) one could hope to say the issues raised have actually been dealt with.

Themes

In any case, part of the point is that for a philosophical take on what mathematicians are actually studying, we need to look at some details. In the previous post I outlined the summary (from philosopher Albert Lautman) of the themes of “Elementary” and “Classical” mathematics. Lautman introduced five themes apropos to the “Modern” period – characterizing what was new compared to the “Classical” (say, pre-1900 or so). Zalamea’s claim, which seems correct to me, is that all of these themes are still present today, but some new ones have been added.

That is, mathematics is cumulative: all the qualities from previous periods stay important, but as it develops, new aspects of mathematics become visible. Thus, Lautman had five points, which are somewhat detailed, but the stand-out points to my mind include:

The existence of a great many different axiomatic systems and theories, which are not reducible to each other, but are related in various ways . Think of the profusion of different algebraic gadgets, such as groups, rings, quandles, magmas, and so on, each of which has its own particular flavour. Whereas Classical mathematics did a lot with, say, the real number system, the Modern period not only invented a bunch of other number systems and algebras, but also a range of different axiom systems for describing different generalizations. The same could be said in analysis: the work on the Calculus in the Classical period leads into the definition of a metric space and then a topological space in the Modern period, and an increasing profusion of specific classes of them with different properties (think of all the various separation axioms, for example, and the web of implications relating them).

The study of a rich class of examples of different axiomatic systems. Here the standout example to me is the classification of the finite groups, where the “semantics” of the classification is much more complex than the “syntax” of the theory. This reminds me of the game of Go (a.k.a. Wei Chi in China, or Baduk in Korea), which has gained some recent fame because of the famous AlphaGo victories. The analogy: that the rules of the game are very simple, but the actual practice of play is very difficult to master, and the variety of examples of games is huge. This is, essentially, because of a combinatorial explosion, and somewhat the same principle is at work in mathematics: the theory of groups has, essentially, just three axioms on one set with three structures (the unit, the inverse, and the multiplication – a 0-ary, unary, and binary operation respectively), so the theory is quite simple. Yet the classification of all the examples is complicated and full of lots of exceptions (like the sporadic simple groups), to the point that it was only finished in Contemporary times. Similar things could be said about topological spaces.

A unity of methods beyond apparent variety. An example cited being the connection between the Galois group of field extensions and the group of deck transformations of a certain kind of branched cover of spaces. In either case, the idea is to study a mathematical object by way of its group of automorphisms over some fixed base object – and in particular to classify intermediate objects by way of the subgroups of this big group. Here, the “base object” could refer to either a sub-field (which is a sub-object in the category of fields) or a base space for the cover (which is not – it’s a quotient, or more generically the target of a projection morphism). These are conceptually different kinds of things on the face of it, but the mechanism of studying “homomorphisms over” them is similar. In fact, following through the comparison reveals a unification, by considering the fields of functions on the spaces: a covering space then has a function field which is an extension of the base case, and the two apparently different situations turn out to correspond exactly.

A “dialectical movement that is a back-and-forth between the One and the Many”. This is one of those jargon-sounding terms (especially the Hegelian-sounding term “dialectic”) and is a bit abstract. The examples given include:

  • The way multiple variants on some idea are thought up, which in turn get unified into a more general theory, which in turn spawns its own variants, and so on. So, as people study different axiom systems for set theory, and their models, this diversity gets unified into the study of the general principles of how such systems all fit together. That is, as “meta-mathematics”, which considers which models satisfy a given theorem, which axioms are required to prove it, etc.
  • The way branches of mathematics (algebra, geometry, analysis, etc.) diverge and develop their own distinct characters, only to be re-unified by mixing them together in new subjects: algebraic geometry, analytic number theory, geometric analysis, etc. intil they again seem like parts of a big interrelated whole. Beyond these obvious cases, the supposedly different sub-disciplines develop distinctive ideas, tools, and methods, but then borrow them from each other as fast as they specialize. This back-and-forth between specialization and cross-fertilization is thus an example of “dialectic”.

Zalamea suggests that in the Contemporary period, all these themes are still present, but that some new ones have become important as well:

Structural Impurity of Arithmetic” – this is related to subjects outside my range of experience, like the Weil Conjectures and the Langlands Program, so I don’t have much to say about it, except to note that, by way of arithmetic functions like zeta functions, they relate number theory to algebraic curves and geometry, and constitute a huge research program that came into being in the Contemporary period (specifically the late 1960’s and early 1970’s). (Edward Frenkel described the Langlands program as “a kind of grand unified theory of mathematics” for this among other reasons.)

Geometrization of Mathematics – essentially, the migration of tools and methods originally developed for like the way topos theory turns logic into a kind of geometry in which the topology of a space provides the algebra of possible truth values. This feeds into the pervasive use of sheaves in modern mathematics, etc. Or, again, the whole field of noncommutative geometry, geometric ideas about space are interpreted as  (necessarily commutative) algebra of functions on that space with pointwise multiplication: differential operators like the Lagrangian, for instance, capture metric geometry, while bundles over a space have an interpretation in terms of modules over the algebra. These geometric concepts can be applied to noncommutative algebras A, thus treating them as if they were spaces.

“Schematization”, and becoming detached from foundations: in particular, the idea that what it means to study, for instance, “groups” is best understood in terms of the properties of the category of groups, and that an equivalent category, where the objects have some different construction, is just as good. You see this especially in the world of n-categories: there are many different definitions for the entities being studied, and there’s increasingly an attitude that we don’t really need to make a specific choice. The “homotopy hypotesis” for \infty-groupoids is an example of this: as long as these form a model of homotopy types, and their collectivity is a “homotopy theory” (with weak equivalences, fibrations, etc.) that’s homotopy-equivalent to the corresponding structure you get from another definition, they are in some sense “the same” idea. The subject of Univalent Foundations makes this very explicit.

Fluxion and Deformation” of the boundaries of some previously fixed subject. “Fluxion” is one of those bits of jargon-sounding words which is suggestive, but I’m not entirely clear if it has a precise measing. This gets at the whole area of deformation theory, quantization (in all its various guises), and so on. That is, what’s happening here is that previously-understood structures which seemed to be discrete come to be understood as points on a continuum. Thus, for instance, we have q-deformation: this starts a bit earlier than the Contemporary period, with the q-integers, which are really power series in a variable q, which just amount to the integers they’re deformations of when q has the value 1. It really takes off later with the whole area of q-deformations of algebra – in which such power series take on the role of the base ring.  Both of these have been studied by people interested in quantum physics, where the parameter q, or the commutators in A are pegged to the Planck constant \hbar.

There’s also reflexivity of modern mathematics, theories applied to themselves. This is another one of those cases where it’s less clear to me what the term is meant to suggest (though examples given include fixed point theorems and classification theorems.)

There’s a list near the beginning of notable mathematicians who illustrate

Zalamea synthesizes these into three big trends described with newly coined terms: “eidal“, “quiddital“, and “archaeal” mathematics. He recognizes these are just convenient rules of thumb for characterizing broad aspects of contemporary research, rather than rigorously definable ideas or subfields. This is a part of the book which I find more opaque than the rest – but essentially the distinction seems to be as follows.

Roughly, eidal mathematics (from the Greek eidos or “idea”) seems to describe the kind that involves moving toward the abstract, and linking apparently unrelated fields of study. One big example referenced here is the Langlands program, which is a bunch of conjectures connecting number theory to geometry. Also under this umbrella he references category theory, especially via Lawvere, which subsumes many different fields into a common framework – each being the study of some particular category such as Top, perhaps by relating it to some other category (such as, in algebraic topology, Grp).

The new term quiddital mathematics (from Latin quidditas, “what exists” or literally “whatness”) appears to refer to the sort which is intimately connected to physics. The way ideas that originate in physics have driven mathematics isn’t totally new: Calculus is a classical example. But more recently, the study of operator algebras was driven by quantum mechanics, index theory which links differential operators and topology was driven by quantum field theory, and there’s a whole plethora of mathematics that has grown out of String Theory, CFT, TQFT, and so forth – which may or may not turn out to be real physics, but were certainly inspired by theorizing about it. And, while it hasn’t had as deep an effect on pure mathematics, as far as I know, I imagine this category would include those areas of study that arose out of other applied studies, such as the theory of networks or the dynamics of large complex systems.

The third new coinage, archaeal mathematics (from arche, or “origin”, also giving the word “archetype”) is another one whose meaning is harder for me to pin down, because the explanation is quite abstract. In the structure of the explanation, this seems to be playing a role that mediates between the first two: something that mediates moving between very abstract notions and concrete realizations of them. One essential idea here is the finding of “invariants”, though what this really seems to mean is more like finding a universal structure of a given type. A simple case might be that between the axioms of groups, and particular examples that show up in practice, we might have various free groups – they’re concrete but pure examples of the theory, and other examples come from imposing more relations on them.

I’m not entirely sure about these three categories, but I do think there’s a good point here. This is that there’s a move away from the specifics and toward general principles. The huge repertoire of “contemparary” mathematics can be sort of overwhelming, and highly detailed. The five themes listed by Lautman, or Zalamea’s additional five, are an attempt to find trends, or deal descriptively with that repertoire. But it’s still, in some ways, a taxonomy: a list of features. Reducing the scheme to these three, whether this particular system makes sense to you or not, is more philosophical: rather than giving a taxonomy, it’s an effort to find underlying reasons why these themes and not others are generating the mathematics we’re doing.  So, while I’m not completely convinced about this system as an account of what contemporary mathematics is about, I do find that thinking about this question sheds light on the mass of current math.

Some Thoughts

In particular, a question that I wonder about, which a project like this might help answer, is the question of whether the mathematics we’re studying today is inevitable. If, as the historical account suggests, mathematics is a time-bound process, we might well ask whether it could have gone differently. Would we expect, say, extraterrestrials, or artificial intelligences, or even human beings in isolated cultures, to discover essentially the same things as ourselves? That is, to what extent is the mathematics we’ve got culturally dependent, and

In Part I, I made an analogy between mathematics and biology, which was mainly meant to suggest why a philosophy of mathematics that goes beyond foundational questions – like the ontology of what mathematical objects are, and the logic of how proof works – is important. That is to say, mathematical questions themselves are worth studying, to see their structure, what kinds of issues they are asking about (as distinct from issues they raise by their mere existence), and so on. The analogy with biology had another face: namely, that what you discover when you ask about the substance of what mathematics looks at is that it evolves over time – in particular, that it’s cumulative. The division of mathematical practice into periods that Zalamea describes in the book (culminating in “Contemporary” mathematics, the current areas of interest) may be arbitrary, but it conveys this progression.

This biological analogy is not in the book, though I doubt it’s original to me. However, it is one that occurs to me when considering the very historically-grounded argument that is there. It’s reminiscent, to my mind, of the question of whether mathematics is “invented or discovered”. We could equally well ask whether evolution “invents” or “discovers” its products. That is, one way of talking about evolution pictures the forms of living things as pre-existing possibilities in some “fitness landscape”, and the historical process of evolving amounts to a walk across the landscape, finding local optima. Increases in the “height” of the fitness function lead, more or less by definition, to higher rates of reproduction, and therefore get reinforced, and so we end up in particular regions of the landscape.

This is a contentious – or at least very simplified – picture, since some would deny that the space of all possibilities could be specified in advance (for example, Lee Smolin and Stuart Kauffman have argued for this view.) But suppose for the moment it’s the case and let’s complete the analogy: we could imagine mathematics, similarly, as a pre-existing space of possibilities, which is explored over time. What corresponds to the “fitness” function is, presumably, whatever it is about a piece of mathematics that makes it interesting or uninteresting, and ensures that people continue to develop it.

I don’t want to get too hung up on this analogy. One of the large-scale features Zalamea finds in contemporary mathematics is precisely one that makes it different from evolution in biology. Namely, while there is a tendency to diversification (the way evolution leads to speciation and an increase in the diversity of species over time), there is also a tendency for ideas in one area of mathematics to find application in another – as if living things had a tendency to observe each other and copy each others’ innovations. Evolution doesn’t work that way, and the reason why not has to do with specifics of exactly how living things evolve: sexual reproduction, and the fact that most organisms no longer transfer genes horizontally across species, but only vertically across generations. The pattern Zalamea points out suggests that, whatever method mathematicians are using to explore the landscape of possible mathematics, it has some very different features. One of which seems to be that it rewards results or concepts in one sub-discipline for which it’s easy to find analogies and apply them into many different areas. This tendency works against what might otherwise be a trend toward rampant diversification.

Still, given this historical outlook, one high-level question would be to try to describe what features make a piece of mathematics more rewarding and encourage its development. We would then expect that over time, the theorems and frameworks that get developed will have more of those properties. This would be for reasons that aren’t so much intrinsic to the nature of mathematics as for historical reasons. Then again, if we had a satisfactory high-level account of what current mathematics is about – something like what the three-way division into eidal, quiddital, and archaeal mathematics is aiming at – that would give us a way to ask whether only certain themes, and only certain subjects and theorems, could really succeed.

I’m not sure how much this gains us within mathematics, but it might tell us how we ought to regard the mathematics we have.

Advertisements

This post – which I’ve split up into parts – is a bit of a departure from talking about the subject matter of mathematical ideas, and more about mathematics in general. In particular, a while ago I was asked a question by a philosopher friend about topology and topos theory as he was trying to understand Alain Badiou’s writings about ontology. That eventually led to my reading a bit more about what recent philosophers have to say about mathematics, or to use it for. This eventually led me to look at Fernando Zalamea’s book “The Synthetic Philosophy of Contemporary Mathematics”. It’s not a new book, unless 2009 counts as new at this point 8 years later. But that’s okay: this isn’t a book review either (though I did find one here). However, it’s the book which was the main jumping off point for the thoughts I’m putting down here. It’s also an interesting book, which speaks to a lot of the same concerns that I’ve been interested in for a while, and while it has some flaws (which I’ll speak to briefly in part II), mostly I want to treat it as a starting point.

I suppose the first issue in talking about the philosophy of mathematics, if your usual audience is for talking about mathematics itself, is justifying why philosophy of mathematics in general ought to be of interest to mathematicians. I’m not sure if this is more, or less, true because I’m not a philosopher, but a mathematician, so my perspective isn’t a very sophisticated reader of the subject, but as someone seeing what it has to say about the field I practice. We mathematicians aren’t the only ones to be skeptical about philosophy and its utility, of course, but there are some particular issues there’s a lot of skepticism about – or at least which lead to a lack of interest.

Why Philosophy Then

My take is that “doing philosophy” is most relevant when you’re trying to carefully and systematically talk about subjects where the concepts that apply are open to doubt, and the principles of reasoning aren’t yet finally defined. This is why philosophers tend to make arguments, challenge each others’ terms, get accused of opacity (in some cases) and so on. It’s also one reason mathematicians tend to be wary of the process, since that’s not a situation we like. The subject matter could be anything, insofar as there are conceptual issues that need clarifying. One result is that, to the extent the project succeeds at pinning down particular concepts and methods, whole areas of philosophy have tended to get reframed under new names: “science”, a more systematic and stable version of the older “natural philosophy”, would be one example. To simplify the history a lot, we could say that by systematically describing something called the “scientific method”, or variations on the theme, science was distinguished from natural philosophy in general. But the thinking that came before this method was described explicitly, which led to its description, was philosophical thinking. The fact that what’s left is necessarily contentious and subject to debate is probably part of why academics in other fields are often dubious about philosophy.

Similarly, there’s the case of logic, which began its life in philosophy as an effort to set down systematic ways of being sure one is thinking rigorously (think Aristotle’s exposition of how syllogisms work). But later on it becomes a topic within mathematics (particularly following Boole turning it into a branch of algebra, which now bears his name). When it comes to philosophy of mathematics in particular, we could say that something similar happened as certain aspects of the topic got formalized and become the field now called “metamathematics” (which studies things such as whether given theorems are provable within specified axiom systems). So one reason philosophy might be important to mathematicians is that the boundary between the two is rather porous. Yet maybe the most common complaint you hear about philosophy is that it seems to have become stuck at just the period when this occurred – around 1900-1940 or so, motivated by things like Hilbert’s program, Cantorian set theory, Whitehead and Russell’s Principia, and Gödel’s theorem. So that boundary seems to have become less permeable.

On the other hand, one of the big takeaways from Zalamea’s book is that the philosophy of mathematics needs to pick up on some of the themes which have appeared in mathematics itself in the “contemporary” period (roughly, since about 1950). So the two fields have a history of exchanging ideas, and potential to keep doing so.

One of these is the sort of thing you see in the context of toposes of sheaves on a site (let’s say it’s a topological space, for definiteness). A sheaf is a kind of object which is defined by what it looks like locally in each open set of the space, which is constrained by having to fit together nicely by gluing on overlaps – with the sheaf condition describing how to pass from local to global. Part of Zalamea’s program is that philosophy of mathematics should embrace this view: it’s the meaning of the word “Synthetic” in the title, as contrasted with “Analytic” philosophy, which is more in the spirit of the foundational approach of breaking down structures into their simplest components. Instead, the position is that lots of different perspectives, each useful for understanding one theme, some part or aspect of mathematics, can be developed, and integrated together by being careful to account for how to reconcile them in areas where they both have something to say. This is another take on the same sort of notion I was inspired by when I chose the name of this blog, so naturally, I’m predisposed to be interested.

Now, maybe it’s not surprising that the boundary between the two areas of thought has been less permeable of late: the part of the 20th century when this seems to have started was also when many fields of academia started to become more specialized as the body of knowledge in each became so huge that it was all anyone could do to master one special discipline. (One might also point to things like the way science became a much bigger group enterprise, as witness things like the Manhattan Project, which led into big government-funded agencies doing “big science” in an institutional setting where specialization and the division of labour was the norm. But that’s a big digression and probably needs more data to support it than I’ve actually got.)

Anyway, Whitehead and Russell’s work seems to have been the last time a couple of philosophers famously made a direct contribution to mathematics. There, the point was to nail down a definite answer to the question of how we know truth, what mathematical entities are, how logic functions and gives rise to more complex mathematics, and so on. When Gödel, working as a mathematician, showed how it was incomplete, and was construed as doing mathematics (and if you read his paper, it’s hard to construe it as much else), that probably contributed a lot to mathematicians drifting away from philosophers, many of whom continued to be interested in the same questions.

Big and Small Scales

Still, even if we just take set-theoretic foundations for granted (which is no longer quite as universal as it used to be), there’s a distinction here. Just because mathematics can be reduced to set theory and logic, this doesn’t mean that the philosophy of mathematics has to reduce to the philosophy of set theory and logic. Whatever the underlying foundations, an account of what happens at large scales might have very different features. Someone with physics inclinations might describe it as characterizing the “effective theory” of mathematics – the way fluid dynamics is an effective theory of particular kinds of big statistical ensembles of atoms, and there are interesting things to say about each level in its own right.

Another analogy that occurs to me is with biology. Suppose we accept that biology ultimately reduces to chemistry, in the sense that all living things are made of chemicals, which behave exactly as a thorough understanding of chemistry would say they do. This doesn’t imply that, in thinking about biology, there’s nothing to think about except chemistry: the philosophy of biology would have its own issues, because biology entails new concepts, regardless of whether there happens to be some non-chemical “vital fluid” that makes living things different from non-living things. To say that there is no such vital fluid is an early, foundational, part of the philosophy of biology, in this analogy. It doesn’t exhaust what there is to say about living things, though, or imply that one should just fall back on the philosophy of chemistry. A big-picture consideration of biology would have to take into account all sorts of knowledge and ideas that get used in the field.

The mechanism of evolution, for example, doesn’t depend on the thermodynamic foundations of life: it can be applied to processes based on all sorts of different substrates – which is how it could inspire the concept of genetic algorithms. Similarly, the understanding of ecosystems in terms of complex systems – starting with simple situations like predator-prey models and building up from there – doesn’t depend at all on what kind of organisms are involved, or even that they are living things. Both of these are bodies of knowledge, concepts, and methods of analysis, that play a big role in studying living things, but that aren’t related at all to the foundational questions of what kind of physical implementation they have. Someone thinking through the concepts in the field would have to take them into account and take into account their own internal logic.

The situation with mathematics is similar: high-level accounts of what kinds of ideas have an influence on mathematical practice could be meaningful no matter what context they appear in. In fact, one of the most valuable things a non-rigorous approach – that of philosophy rather than, say, metamathematics as such – has to offer is that it can comment when the same themes show up even in formally very different sub-disciplines within mathematics. Recognizing these sorts of themes, before they can be formalized and made completely precise, is part of describing what mathematicians are up to, and what the significant features of that practice may be. Discovering those features, and hopefully pinning them down enough to get one or more ways to formalize them that are rigorous to use, is one of the jobs philosophy ought to be able to do. Zalamea suggests a few such broad patterns, which I’ll try to unpack and comment on a little in Part II of this post.

 

Historical Change

Even granted the foundational questions about mathematics, there are still distinctive features of what people researching it today are doing which are part of the broader picture. This leads into the distinction which Zalamea makes between the different characteristics of the particular mathematics current in different periods. Part of the claim in the book is exactly that this distinction isn’t only an arbitrary division of the timeline. Rather, the claim is that what mathematicians generally are doing at any given time has some recognizable broad features, and these have changed over time, depending on what were seen as the interesting problems and productive methods.

One reason mathematicians may have tended to be skeptical of philosophy (beyond the generic reasons) is that by focusing on the typical problems and methods of work of the “Modern” period, it hasn’t had much to say about the general features that actually come up in contemporary work. David Corfield made a similar argument in “Toward a Philosophy of Real Mathematics”, where “real” meant something like what Zalamea calls “contemporary”: namely, what mathematicians are actually doing these days.

This outlook suggests that, just as art has evolved as new ideas are created and brought into the common practice, so has mathematics. It contrasts with the usual way mathematicians think of themselves as discovering and exploring truths rather than creating the way artists do. It probably doesn’t have to be: the continents are effectively eternal in comparison to human time, but different people have come across them and explored them at different times. Since the landscape of possible mathematics is huge, and merely choosing a direction in which to explore and by what methods (in the analogy, perhaps the difference between boating on a river and walking overland) has a creative aspect to it, the distinction is a bit hazier. It does put the emphasis on the human side of that historical process rather than the eternal part (already a philosophical stance, but a reasonable one). Even if the periodization is a bit arbitrary, it’s a way of highlighting some changes over time, and making clear why there might be new trends that need some specific attention.

Thus, we start with “Elementary” mathematics – the kind practiced in antiquity, up through about the time of invention of Calculus. The mathematics in this period was closely connected to the familar world: geometry as a way to talk about space, arithmetic and algebra as tools for manipulating numbers, and so forth. There were plenty of links to applications that could easily be understood as being about the everyday world – solving polynomial equations, for example, amounts to finding quantities that have special properties with respect to some fairly straightforward computation that can be done with them. Classical straightedge-and-compass constructions in geometry give a formal, idealized way to talk about what can be more-or-less well approximated by literal physical operations. “Elementary” doesn’t necessarily mean simple: there are very complex bits of mathematics in these areas, but the building blocks are fairly simple elements. Still, in this period, it was possible to think of mathematics itself as a kind of philosophy of real things – abstracting out ideal properties and analyzing them, devising rules of logic and calculation, and so on. The sort of latent Platonism of a lot of mathematical thinking – that these abstract forms are an underlying reality of particular physical things –

Then “Classical” (the period when Leibniz, Euler, Gauss, et. al.) when mathematics was still a fairly unified field, but with new methods, like the rigorous use of infinite processes. It’s also a period when mathematics itself begins to generate more conceptual issues that needed to be talked about in external language.Think of the controversy over the invention of Calculus, and the use of infinitesimals in derivatives and integrals, the notion that an infinite series might converge, and so on. Purely mathematical solutions were the answer, but arose only after there’d been an outside-the-formalism discussion about the conceptual problems with infinitesimals. This move away from elements that directly and obviously corresponded to real things was fruitful, though, and not only because it led to useful tools like Calculus. Also because that very fact of thinking about idealized or abstract entities opened up many of the areas of mathematics that came later, and because trying to justify these methods against objections led to refinements like the concept of a limit, which led into analytical arguments with “epsilons” and “deltas”, and more sophisticated use of quantifiers like the “for all” and “there exists”. Refining this language opened up combinatorial building-blocks of all sorts of abstract concepts.

This leads into the “Modern” period (about 1850-1950), where people became concerned with structure, axiomatization, foundational questions. This move was in part a response to the explosion of general concepts which those same combinatorial building blocks encouraged. Particular examples of, for instance, groups may very well have lots of practical applications, but here we start to see the study of the abstract concept of a group as such, proof of formal theorems about it, and so on. In algebra, Jordan and Cayley formally set out the axioms of rings, groups, fields, etc. which people had been studying for some time in a less explicit way (as, for instance, in Galois theory). The systematization of geometry by Klein, Riemann, Cartan, and so forth, was similar: particular geometries may well have physical relevance, or be interesting as examples, but by systematizing and proving general theorems, it’s the abstractions themselves that become the real objects of study for mathematicians as such.

As a repertoire of such concepts started to accumulate, the foundational questions became important: if the actual entities mathematicians were paying attention to were not the elementary ones, but these new abstractions, people were questioning what, in ontological terms, those things actually were. This is where the investigation topics like the relation of set theory to logic, the existence of set-theoretic models of formal theories, the relation between provability of theorems and the existence of models with particular properties, consistency of axioms, and so on, came to the forefront.

Zalamea’s book starts with an outline of Lautman’s description of five big themes in mathematics that became prominent in the Modern period, and then extends them into the “Contemporary” period (roughly, after 1950) by saying that all the same trends continue, but a bunch of new ones appear.  One of these is precisely a move away from a focus on the specific foundational description of a structure – in categorical language, we’ve tended to focus less on the set-theoretic details of a structure than on features that are invariant under isomorphisms that change all of that. But this gets into a discussion I’ll save for the second part of this post.

For now, I’ll say just a couple of things about this approach to wrap up this part. I have some doubts about the notion that the particular historical evolution of which themes mathematics is exploring represent truly different “kinds” of math, but that’s not really the claim. What seems true to me is that, even in what I described above, you can see how spending time exploring one issue generates subject matter that becomes a concern for the next. Mathematicians on another planet, or even here if we could somehow run through history again, might have taken different paths and developed different themes – or then again, this sequence might have been as necessary as any logical deduction. This is a lot harder to see, though the former seems more natural to my mind. Highlighting the historical process of how it happened does, at least, help to throw some light on some of the big features of what we’ve been discovering. Zalamea’s book (which, again, I’ll come back to) makes a particular attempt to do so by suggesting three main kinds of contemporary math (with their own neologisms describing them). Whatever you think of the details of this, I think it makes a strong case that looking at these changes over time can reveal something important.

 

 

This is a summary of talks at the conference in Lisbon, continuing from the previous post. The ones I classified under “Field Theory” were a subjective choice, and the categories I list here are even more so, but I think they roughly summarize some of the big themes. I’m hoping to get back to posting here somewhat often, maybe with a wider variety of topics – but for now, this seems like a good start.

Infinity-Categorical Structures

Simona Paoli gave an overview of infinity-categories generallycalled “Segal-Type Models of Higher Categories“, which was based on her recent monograph of the same title. The talk, which basically summarizes the first chapter in the monograph that lays out the groundwork, described the state of the art on various kinds of higher categories constructed by simplicial methods (like the definitions of Tamsamani and Simpson, etcetera). Since I discussed this at some length back when I was describing the seminar on this subject we did at Hamburg, I’ll just say that Simona’s talk was a nice summary of the big themes we looked at there.

The first talk of the conference was by Dave Carchedi, called “Higher Orbifolds as Structured Infinity-Topoi” (it was a board talk, so there are no slides, but it appears to be based on this paper). There was some background about what “higher orbifolds” might be – to begin with, looking at the link between orbifolds and toposes. The basic idea, as I understand it, being that you can think of nice toposes as categories of sheaves over groupoids, and if the toposes have some good properties and extra structure – like a commutative ring object – you can think of them as being like the sheaves of functions over an orbifold. The commutative ring object is then the structure sheaf for the orbifold, thought of as a ringed space. In fact, you may as well just say such a topos with this structure is exactly what you mean by an orbifold, since there’s a simple correspondence. The way to say this is that orbifolds “are” just Étale stacks. (“Étale”, of a groupoid, means that the source map from morphisms to objects is a local homeomorphism – basically, a sheeted cover. An Étale stack is one presented by an Étale groupoid.)

So then the idea is that a “higher orbifold” should be a gadget that has a similar relation to higher toposes: Étale \infty-stacks. One interesting thing about \infty-toposes is that the totality of them forms an \infty-topos itself. The novel part here is showing that the \infty-topos of higher orbifolds also, itself, has the same properties – in particular, a universal structure sheaf called \theta_U. This means that it is, in itself, an orbifold! (Someone raised size concerns here: obviously, a category which is one of its own objects presents foundational issues. So you do have to restrict to orbifolds in a universe of some given maximum size, and then you get that the \infty-category of them is itself an orbifold – but in a larger universe. However, the main issue is that there’s a universal structure sheaf which gives the corresponding structure to the category of all such objects.)

Categorification

There was one by Imma Galvez-Carillo, called “Restriction Species“, which talks about categorifying linear algebra, and in particular coalgebra. The idea of using combinatorial species for this purpose has been around for a while – it takes the ideas of Baez and Dolan on groupoid cardinality and linear functors (which I’ve written about plenty in this blog and elsewhere). Here, the move beyond that is to the world of \infty-groupoids, using simplicial models. A lot of the same ideas appear: slice categories \mathbf{Gpd}/B over the groupoid of finite sets and bijections B \in \mathbf{Gpd} can be seen as generalizing vector spaces with a specified basis (consisting of the cardinalities 0, 1, 2, … of the finite sets). Individual maps into B are “species”, and play the role of vectors. The whole apparatus of groupoidification gives a way to understand this: the groupoid cardinality of the fibre over each n becomes the component of a vector, and so on. Spans are then linear maps, since composition using the fibre product has the same structure as matrix multiplication. This talk considered an \infty-groupoid version of this idea – groupoid cardinality generalizes to a kind of Euler characteristic – and talked about what incidence coalgebras look like in such a context. Another generalization is that decomposition spaces – related to restriction species, which are presheaves on I, the category (no longer a groupoid) of finite sets and injections, which carry information about how structures “restrict” along injections. The talk discussed how this gives rise to coalgebras. An example of this would be the Connes-Kreimer bialgebra, whose elements are forests. It turns out this talk just touched on one part of a large project with Joachim Kock and Andrew Tonks – the most obviously relevant references here being this, on the categorified concept called “homotopy linear algebra” involved, and this, about restriction species and decomposition spaces.

One by Vanessa Miemietz on 2-Representations of finitary 2-Categories also tied into the question of algebraic categorification (Miemietz is a collaborator of Mazorchuk, who wrote these notes on that topic). The idea here is to describe monoidal categories which are sufficiently algebra-like to carry an interesting representation theory that’s a good 2-categorical analog for the usual kind for Lie algebras, and then develop that theory. These turn out to be “FIAT” categories (an acronym for “finitary, involution, adjunction, two-category”, which summarizes the kind of structures they need to have). This talk developed some of this theory, including an analog of the Artin-Wedderburn Theorem (which says that all reasonably nice rings are essentially just sums of matrix rings over division algebras), and used that to talk about the representation theory of FIAT categories.

Christian Blohmann spoke about “Morita Equvialence of Higher Geometric Groupoids”. The basic idea was to generalize, to \infty-stacks, the correspondence between three different definitions of principle G-bundles: in terms of local trivializations and gluing functions; in terms of a bundle over X with a free, proper action of G; and in terms of classifying maps X \rightarrow BG. The first corresponds to a picture involving anafunctors and a complex of n-fold intersections U_i \cap U_j \cap U_k and so on; the third generalizes naturally by simply taking BG to by any space, not just a homotopy 1-type. The talk concentrated on the middle term of the three. A big part of it amounted to coming up with a suitable definition for a “principal” action of an \infty-group: it’s this which turns out to involve Morita equivalences, since one doesn’t necessarily want to insist that the action gives strict isomorphisms, but only Morita equivalence.

A talk I didn’t follow very well, but seemed potentially pretty interesting, was Charles Cascauberta’s “Homotopy Algebras vs. Algebras up to Homotopy“. This involved the relation between the operations of taking algebras of a monad T in a model category $M$, and taking the homotopy category. The question has to do with the relation between the two possible orders in which this can be done, and in particular the fact that the two orders give different results.

Topology and Geometry

Ronnie Brown gave a talk called “Homotopical Excision“, which surveyed some of the ways crossed modules and higher structures can be used in topology. As with a lot of Ronnie Brown’s surveys, this starts with the groupoid version of the van Kampen theorem, but grows from there. Excision is about relative homotopy groups of spaces X with distinguished subspaces A. In particular, this talks about unions of those spaces. As one starts taking unions, crossed modules come into the picture, and then higher crossed structures: crossed modules OF crossed modules (which are squares of groups satisfying a bunch of properties), and analogous structures that take the shape of n-cubes. There’s a lot of background, so check out the slides, or other work by Brown and others if you’re interested.

Manuel Barenz gave a talk called “Extending the Crane-Yetter Model” which talked about manifold invariants. There’s a lot of categorical machinery used in building these invariants, which uses various kinds of string diagrams: particular sorts of monoidal categories with some nice properties let you interpret knot-like diagrams as morphisms. For example, you need to be able to interpret a bend in which an upward-oriented strand turns around and becomes downward-oriented. There is a whole zoo of different kinds of monoidal category, each giving rise to a language of string diagrams with its own allowed operations. In this example, several different properties come up, but the essential one is pivotality, which says that there is a kind of duality for which this bend is interpreted as the morphism which pairs an object with its dual. If your category is enriched in vector spaces, a knot or link ends up giving you a complex number to compute (a morphism from the identity object to itself). The “string net space” for a manifold amounts to a space spanned by all the ways of embedding this type of graph in the manifold. Part of what this talk speaks to is the idea that such a construction can give the same state space as the Crane-Yetter (originally constructed as a state-sum invariant, based on a totally different construction).

For 4-manifolds, the idea is then that you can produce diagrams like this using the Kirby calculus, which is a way of summarizing a decomposition of the manifold into handle-bodies (the diagrams arise from marking where handles are attached to a 3-sphere boundary of a 4-disk). These diagrams can be transformed, because different handle-body decompositions can be deformed into each other by handle-slides and so forth. So part of the issue in creating an invariant is to identify just what kind of monoidal categories, and what kind of labellings, has just the kind of allowable moves to get along with the moves allowed in the Kirby calculus, and so ensure that the resulting diagrams actually give the same value. This type of category will then naturally be what you want to describe 4-manifold invariants.

Other talks:

Here are a few talks which, I must admit, went either above my head, or are outside my range of skill to summarize adequately, or in some cases just went by too quickly for me to take adequate notes on, but which some readers might be interested to know about…

Some physics-related talks:

Christian Saemann gave an interesting talk about the relation between the self-dual string in string theory and higher gauge theory using the string 2-group, which is a sort of natural 2-group analog of the spin group Spin(n), and for that reason is surely bound to be at least mathematically important. Martin Wolf’s talk, “Super Yang-Mills Theory from Higher Chern-Simons Theory“, which relates the particular 6-dimensional chiral, superconformal field theory SYM to some combination of twistor geometry with the geometry of gerbes (or “categorified principal bundles”). Branislav Jurco spoke about “Homological Perturbation, Minimal Models, and Effective Actions”, which involved effective actions in the Lagrangian formulation of quantum theories to some higher-algebraic gadgets, such as homotopy algebras.

Domenico Fiorenza’s talk on T-duality in rational homotopy theory seemed interesting (in particular, it touched on the Fourier transform as a special case of the “pull-push” construction which I’m very interested in), but I will have to think about this way of talking about it before I could give a good summary. Perhaps reading the associated paper would be a good start.

Operads

Operads aren’t really my specialty. The general idea is that they formalize the situation of having operations taking variable numbers of inputs to a given output and describe the structure of the situation. There are many variations which describe possible conditions which can apply, and “algebras” for an operad are actual implementations of such a structure on particular spaces with particular operations. The current theory is rather more advanced, though, and in particular \infty-operads seem to be under lots of development right now.

There was a talk by Hongyi Chu on “Enriched Infinity-Operads“, which described how to give a categorification of the notion of “operad” to something which is only homotopy-coherent. Philip Hackney’s talk, “Homotopy Theory of Segal Cyclic Operads” likewise used simplicial presheaves to talk about operads having the property, “cyclic”, which allows inputs to be “rotated” into outputs and vice versa in a particular way.

Other

Andrew Tonks’ talk “Tilings, Trees, DG2A’s and B_{\infty}-algebras” (DG2A’s stands for “differential graded 2-algebras”) was quite interesting to me, partly because of the nice combinatorial correspondence it used between certain special kinds of tile-arrangements and particular kinds of trees with coloured nodes. These form elements of those 2-algebras in question, and a lot of it involved describing the combinatorial operations that correspond to composition, the differential, and other algebraic structures.

Johannes Hubschmann’s talk, “Multi-derivative Maurer-Cartan Algebras and Lie-Reinhart Algebras” used a lot of algebraic machinery I’m not familiar with, but essentially the idea is that these are algebras with derivations on them, and a “higher structure” version of such an algebra will have several different derivatives in a coherent way.

Ahmad al-Yasry spoke on “Graph Homologies and Functoriality“, which talked about some work which seems to be closely related to span constructions I’m interested in, and bicategories. In this case, the spans are of manifolds with embedded graphs, and they need to be special branched coverings. This importantly geometric setup is probably one reason I’m less comfortable with this than I’d like to be (considering that Masoud Khalkhali and I spent some time discussing a related paper back when I was at U of Western Ontario. This feeds somehow into the idea – popular in noncommutative geometry – of the Tomita flow, a kind of time-evolution that naturally appears on certain algebras. In this case, there’s a bicategory, and correspondingly two different time evolutions – horizontal and vertical.

 

So those are the talks at HSL-2017, as filtered through my point of view. I hope to come back in a while with more new posts.

Update

This blog has been on hiatus for a while. I’ve spent the past few years in several short-term jobs which were more teaching-heavy than the research postdocs I was working in when I started it, so a lot of my time went to a combination of teaching and applying for jobs. I know: it’s a common story.

However, as of a year ago, I’m now in a tenure-track position at SUNY Buffalo State College, in the Mathematics Department. Given the academic job market is these days, I feel pretty lucky to find such a job, and especially since it’s a supportive department with colleagues I get along with well. It’s a relatively teaching-oriented position, but they’re supportive of research too, so I’m hoping I’ll get back to updating the blog semi-regularly.

In particular, since I’ve been here I’ve been able to get out to a couple of conferences, and I’d like to take a little time to make a post about the most recent. The first I went to was the Union College Mathematics Conference, in Schenectady, here in New York state. The second was Higher Structures in Lisbon. I was able to spend some time there talking with Roger Picken, about our ongoing series of papers, and John Huerta about a variety of stuff, before the conference, which was really enjoyable.

Here’s the group picture of the participants:hs-lisbon-2017

The talks from the conference that had slides are all linked to from their abstracts page, but there are a few talks I’d like to comment on further. Mine was similar to talks I’ve described here in the past, about transformation structures and higher gauge theory. Hopefully there will be an arXiv paper reasonably soon, so I’ll pass over that for now. I’ll summarize what I can, though, focusing on the ones that are particularly interesting (or comprehensible) to me. I’ve linked to the slides, where available (some were whiteboard talks). I’ve grouped them into different topics. This post summarizes talks that fall under the general category of “field theory”, while the others will be in a follow-up post.

Field Theory

One popular motivation for the use of “higher structures” is field theory, in its various forms. This makes sense: most modern physical theories are of this kind, one way or another, and physics is a major motivation for math. Specifically, one of the driving ideas is that when increasing the dimension of the theory, concepts which are best expressed with categories in low dimensions need higher n-categories to express them in higher categories – we see this in fully-extended TQFT’s, for instance, but also the idea that to express the homotopy n-type of a space (what you want, generally, for an n-dimensional space), you need an n-groupoid as a model. There are some other situations where they become useful, but this is an important one.

Ana Ros Camacho was a doctoral student with Ingo Runkel in Hamburg while I was a postdoc there, so I’ve seen her talk about her research several times (thesis). This talk, “Toward a Higher-Categorical Statement for the Landau-Ginzburg/CFT Correspondence”, was maybe the clearest overview I’ve seen so far, so this was a highlight for me. Essentially, it’s a fact of long standing that there’s a correspondence between 2D rational conformal field theory and the Landau-Ginzburg model – a certain Sigma-model (field theory where the fields are maps into some classifying space) characterized by a potential. The idea was that there’s some evidence for a conjecture (but not yet a proof that turns it into a theorem) which says that this correspondence comes from some sort of relationship – yet to be defined precisely – between two monoidal categories.

One is a category of matrix factorizations, and the other is a category which comes from representations of a vertex operator algebra \mathcal{V} associated with the CFT’s. Matrix factorizations work like this: start with the polynomial ring S = k[x_1,x_2,\dots,x_n], and pick a polynomial W \in S. If the dimension of the quotient ring of S by all the derivatives of W is finite-dimensional, it’s a “potential”.

This last condition is what makes it possible to talk about a “matrix factorization” of (S,W), which consists of (M,d), where M = M_0 \oplus M_1 is a free \mathbb{Z}_2-graded S-module, and d : M \rightarrow M is a “twisted differential” – an S-linear map in degree 1 (meaning it takes M_0 to M_1 and vice versa) such that d^2 = W \cdot Id_M. (That is, the differential is a kind of “square root” of the potential, in this special degree-1 sense.) There is a whole bicategory of such matrix factorizations, called LG (for “Landau-Ginzburg”). Its objects are algebras with a potential, (S,W). The morphisms from (S_1,W_1) to (S_2,W_2) are matrix factorizations for (S_1 \otimes S_2, W_1 - W_2) (which can be defined in a natural way), which can be composed by a kind of tensor product of modules, and the 2-morphisms are just bimodule maps.

The notion, then, is that this 2-category LG is supposed to be related in some fashion to a category Rep(\mathcal{V}) of representations of some vertex algebra associated to a CFT. There are some partial results to the effect that there are monoidal equivalences between certain subcategories of these in particular cases (namely, for special potentials W). The hope is that this relationship can be expanded to explain the known relationship between the two sorts of field theory.

Tim Porter talked about “HQFT’s and Beyond” – which I’ll skimp on here mainly because I’ve written about a similar talk in a previous post. It did get into some newer ideas, such as generalizing defect-TQFT’s to HQFT and more.

Nils Carqueville gave a couple of talks – one for himself, and one for Catherine Meusburger, who had to cancel – on some joint work of theirs. One was “3D Defect TQFT’s and their Tricategories“, and the other “Orbifolds of Defect TQFT’s“. This is a use of “orbifold” that I don’t entirely understand, but I think roughly the idea is that an “orbifold completion” of a category is an extension in the same way that the category of orbifolds extends that of manifolds, and it’s connected to the idea of equivariantization – addressing symmetry.

In any case, what it’s applied to here is the notion of TQFT’s which are defined, not on just categories of manifolds with cobordisms as the morphisms, but something more general, where all of these spaces are allowed to have “defects”: embedded submanifolds of lower dimension, which can meet at still lower-dimensional junctions, and so on. The term suggests, say, a crystal in solid-state physics, where two different crystal structures meet at a “defect” plane. In defect TQFT, one has, essentially, one TQFT living on one side of the defect, and another on the other side. Then the “tricategories” in question have objects assigned to regions, morphisms to defects where regions meet, and so on (thus, this is a 3D theory). A typical case will have monoidal categories as objects, bimodule categories as morphisms, and then functors and natural transformations. The monoidal categories might be, say, representation categories for some groupoid, which is what you’ll see if the theory on each region is a gauge theory. But the formalism makes sense in a much broader situation. A later talk by Daniel Scherl addressed just such a case (the tricategory of bimodule categories) and the orbifold completion construction.

Dmitri Pavlov’s “Extended QFT’s are Local” was structured around explaining and one main theorem (and the point of view that gives it a context): that field theories FT^G_V(T) : Man^{op} \rightarrow sSet, which is to say covariant functors which take manifolds into simplicial sets (or, more generally, some other model of \infty-groupoids) have a particular kind of structure. This amounts to showing that being a field theory requires that it should have some properties. First, it should be a local theory: this amounts to the functor being a sheaf, or stack (that is, there are the usual gluing conditions which relate the \infty-groupoids$ assigned to overlapping neighborhoods, and their union). Next, that there should be a classifying object \mathcal{E}FT^G_V in simplicial sets so that, up to homotopy, there’s an equivalence between concordance classes of fields (which might be, say, connections on bundles, or geometric structures, or various other things) and maps into the classifying space. Then, that this classifying space can be built as a homotopy colimit in a particular way. This theorem seems like a snazzier version of the Brown Representability Theorem, which roughly says that functor satisfying some nice axioms making it somewhat like a cohomology theory (now extended to specify a “field theory” in a more physics-compatible sense) has a classifying object. The talk finished by giving examples of what the classifying object looks like for, say, the theory of vector bundles with connection, for the Stolz-Teichner theory, etc.

In a similar spirit, Alexander Schenkel’s “Towards Homotopical Algebraic QFT” is an efford to extend the formalism of Algebraic QFT (developed by people such as Roberts, and Haag) to an \infty-categorical – or homotopical – situation. The idea behind AQFT was that such a field theory would be a functor F : Loc \rightarrow Alg, which takes some category of spacetimes to a category of algebras, which are supposed to be the algebra of operators on the fields on that bit of spacetime. Then breaking down spacetime into regions, you get a net of algebras that fit together in a particular way. The axioms for AQFT say things like: the algebras for two spacelike-separated regions of space should commute with each other (as subalgebras inside the one associated to a larger region containing both). This gets at the idea that the theory is causal – acting on one region doesn’t affect the other, if there’s no timelike path from one to the other. The other conditions say that when one region is embedded in another, the algebra is also embedded; and that if a small region contains a Cauchy surface for a larger region, the two algebras are actually isomorphic (i.e. having a Cauchy surface determines the whole region). These regions get patched together by local-to-global gluing condition which makes the functor into a cosheaf (not a sheaf: it’s covariant because in general bigger regions have bigger algebras of observables). The problem was that this framework is not enough to account for things like gauge theories, essentially because the gluing has some flexibility up to gauge equivalence. So the talk describes how to extend the framework of AQFT to homotopical algebra so that the local-to-global gluing condition is a homotopy sheaf condition, and went on to talk about what such a theory looks like in some detail, including the extension to categories of structured spacetimes (in somewhat the same vein as HQFT mentioned above).

Stanislaw Szawiel spoke about “Categories of Physical Processes“, which was motivated by describing this as a “non-topological TQFT”. That is, like the Atiyah approach to TQFT, it uses a formalism of categories and functors into some category of algebras to describe various physical systems. Rather than specifically the category of bordisms used in TQFT, the precise category Phys being used depends on what system one wants to model. But functors into *Mod, of C^*-algebras and bimodules, are seen as assigning algebraic data to physical content. There are a lot of details out of the theory of C^*-algebras, such as the GNS theorem, unitarity, and more which come into play here, which I won’t attempt to summarize. It’s interesting, though, that a bunch of different physical systems can be described with this formalism: classical Markov processes, particle scattering, and so forth. One of the main motivations seemed to be to give a language for dealing with the “Penrose Problem”, where evolution of spacetime is speculated to be dynamically related to “state vector collapse” in quantum gravity.

Theo Johnstone-Freyd’s talk on “The Moonshine Anomaly” succeeded in getting me interested in the Monster group and its relation to CFT. He did mention a couple of recent papers that calculate some elements of the fourth cohomology of the super-sized sporadic groups C_0 and M (the Monster) which have interesting properties, and then proceeded to explain what this means. That explanation pulls in how these groups relate to the Leech Lattice – a 24-dimensional lattice with nice properties, of which they’re symmetry groups. This relates to CFT, since these are theories where the algebra of observables is a certain chiral algebra (typically described as a vertex algebra). The idea, as I understand it, is that the groups act as symmetries on some such operator, and a “gauged” or “orbifolded” theory (a longstanding idea, which is described here) ends up being related to the category of twisted representations of the group G. The “twisting” requires a cohomology class (which is the – nontrivial – associated of that category), and this class is what’s called the “anomaly” of the theory, which gets used in the Lagrangian action for this CFT. So the calculation of that anomaly in the papers above – an element of the Monster group’s fourth cohomology – also helps get a handle on the action of the corresponding CFT.

(More talks to come in part II)

So apparently the “Integral” gamma-ray observatory has put some pretty strong limits on predictions of a “grain size” for spacetime, like in Loop Quantum Gravity, or other theories predicting similar violations of Lorentz invariants which would be detectable in higher- and lower-energy photons coming from distant sources.  (Original paper.)  I didn’t actually hear much about such predictions when I was the conference “Quantum Theory and Gravitation” last month in Zurich, though partly that was because it was focused on bringing together people from a variety of different approaches , so the Loop QG and String Theory camps were smaller than at some other conferences on the same subject.  It was a pretty interesting conference, however (many of the slides and such material can be found here).  As one of the organizers, Jürg Fröhlich, observed in his concluding remarks, it gave grounds for optimism about physics, in that it was clear that we’re nowhere near understanding everything about the universe.  Which seems like a good attitude to have to the situation – and it informs good questions: he asked questions in many of the talks that went right to the heart of the most problematic things about each approach.

Often after attending a conference like that, I’d take the time to do a blog about all the talks – which I was tempted to do, but I’ve been busy with things I missed while I was away, and now it’s been quite a while.  I will probably come back at some point and think about the subject of conformal nets, because there were some interesting talks by Andre Henriques at one workshop I was at, and another by Roberto Longo at this one, which together got me interested in this subject.  But that’s not what I’m going to write about this time.

This time, I want to talk about a different kind of topic.  Talking  in Zurich with various people – John Barrett, John Baez, Laurent Freidel, Derek Wise, and some others, on and off – we kept coming back to kept coming back to various seemingly strange algebraic structures.  One such structure is a “loop“, also known (maybe less confusingly) as a “quasigroup” (in fact, a loop is a quasigroup with a unit).  This was especially confusing, because we were talking about these gadgets in the context of gauge theory, where you might want to think about assigning an element of one as the holonomy around a LOOP in spacetime.  Limitations of the written medium being what they are, I’ll just avoid the problem and say “quasigroup” henceforth, although actually I tend to use “loop” when I’m speaking.

The axioms for a quasigroup look just like the axioms for a group, except that the axiom of associativity is missing.  That is, it’s a set with a “multiplication” operation, and each element x has a left and a right inverse, called x^{\lambda} and x^{\rho}.  (I’m also assuming the quasigroup is unital from here on in).  Of course, in a group (which is a special kind of quasigroup where associativity holds), you can use associativity to prove that x^{\lambda} = x^{\rho}, but we don’t assume it’s true in a quasigroup.  Of course, you can consider the special case where it IS true: this is a “quasigroup with two-sided inverse”, which is a weaker assumption than associativity.

In fact, this is an example of a kind of question one often asks about quasigroups: what are some extra properties we can suppose which, if they hold for a quasigroup Q, make life easier?  Associativity is a strong condition to ask, and gives the special case of a group, which is a pretty well-understood area.  So mostly one looks for something weaker than associativity.  Probably the most well-known, among people who know about such things, is the Moufang axiom, named after Ruth Moufang, who did a lot of the pioneering work studying quasigroups.

There are several equivalent ways to state the Moufang axiom, but a nice one is:

y(x(yz)) = ((yx)y)z

Which you could derive from the associative law if you had it, but which doesn’t imply associativity.   With associators, one can go from a fully-right-bracketed to a fully-left-bracketed product of four things: w(x(yz)) \rightarrow (wx)(yz) \rightarrow ((wx)y)z.  There’s no associator here (a quasigroup is a set, not a category – though categorifying this stuff may be a nice thing to try), but the Moufang axiom says this is an equation when w=y.  One might think of the stronger condition that says it’s true for all (w,x,y,z), but the Moufang axiom turns out to be the more handy one.

One way this is so is found in the division algebras.  A division algebra is a (say, real) vector space with a multiplication for which there’s an identity and a notion of division – that is, an inverse for nonzero elements.  We can generalize this enough that we allow different left and right inverses, but in any case, even if we relax this (and the assumption of associativity), it’s a well-known theorem that there are still only four finite dimensional ones.  Namely, they are \mathbb{R}, \mathbb{C}, \mathbb{H}, and \mathbb{O}: the real numbers, complex numbers, quaternions, and octonions, with real dimensions 1, 2, 4, and 8 respectively.

So the pattern goes like this.  The first two, \mathbb{R} and \mathbb{C}, are commutative and associative.  The quaternions \mathbb{H} are noncommutative, but still associative.  The octonions \mathbb{O} are neither commutative nor associative.  They also don’t satisfy that stronger axiom w(x(yz)) = ((wx)y)z.  However, the octonions do satisfy the Moufang axiom.  In each case, you can get a quasigroup by taking the nonzero elements – or, using the fact that there’s a norm around in the usual way of presenting these algebras, the elements of unit norm.  The unit quaternions, in fact, form a group – specifically, the group SU(2).  The unit reals and complexes form abelian groups (respectively, \mathbb{Z}_2, and U(1)).  These groups all have familiar names.  The quasigroup of unit octonions doesn’t have any other more familiar name.  If you believe in the fundamental importance of this sequence of four division algebras, though, it does suggest that a natural sequence in which to weaken axioms for “multiplication” goes: commutative-and-associative, associative, Moufang.

The Moufang axiom does imply some other commonly suggested weakenings of associativity, as well.  For instance, a quasigroup that satisfies the Moufang axiom must also be alternative (a restricted form of associativity when two copies of one element are next to each other: i.e. the left alternative law x(xy) = (xx)y, and right alternative law x(yy) = (xy)y).

Now, there are various ways one could go with this; the one I’ll pick is toward physics.  The first three entries in that sequence of four division algebras – and the corresponding groups – all show up all over the place in physics.  \mathbb{Z}_2 is the simplest nontrivial group, so this could hardly fail to be true, but at any rate, it appears as, for instance, the symmetry group of the set of orientations on a manifold, or the grading in supersymmetry (hence plays a role distinguishing bosons and fermions), and so on.  U(1) is, among any number of other things, the group in which action functionals take their values in Lagrangian quantum mechanics; in the Hamiltonian setup, it’s the group of phases that characterizes how wave functions evolve in time.  Then there’s SU(2), which is the (double cover of the) group of rotations of 3-space; as a consequence, its representation theory classifies the “spins”, or angular momenta, that a quantum particle can have.

What about the octonions – or indeed the quasigroup of unit octonions?  This is a little less clear, but I will mention this: John Baez has been interested in octonions for a long time, and in Zurich, gave a talk about what kind of role they might play in physics.  This is supposed to partially explain what’s going on with the “special dimensions” that appear in string theory – these occur where the dimension of a division algebra (and a Clifford algebra that’s associated to it) is the same as the codimension of a string worldsheet.  J.B.’s student, John Huerta, has also been interested in this stuff, and spoke about it here in Lisbon in February – it’s the subject of his thesis, and a couple of papers they’ve written.  The role of the octonions here is not nearly so well understood as elsewhere, and of course whether this stuff is actually physics, or just some interesting math that resembles it, is open to experiment – unlike those other examples, which are definitely physics if anything is!

So at this point, the foregoing sets us up to wonder about two questions.  First: are there any quasigroups that are actually of some intrinsic interest which don’t satisfy the Moufang axiom?  (This might be the next step in that sequence of successively weaker axioms).  Second: are there quasigroups that appear in genuine, experimentally tested physics?  (Supposing you don’t happen to like the example from string theory).

Well, the answer is yes on both counts, with one common example – a non-Moufang quasigroup which is of interest precisely because it has a direct physical interpretation.  This example is the composition of velocities in Special Relativity, and was pointed out to me by Derek Wise as a nice physics-based example of nonassociativity.  That it’s also non-Moufang is also true, and not too surprising once you start trying to check it by a direct calculation: in each case, the reason is that the interpretation of composition is very non-symmetric.  So how does this work?

Well, if we take units where the speed of light is 1, then Special Relativity tells us that relative velocities of two observers are vectors in the interior of B_1(0) \subset \mathbb{R}^3.  That is, they’re 3-vectors with length less than 1, since the magnitude of the relative velocity must be less than the speed of light.  In any elementary course on Relativity, you’d learn how to compose these velocities, using the “gamma factor” that describes such things as time-dilation.  This can be derived from first principles, nor is it too complicated, but in any case the end result is a new “addition” for vectors:

\mathbf{v} \oplus_E \mathbf{u} = \frac{ \mathbf{v} + \mathbf{u}_{\parallel} + \alpha_{\mathbf{v}} \mathbf{u}_{\perp}}{1 + \mathbf{v} \cdot \mathbf{u}}

where \alpha_{\mathbf{v}} = \sqrt{1 - \mathbf{v} \cdot \mathbf{v}}  is the reciprocal of the aforementioned “gamma” factor.  The vectors \mathbf{u}_{\parallel} and \mathbf{u}_{\perp} are the components of the vector \mathbf{u} which are parallel to, and perpendicular to, \mathbf{v}, respectively.

The way this is interpreted is: if \mathbf{v} is the velocity of observer B as measured by observer A, and \mathbb{u} is the velocity of observer C as measured by observer B, then \mathbf{v} \oplus_E \mathbf{u} is the velocity of observer C as measured by observer A.

Clearly, there’s an asymmetry in how \mathbf{v} and \mathbf{u} are treated: the first vector, \mathbf{v}, is a velocity as seen by the same observer who sees the velocity in the final answer.  The second, \mathbf{u}, is a velocity as seen by an observer who’s vanished by the time we have \mathbf{v} \oplus_e \mathbf{u} in hand.  Just looking at the formular, you can see this is an asymmetric operation that distinguishes the left and right inputs.  So the fact (slightly onerous, but not conceptually hard, to check) that it’s noncommutative, and indeed nonassociative, and even non-Moufang, shouldn’t come as a big shock.

The fact that it makes B_1(0) into a quasigroup is a little less obvious, unless you’ve actually worked through the derivation – but from physical principles, B_1(0) is closed under this operation because the final relative velocity will again be less than the speed of light.  The fact that this has “division” (i.e. cancellation), is again obvious enough from physical principles: if we have \mathbf{v} \oplus _E \mathbf{u}, the relative velocity of A and C, and we have one of \mathbf{v} or \mathbf{u} – the relative velocity of B to either A or C – then the relative velocity of B to the other one of these two must exist, and be findable using this formula.  That’s the “division” here.

So in fact this non-Moufang quasigroup, exotic-sounding algebraic terminology aside, is one that any undergraduate physics student will have learned about and calculated with.

One point that Derek was making in pointing this example out to me was as a comment on a surprising claim someone (I don’t know who) had made, that mathematical abstractions like “nonassociativity” don’t really appear in physics.  I find the above a pretty convincing case that this isn’t true.

In fact, physics is full of Lie algebras, and the Lie bracket is a nonassociative multiplication (except in trivial cases).  But I guess there is an argument against this: namely, people often think of a Lie algebra as living inside its universal enveloping algebra.  Then the Lie bracket is defined as [x,y] = xy - yx, using the underlying (associative!) multiplication.  So maybe one can claim that nonassociativity doesn’t “really” appear in physics because you can treat it as a derived concept.

An even simpler example of this sort of phenomenon: the integers with subtraction (rather than addition) are nonassociative, in that x-(y-z) \neq (x-y)-z.  But this only suggests that subtraction is the wrong operation to use: it was derived from addition, which of course is commutative and associative.

In which case, the addition of velocities in relativity is also a derived concept.  Because, of course, really in SR there are no 3-space “velocities”: there are tangent vectors in Minkowski space, which is a 4-dimensional space.  Adding these vectors in \mathbb{R}^4 is again, of course, commutative and associative.  The concept of “relative velocity” of two observers travelling along given vectors is a derived concept which gets its strange properties by treating the two arguments asymmetrically, just like like “commutator” and “subtraction” do: you first use one vector to decide on a way of slicing Minkowski spacetime into space and time, and then use this to measure the velocity of the other.

Even the octonions, seemingly the obvious “true” example of nonassociativity, could be brushed aside by someone who really didn’t want to accept any example: they’re constructed from the quaternions by the Cayley-Dickson construction, so you can think of them as pairs of quaternions (or 4-tuples of complex numbers).  Then the nonassociative operation is built from associative ones, via that construction.

So are there any “real” examples of “true” nonassociativity (let alone non-Moufangness) that can’t simply be dismissed as not a fundamental operation by someone sufficiently determined?  Maybe, but none I know of right now.  It may be quite possible to consistently hold that anything nonassociative can’t possibly be fundamental (for that matter, elements of noncommutative groups can be represented by matrices of commuting real numbers).  Maybe it’s just my attitude to fundamentals, but somehow this doesn’t move me much.  Even if there are no “fundamentals” examples, I think those given above suggest a different point: these derived operations have undeniable and genuine meaning – in some cases more immediate than the operations they’re derived from.  Whether or not subtraction, or the relative velocity measured by observers, or the bracket of (say) infinitesimal rotations, are “fundamental” ideas is less important than that they’re practical ones that come up all the time.

I recently got back to London, Ontario from a trip to Ottawa, the first purpose of which was to attend the Ottawa Mathematics Conference. The other purpose was to visit family and friends, many of whom happen to be located there, which is one reason it’s taken me a week or so to get around to writing about the trip. Now, the OMC was a general-purpose conference, mainly for grad students, and some postdocs, to give short talks (plus a couple of invited faculty from Ottawa’s two universities – the University of Ottawa, and Carleton University – who gave lengthier talks in the mornings). This is not a type of conference I’ve been to before, so I wasn’t sure what to expect.

From one, fairly goal-oriented, point of view, the style of the conference seemed a little scattered. There was no particular topic of focus, for instance. On the other hand, for someone just starting out in mathematical research, this type of thing has some up sides. It gives a chance to talk about new work, see what’s being done across a range of subjects, and meet people in the region (in this case, mainly Ottawa, but also elsewhere across Eastern and Southern Ontario). The only other general-purpose mathematics conference I’ve been to so far was the joint meeting of the AMS in New Orleans in 2007, which had 5000 people and anyone attending talks would pick special sessions suiting their interests. I do think it’s worthwhile to find ways of circumventing the various pressures toward specialization in research – it may be useful in some ways, but balance is also good. Particularly for Ph.D. students, for whom specialization is the name of the game.

One useful thing – again, particularly for students – is the reminder that the world of mathematics is broader than just one’s own department, which almost certainly has its own specialties and peculiarities. For example, whereas here at UWO “Applied” mathematics (mostly involving computer modelling) is done in a separate department, this isn’t so everywhere. Or, again, while my interactions in the UWO department focus a lot on geometry and topology (there are active groups in homotopy theory and noncommutative geometry, for example), it’s been a while since I saw anyone talk about combinatorics, or differential equations. Since I actually did a major in combinatorics at U of Waterloo, it was kind of refreshing to see some of that material again.

There were a couple of invited talks by faculty. Monica Nevins from U of Ottawa gave a broad and enthusiastic survey of representation theory for graduate students. Brett Stevens from Carleton talked about “software testing”, which surprised me by actually being about combinatorial designs. Basically, it’s about the problem of how, if you have many variables with many possible values each, to design a minimal collection of “settings” for those variables which tests all possible combinations of, say, two variables (or three, etc.). One imagines the variables representing circumstances software might have to cope with – combinations of inputs, peripherals, and so on – so the combinatorial problem is if there are 10 variables with 10 possible values each, you can’t possibly test all 10 billion combinations – but you might be able to test all possible settings of any given PAIR of variables, and much more efficiently than just an exhaustive search, by combining some tests together.

Among the other talks were several combinatorial ones – error correcting codes using groups, path ideals in simplicial trees (which I understand to be a sort of generalization to simplicial sets of what trees are for graphs), heuristic algorithms for finding minimal cost collections of edges in weighted graphs that leave the graph with at least a given connectivity, and so on. Charles Starling from U of O gave an interesting talk about how to associate a topological space to an aperiodic tiling (roughly, any finite-size region in an aperiodic tiling is repeated infinitely many times – so the points of the space are translations, and two translations are within \epsilon of one another if they produce matching regions about the origin of size \frac{1}{\epsilon} – then the thing is to study cohomology of such spaces, and so forth).

The talk immediately following mine was by Mehmetcik Pamuk about homotopy self-equivalences of 4-manifolds, which used a certain braid of exact sequences of groups of automorphisms (among other things). I expected this to be very interesting, and it was certainly intriguing, but I can’t adequately summarize it – whatever he was saying, it proved to be hard to pick up from just a 25 minute talk. I did like something he said in his introduction, though: nowadays, if a topologist says they’re doing “low-dimensional” topology, they mean dimension 3, and “high-dimensional” means dimension 4. This is a glib but indicative way to point out that topology of manifolds in dimensions 1 and 2 is well understood (the connected components are, respectively, circles and n-holed tori), and in dimension 5 and above have been straightened out more recently thanks to Smale.

There were some quite applied talks which I missed, though I did catch one on “gravity waves”, which turn out not to be gravitational waves, but the kind of waves produced in fluids of varying density acted on by gravity. (In particular, due to layers of temperature and pressure in the atmosphere, sometimes denser air sits above less dense air, and gravity is trying to reverse this, producing waves. This produces those long rippling patterns you sometimes see in high-altitude clouds. Lidia Nikitina told us about some work modelling these in situations where the ground topography matters, such as near mountains – and had some really nice pictures to illustrate both the theory and the practice.)

On the second day there were quite a few talks of an algebraic or algebra-geometric flavour – about rings of algebraic invariants, about enumerating lines in special “blow-up” varieties, function fields associated to hyperelliptic curves, and so on – but although this is interesting, I had a harder time extracting informative things to say about these, so I’ll gloss over them glibly. However, I did appreciate the chance to gradually absorb a little more of this area of math by osmosis.

The flip side of seeing what many other people are doing was getting a chance to see what other people had to say about my own talk – about groupoids, spans, and 2-vector spaces. One of the things I find is that, while here at UWO the language of category theory is widely used (at least by the homotopy theorists and noncommutative geometry people I’ve been talking to), it’s not as familiar in other places. This seems to have been going on for some time – since the 1970’s if I understand the stories correctly. After MacLane and Eilenberg introduced categories in the 1940’s, the concept had significant effects in algebraic geometry/topology, homological algebra, and spread out from there. There was some deep enthusiasm – possibly well-founded, though I won’t claim so – that category theory was a viable replacement for set theory as a “foundation” for mathematics. True or not, that idea seemed to be one of those which was picked up by mathematicans who didn’t otherwise know much about category theory, and it seems to be one that’s still remembered. So maybe it had something to do with the apparent fall from fashion of category theory. I’ve heard that theory suggested before: roughly, that many mathematicians thought category theory was supposed to be a new foundation for mathematics, couldn’t see the point, and lost interest.

Now, my view of foundations is roughly suggested in my explanation of the title of this blog. I tend to think that our understanding of the world comes in bits and pieces, which we refine, then try to stick together into larger and more inclusive bits and pieces – the “Atlas” of charts of the title. This isn’t really just about the physical world, but the mathematical world as well (in fact I’m not really a Platonist who believes in a separate “world” of mathematical objects – though that’s a different conversation). This is really just a view of epistemology – namely, empirical methods work best because we don’t know things for sure, not being infinitely smart. So the “idealist”-style program of coming up with some foundational axioms (say, for set theory), and deriving all of mathematics from them without further reference to the outside doesn’t seem like the end of the story. It’s useful as a way of generating predictions in physics, but not of testing them. In mathematics, it generates many correct theorems, but doesn’t help identify interesting, or useful, ones.

So could category theory be used in foundations of mathematics? Maybe – but you could also say that mathematics consists of manipulating strings in a formal language, and strings are just words in a free monoid, so actually all of mathematics is the theory of monoids with some extra structure (giving rules of inference in the formal language). Yet monoid theory – indeed, algebra generally – is not mainly interesting as foundations, and probably neither is category theory.

On the whole, it was an interesting step out of the usual routine.

I’m going up to Ottawa for a few days, in part to talk about spans and groupoids (basically, some cross section of the material in these posts here) at a conference put on by the Ottawa U math department, primarily for grad students and postdocs in the general vicinity. This is nice – gives me a chance to visit my parents and friends there (the fraction of my life I lived in Ottawa is now creeping down toward a mere third, but it probably has as strong a claim to “home” as anywhere). May is also one of the most tolerable months to be there. One of the grad students in our department is also going. Enxin Wu recently decided to start working with Dan Christensen too, so probably in future we’ll have various things to talk about. Last week, he gave a seminar talk on algebra deformation that was a long version of the one he’ll be giving in Ottawa.

Enxin is one of those guys who seems to really understand – it’s tempting to say grok– algebra, which I always find impressive. I’m a predominantly visual thinker, and the kind of symbolic computations common in algebra always seem a little mysterious to me at first until I can find a picture, or at least practice them a lot. Lie groups, for instance, make some sense to me – you can picture rotation groups, or at least keep a geometric picture of a manifold in mind. Lie algebras, being infinitesimal versions of Lie groups, are also not so hard to visualize. General associative algebras? Harder.

The talk was about associative algebras, to give some background on deformation, but the things whose deformations Enxin has been thinking about are A_{\infty}-algebras (see this brief intro, for instance), an “invention” of Stasheff. The talk was about deformation of these algebras – the kind of deformation that pertains to deformation quantization. This has been studied by Kontsevich. Deformation quantization has to do with replacing things valued in some algebra A by new things, valued in the bigger algebra A[[t]] of formal power series in t with coefficients in A, so that the original structure you started with is just the constant part that appears when you set t=0. (The term “quantization” applies when you consider algebras of functions on a manifold, with a Poisson bracket – in other words, algebras of observables of a physical system).

Some of the main results have to do with the Hochschild cohomology for some complex associated to the algebra you start with, and the fact that this cohomology classifies obstructions to the deformation. I expected to get lost in a maze of notation – and there certainly is a lot – but as it turns out, I had some mental pictures to attach to these things, because related things came up a few years ago in the quantum gravity seminar at UCR (week 8 on that page especially), which provides a few pictures that helped a lot. Diagrammatic notation makes algebra a lot more comprehensible to me.

So let’s get more specific.

The point is to replace a multiplication operator m : A \otimes A \rightarrow A with a power series whose coefficients are “multiplication” operators. That is, a deformation of an associative algebra (A,m) (where m : A \otimes A \rightarrow A is the multiplication for A) is (A[[t]],m_t), where the new multiplication m_t is defined (by linearity) by its action on elements of A, which works like this:

m_t(a,b) = \sum_{i=0}^{\infty} {\alpha_i}(a,b){t^i}

for some operators \alpha_i : A \otimes A \rightarrow A. Then there are a bunch of conditions on the \alpha that are needed to make m_t associative. There’s one condition for each power of t, since the coefficients in the associator should be zero:

\sum_{i+j=n\\i,j>0} \alpha_i( (\alpha_j \otimes 1) - (1 \otimes \alpha_j)) = 0

The n=0 condition just says that \alpha_0 is associative – so it’s the m from the original algebra, which you get back when t=0.

Then given an algebra A, you can create the deformation category \mathcal{D} of A whose objects are its deformations. The morphisms are continuous algebra homomorphisms that get along with the multiplication operations. It turns out that since formal power series with nonzero n=0 term are invertible (a consequence of the Lagrange theorem) this \mathcal{D} is actually a groupoid. Then the question is to classify the isomorphism classes of deformations – that is, \Pi_0(\mathcal{D}). One can easily imagine that there might be no nontrivial deformations of some algebra – that is, every one is isomorphic to the deformation where all the \alpha_i are trivial except \alpha_0 = m. So when does this happen? More generally, how can one classify the deformations up to isomorphism?

The answer has to do with Hochschild cohomology, which is related to a complex you can make from A. Taking C^n(A) = hom(A^{\otimes n},A), the space of n-ary multilinear operations on A, you build this complex:

0 \stackrel{d_0}{\longrightarrow} C^0(A) \stackrel{d_1}{\longrightarrow} C^1(A) \stackrel{d_2}{\longrightarrow} \dots

where the differential maps are d_n : C^n(A) \rightarrow C^{n+1}(A) defined by an alternating sum:

d(f)(a_1, \dots, a_n) = a_1  f(a_2, \dots, a_{n+1}) + \sum_{i=1}^{n} (-1)^i f(a_1, \dots, a_i a_{i+1}, \dots, a_{n+1}) + (-1)^{n+1} f(a_1, \dots,a_n) a_{n+1}

(Intuitively: there are too many arguments, so you start with the extra one on the left, push it into the middle as a “lump under the rug” where two arguments are combined, and push the lump all the way to the right. To ensure that d^2 = 0, you do this with alternating signs. This kind of algebraic manipulation is the kind of thing I can do, and clearly works, but I don’t exactly grok.)

Then you take the Hochschild cohomology groups in the standard cohomology way: HH^i = \frac{ker(d_{i+1})}{Im(d_i)}. A cohomology class in one of these groups is a class of multilinear maps from n copies of A to A (up to a factor which is d_n of something). As usual with cohomology, they describe obstructions to something – to exactness. Exactness, in this setting, would mean that A has no interesting deformations at the n^{th} level.

What does “level” mean here? Well, for example, at level 2 we’re talking about maps A \otimes A \rightarrow A, such as the multiplication map. In fact, we have d_3(m) = 0 for an associative algebra – you can check that d(m) is twice the associator a_1(a_2a_3) - (a_1a_2)a_3, which is zero. So m is a cochain. Is it a coboundary? Sure – it’s d_2(1). So m is in the trivial class in HH^2(A). The point then is that it turns out that if this is the only class – if HH^2(A) = 0 – then there are no interesting deformations of the multiplication of A in the sense described above. The groupoid $\mathcal{D}$ has just one object. (One thing that occurs to me is that this makes it a group – which group is something Enxin didn’t discuss. My algebra instincts aren’t quite up to answering that off the top of my head.) For example, if A = \mathbb{C} (as an algebra over \mathbb{R}), there are no nontrivial deformations: HH^2(\mathbb{C}) = 0.

What do the other levels mean? Really, this is where you’d want to look at the generalization from associative algebras to A_{\infty}-algebras. Whereas for an associative algebra A, the associator $a(x,y,z) = x(yz) – (xy)z$ is zero, in general an A_{\infty}-algebra will have an associator map a : A^{\otimes 3} \rightarrow A (that is, a \in C^3 in the complex above), which might not be zero, but which is d_3(m).

This is the beginning of a story relating A_{\infty}-algebras to weak \infty-categories: a bicategory, for example, has an associator for composition of morphisms. In a bicategory, you expect the associator to satisfy a certain identity – the Pentagon identity – but in general you’d just ask for a “pentagonator” (something in C^4), and so on (this is where those seminar notes above help me think in pictures, by the way). An A_{\infty}-algebra is a vector space equipped with maps at all these levels – described by Stasheff’s associahedra – satisfying some relations. The general story of deformation relates the Hochschild cohomology groups at different levels to deformations of A_{\infty}-algebras. Enxin didn’t go into this in his talk, but he did say a little something about the next level:

An infinitesimal deformation of A is a deformation not in A[[t]], but in the quotient A[[t]]/(t^2=0). This only needs two maps, \alpha_0 , \alpha_1. The third Hochschild cohomology measures obstructions to extending an infinitesimal deformation to a full deformation in A[[t]] – if HH^3(A) = 0, then any infinitesimal deformation can be extended to a full deformation.

All in all, I thought the talk was interesting – it tied in much more closely to things I already knew about TQFTs and higher categories than I’d expected. I’ll be really impressed if he can condense it into a 25-minute version…