Mathematical physics and M-theory: The study of higher structures

In recent posts we’ve begun to discuss some ideas at the foundation of the duality symmetric approach to M-theory. As we started to review in the last entry, one of the first goals is to formulate and study a general field theory in which T-duality is a manifest symmetry. It was discussed how this was the first-principle goal of double field theory, and it was similarly featured as a motivation in our introductory review of double sigma models. There is a lot to be discussed about the duality symmetric approach moving forward, including the effective theory for this doubled string prior to ultimately looking at lifting to M-theory, where, instead of double field theory we will be working with what is known as exceptional field theory. What also remains an important question has to do with obtaining a global formulation of such duality symmetric actions. What is clear is that higher geometry and algebra are important to achieving such a formulation, and there is much ground to cover on this topic.

Meanwhile, in the present entry I would like to share what I have been studying and learning about as it relates to the other side of my PhD research: the higher structure approach to M-theory. If the duality symmetric approach is a sort of bottom-up way to attack the M-theory proposal, particularly insofar that we are building from the field theory point of view, the higher structure approach can be looked at here as a sort of top-down way to access the question of string theory’s non-perturbative completion. Although this language is a bit schematic, as there is a lot of overlap between the two approaches and their machinery, it does lend some intuition to the different perspectives being undertaken.


In William Thurston’s 1994 essay, ‘On proof and progress in mathematics‘ [1], it was argued that progress in mathematics is driven not only by proof of new theorems. Progress is also made by aiding in human beings ways to think about and understand mathematics. Emily Riehl made this a point of emphasis at the beginning of her notes on categorical homotopy theory [2], including on the usefulness of qualitative insights, and I think a similar emphasis may be made here in the context of our focus in mathematical physics and particularly M-theory. A further point of philosophical emphasis in this essay is Eugene Wigner’s article on the unreasonable effectiveness of mathematics in physics and, finally, the more recent presentation by Robbert Dijkgraaf on the unreasonable effectiveness of string theory in mathematics. In my view, M-theory represents one of a few research topics at the frontier of mathematical physics. What parametrises the boundaries of this frontier is the interface between foundational maths and fundamental physics. Indeed, I take this as Dijkgraaf’s point in his presentation at String Math 2020, namely both the need for this engagement and how, historically, progress is often made when the two sides (mathematics and physics) interact. For myself, I almost joined the maths school prior to deciding my future was in mathematical physics, and I find great interest in working at this interface, where, furthermore, when thinking of M-theory Thurston’s notion of progress appears particularly apt.

The motivation may be stated thusly [3]: there presently exists many interconnected hints in support of the proposed existence of M-theory. But a systematic formulation of the full theory – i.e., string theory’s non-perturbative completion – remains an important open problem. A key issue here ultimately concerns the lack of clarity about the underlying principles of M-theory (there are many references on this point, but as one example see [4]). I look at the current situation as a puzzle or as a patchwork quilt. There are pieces of the total picture that we can identify and start to fill in. There are others that remain unknown, leaving empty spaces in our picture of M-theory. And then, finally, how all of the pieces relate or connect is another question that we need to answer but cannot currently access.

To advance the problem, there is ample reason to suggest and to argue that what is needed is new mathematical machinery. As a new researcher, this need was something that I started thinking about a year or more ago. Let me put it this way: our world is best described by quantum field theory. If M-theory is the correct description of fundamental physics, we should end up with a quantum field theoretic description. But it seems unlikely that M-theory will be captured or defined by some Lagrangian, or some S-matrix, or other traditional approaches [3]. Indeed, the tools we need are more than just fibre bundles, standard topology, or differential geometry. Although much of modern physics is built using tools and approaches that deal with local, approximate, perturbative descriptions of reality, in investigating the M-theory problem we need to find ways of dealing with the global and non-perturbative structure of physical fields, and thus we are dealing with the difficulty of employing non-perturbative methods. Entering into my PhD, this is the challenge that I see. I also see this challenge, from the perspective of fundamental physics, as being similar to the situations that have historically arisen many times. A large part of the history of fundamental physics is described by the search for new mathematical language required to aid the modelling of physical phenomena. Hisham Sati and Urs Schreiber [5] presented the argument well, describing the situation explicitly, when discussing the motivation for pursuing a rigorous mathematical foundation for quantum field theory and perturbative string theory. As an example, they cited the identification of semi-Riemannian differential geometry as the underlying structure of gravity. Or, think of the use of representation theory in particle physics. In truth, there are many examples and, to Dijkgraaf’s point, we should embrace this history.

I think this is why, as I prepare to start my formal PhD years, the 2018 Durham Symposium seems momentous, particularly as I begin to generate my thoughts on M-theory and what sort of research I might find meaningful. Although it was slightly before my time, as I was only a first-year undergraduate when the Durham symposium had taken place, I was already developing an interest in non-perturbative theory and I remember learning of the symposium with enthusiasm. It gave me confidence and, I suppose, assurance that my thoughts are moving in the right direction. I’ve also taken confidence from many other important conferences, such as the 2015 conference organised around the theme of new spaces in mathematics and physics. But, for me, the Durham symposium has become a tremendous reference, because the culmination of this search for new mathematical language is apparent, organised under the study of higher structures, and I find this programme of research immensely stimulating.

Similar to the situation in QFT where, over the past decade or more, progress has been made to understand its fundamental nature – for instance, efforts to define QFT on arbitrary corbordism – higher structures provides a concise language of gauge physics and duality that has seemed, in recent years, to open pathways to rigorously attack the M-theory question. Indeed, efforts toward an axiomatic formulation of QFT (for instance, see recent developments in the area of algebraic and topological QFT [6]) and those toward string theory’s full non-perturbative completion to M-theory have a lot in common. Furthermore, an important motivation for the study of higher structures (and higher differential geometry, higher gauge theory and symmetry algebras, and so on) comes directly from decisive hints about the inner workings of M-theory. Hence, the title of the Durham Symposium and its guiding document, ‘Higher structures in M-theory‘.

To give some immediate examples and sketch a few more introductory thoughts, the higher algebraic structures we know to govern closed string field theory is something I started to investigate as related to my recent MRes thesis. But the most basic example of a higher structure in string theory arguably goes back to the first quantisation of the bosonic string. Indeed, as I described in a past note (I think from my first-year undergrad), if I were to teach strings one day my opening lecture would be on generalising point particle theory and emphasising the motivation on why we want to do this. From this approach, I think one can show in a wonderfully pedagogical way that, when generalising from 0-dimensional point particle theory to the 1-dimensional string (and so on), higher dimensionality is a natural consequence and is essentially forced upon us. (As an aside, I remember reading a comment by Schreiber about this very same point of introduction. I recommend reading Schreiber’s many notes over the years. For instance, here is a forethoughtful contribution from 2004 that begins to motivate some of the concepts we will discuss below. A helpful online resource is also ncatlab that covers many of the topics we will be discussing on this blog, along with appropriate references). And, it turns out, this is one way we might also motivate in fundamental physics the study of higher structures; because, in this picture, the Kalb-Ramond 2-form can be seen as an example of a higher structure as it is generalised from the gauge potential 1-form [3]. Of course, since the mid-1990s, a growing body of evidence urged the string theory community to study extended objects of dimension > 1  , and around the same time attempts were already developing to use category theory (more on categories in a moment) to study string diagrams [7], as one can certainly see that string diagrams possess a powerful logic when it comes to composition.


So what do we mean by higher structures? From my current vantage, I would describe a higher structure as a categorified mathematical structure, which I also take to mean higher homotopy theory. But we can perhaps begin to build toward the idea by reviewing briefly two main ingredients: category theory and homotopy theory. As a matter of correspondence between mathematics and physics, category theory is the mathematical language of duality and homotopy theory is the mathematical language of the gauge principle.

We may think of category theory as being positioned at the foundations of modern mathematics [8], but, in many ways, it is quite elementary. Similar to the use of a venn diagram when teaching basic set theory, we can build the idea of a category in a fairly intuitive way.

A category {\mathcal{C}} consists of the following data [9]:

* A collection of mathematical objects. If {X} is an object of {\mathcal{C}} , then we write {X \in \mathcal{C}} .

* Every pair of objects {X, Y \in \mathcal{C}} , we may define a set of morphisms {X \rightarrow Y} denoted as {\text{Hom}\mathcal{C}(X,Y)} .

* For every {X \in \mathcal{C}} , there needs to exist an identity morphism {Id_{X} \in \mathcal{C}(X,X)} .

* For every triple {X,Y,Z \in \mathcal{C}} , we may define a composition map {\circ : \mathcal{C}(X,Y) \times \mathcal{C}(Y, C) \rightarrow C(X, Z)} .

* Composition is associative and unital.

If category theory is the mathematics of mathematics, I would currently emphasise in a physics context [10] the approach to category theory as the language that describes composition. Think of the trivial example of moving in some space (let’s not get too stuck on definitions at this point). We can compose the journey from points A to B to C to D in the following way,

\displaystyle  A \rightarrow B \rightarrow C \rightarrow D \ (1).

We can also compose the same journey in terms of pairs of vertices or what we are presently calling points such that

\displaystyle  A \rightarrow C, B \rightarrow D \ (2)  ,

and then we may write the entire journey as {A \rightarrow D} giving the same description in (1).

The idea of a category can be constructed using similar logic. Given a collection of objects {A,B,C,D} , paths {A \rightarrow B \rightarrow C \rightarrow D} denoted by the arrows may be defined as the relation amongst the objects in terms of structure preserving maps {f,g,h} called morphisms.


So at its most basic, a category is a collection of objects and arrows between those objects. It is, in some sense, a relational set, which must follow the conditions stated above.

Example. The category of sets, denoted by Set. The category of R-modules, denoted by RMod. A morphism {f : X \rightarrow Y} is said to be an isomorphism if there exists {g : Y \rightarrow X} such that {g \circ f = Id_{X}} and {f \circ g = Id_{Y}} . In the category Set, isomorophisms are bijections.

The concept of functors is of deep importance in this language. In short, a functor is a morphism between categories. If {\mathcal{C}} and {\mathcal{D}} are categories, we may define a functor {F : \mathcal{C} \rightarrow \mathcal{D}} such that it assigns an object {FX \in \mathcal{D}} for any {X \in \mathcal{C}} , and a morphism {Ff : FX \rightarrow FY} for any {f : X \rightarrow Y} , where associativity and unitality are preserved. So, for instance, if {f : X \rightarrow Y} , {g : Y \rightarrow X} , associativity is preserved such that

\displaystyle  Fg \circ Ff = F(g \circ f) \ (3).

We may also define the notion of a natural transformation as a morphism between functors. If {F,G :  \mathcal{C} \rightarrow \mathcal{D}} define two functors, then a natural transformation {F \implies D} assigns any {X \in \mathcal{C}} a morphism {FX \rightarrow GX} .

There is a lot to be said about functors, categorical products, and also the important role duality plays in category theory. In the next entries, we will formally define these ideas as well as many others. For now, I am simply trying to provide some sense of an early introduction into some of the machinery used when we speak of higher structures, such as by giving an intuitive example of a category, with a mind toward formal definition in a following post. The same can be said for all ideas presented here, as, in the present entry, we are simply encircling concepts and sketching a bit of land, similar as a geoscientist would do when first preparing to sketch a topological map.

What one will find, on further inspection, is that category theory is deeply interesting for a number of reasons. At its deepest, there is something to be said about it as a foundational framework. One of the most inspiring realisations about category theory comes from something that seems incredibly basic: the idea in set theory of taking the product of two sets. Indeed, one may have seen this notion of a product as fundamental. But what we observe is that this most basic concept of taking a product of two sets is not fundamental in the way we may have been used to thinking, because one of the amazing things about the story of category theory is how the idea of products is more deeply defined in terms of a categorical product. The reward for this realisation, aside from shear inspiration, is technically immense.

Indeed, a category can contain essentially any mathematical object, like sets, topological spaces, modules, and so on. In many constructions, one will seek to study very generally the products of these objects – so, for example, the product of topological spaces – and the concept of a product in category theoretic language can capture all such instances and constructions. In later discussions we will see how this language allows us to look at mathematics at a large scale, which is to say that, in the abstract, we can take any collection of mathematical objects and study the relations between them. So if the goal is a completely general view, using category theory we are able to strip back a lot of inessential detail so as to drill fundamentally into things.

Additionally, there is a deep relationship between category theory and homotopy theory, which, in this post, I would like to highlight on the way to offering a gentle introduction to the concept of a higher structure. Down the road we will discuss quite a bit about higher-dimensional algebra, such as n-categories and operads, which are algebraic structures with geometric content, as we drive toward a survey of the connection between higher categorical structures and homotopy theory. In physics, there is also connection here with things like topological quantum field theory. Needless to say, there is much to cover, but when thinking of homotopy theory at its most basic, it is appropriate to go all the way back to algebraic topology.

The philosophical motivation is this: there are many cases in which we are interested in solving a geometrical problem of global nature, and, in algebraic topology, the method is generally to rework the problem into a homotopy theoretic one, and thus to reduce the original geometric problem to an algebraic problem. Let me emphasise the key point: it is a fundamental achievement of algebraic topology to enable us to reduce global topological problems into homotopy theory problems. One may motivate the study of homotopy theory thusly: if we want to think about general topological spaces – for example, arbitrary spaces that are not Hausdorff or even locally contractible – what this amounts to is that we relax our interest in the notion of equivalence under homeomorphism (i.e., topological equivalence) and instead work up to homotopy equivalence.

Definition 1 Given maps {f_0,f_1: X \rightarrow Y} , we may write {f_0 \simeq f_1} , which means {f_0} is homotopic to {f_1} , if there exists a continuous map {F : X \times I \rightarrow Y} , called a homotopy, such that {F(x,0) = f_0(x)} and { F(x,1) = f_1(x)} . We may also write {F: f_0 \implies f_1} to denote the homotopy.

As suggested a moment ago, a homotopy relation {\simeq} is an equivalence relation. This is true if {F_{01} : f_0 \implies f_1} and {F_{12} : f_1 \implies f_2} for the family of maps {f_i : X \rightarrow Y} , then

F_{02} (t,x) =  \begin{cases} F_{01}(2t,x)  :  0 \leq t \leq 1/2 \\ F_{01}(2t-1,x)  :  1/2 \leq t \leq 1 \\          \end{cases} \ (4)

gives a homotopy {F_{02} : f_0 \implies f_2} .

As an aside, what is both lovely and interesting is how, from a physics perspective, we may think of homotopy theory and ask how it might relate to the path integral; because, on first look, it would seem intuitive to ask this question. There is a long and detailed way to show it to be true, but, for simplicity, the argument goes something as follows. Think, for starters, of what we’re saying in the definition of homotopy. Given some {X} , which for now we’ll define as a set but later understand as a homotopy type, let us define two elements {x,y \in X} such that we may issue the following simple proposition {x = y} . The essential point, here, is that there may be more than one way that {x} is equal to {y} , or, in other words, there may be more than one reason or more than one path. Hence, we can construct a homotopy {\gamma} such that x \xrightarrow[]{\gamma} y  is a homotopy from {x} to {y} and then an identity map {Id_{X}(x,y)} for the set of homotopies from {x} to {y} in {X} . One can then proceed to follow the same reasoning and construct a higher homotopy by defining a homotopy of homotopy and so on.

The analogy I am drawing is that, in the path integral formalism, given some simply-connected topological space, recall that we can continuously deform the path {x(t)} to {x(y)} . In this deformfation, {\phi[x(t)]} approaches {\phi[y(t)]} continuously such that, taking the limit, we have

\displaystyle  \phi[y(t)]=\lim\phi[x(t)]=e^{iS[y(t)]}, \ \text{as }x(t)\rightarrow y(t) \text{continuously}. \ (5)

The principle of the superposition of quantum states, or, the sum of many paths, in a simply-connected space can be constructed as a single path integral; because, when all of the dust settles, the paths in this space can be shown to contribute to the total amplitude with the same phase (this is something we can lay out rigorously in another post). The result is that we end up with the Feynman path integral.

In homotopy theory, on the other hand, the analogous is true in that paths in the same homotopy class contribute to the total amplitude with the same phase. So, if one defines the appropriate propagator and constrains appropriately to the homotopy class, an equivalent expression for the path integral may be found. And really, one can probably already start to suspect this in the basic example of homotopy theory of topological spaces. Typically, given a topological space {X} and two continuous functions from this space to another topological space {Y} such that

\displaystyle f,g : X \rightarrow Y \ (6)  ,

it is straightforward to define, with two points in the mapping space, {f,g \in \text{Maps}(X,Y)} a homotopy {\eta}

\displaystyle  f \xrightarrow[]{\eta} g \ (7).

This is just a collection of continuous paths between the points.

But I digress. The focus here is to build up to the idea of higher structures.

The reason that a brief introduction to homotopy theory aids this purpose is because, if we think of a higher structure as a categorified mathematical structure, what we are referring to is a phenomenon in which natural algebraic identities hold up to homotopy. In other words, we’re speaking of mathematical structure in homotopy theory and thus of higher algebra, higher geometry, and so forth. Higher algebra consists of algebraic structure within higher category theory [11, 12]. As we discussed earlier, categories have a set of morphisms between objects, and, so, in the example of the category of sets, elements of a set may or may not be equal. Higher categories, much like higher algebra, are a generalisation of these sort of constructions we see in ordinary category theory. In the higher case we now have homotopy types of morphisms, which are called mapping spaces. And so, unless we are working with discrete objects, we must deal with homotopy as an equivalence relation should two so-called elements of a homotopy type, typically represented by vertices, be connected in a suitable way.

When we speak of higher structures as mathematical structures in homotopy theory, this is more specifically a mathematical structure in {(\infty, 1)-\text{category theory}} . This is a special category such that, from within the collection of all {(n, r)-\text{categories}} , which is defined to be an {\infty-\text{category}} satisfying a number of conditions, we find an {(\infty, 1)-\text{category theory}} to be a weak {\infty-\text{category}} in which all n-morphisms for {n \geq 2} are equivalences. I also think of a higher structure almost as a generalisation of a Bourbaki mathematical structure. But perhaps this comment should be reserved for another time.

In summary, if as motivation it is the case that we often want to study homotopy theory of homotopy theories, for instance what is called a Quillen model category, what we find is a hierarchy of interesting structures, which is described in terms of the homotopy theoretic approach to higher categories. And it is from this perspective that homotopy theories are just {(\infty, 1)-\text{category theory}} , where {\infty} denotes structure with higher morphisms (of all levels) and the 1 refers to how all the 1-morphisms and higher morphisms are weakly invertible. Hence, too, in higher category theory we may begin to speak of {(\infty, n)-\text{categories}} , which may be described as:

1. An n-category up to homotopy (satisfying the coherence laws, more on this in a later post);

2. An {(r, n)-\text{categories}} for {r = \infty} ;

3. A weak {\infty-\text{category}} or {\omega-\text{category}} where all k-morphisms are equivalences satisfying the condition {k > n} .

There are different ways to define {(\infty, n)-\text{categories}} , and their use can be found in such places as modern topological field theory. If category theory is a powerful language to study the relation between objects, n-categories enables us to then go on and study the relations between relations, and so on. As an example, consider the category of all small categories. For two categories {\mathcal{C}} , {\mathcal{D}} , whose morphisms are functors, the set or collection of all morphisms hom-set {\text{Fun}(C, D)} are then functors from {\mathcal{C}} to {\mathcal{D}} . This forms a functor category in which all morphisms are natural transformations, given that the natural transformations are morphisms between morphisms (functors). Hence, in this way, we scratch the surface of the idea of higher categories, because, taking from what was mentioned above, these are categories equipped with higher {n} -morphisms between {(n-1)}-morphisms for all {n \in \mathbb{N}} .

Moreover, if in ordinary category theory there are objects and morphisms between those objects, from the higher category view these are seen as 1-morphisms. Then, we may define a 2-category, which is just a generalisation that includes 2-morphisms between the 1-morphisms. And we can therefore continue this game giving definition to {n} -category theory. We will eventually get into more detail about the idea of {n} -categories, including things like weak {n} -categories where associativity and identity conditions are no longer given by equalities (i.e., they are no longer strict), instead satisfied up to an isomorphism of the next level. But for now, in thinking of the basic example of a composition of paths and this notion of generalising to 2-morphisms between the 1-morphism, the emphasis here is on the idea that the two conditions of associativity and identity must hold up to reparameterisation (the topic of reprematerisation being a whole other issue) – hence, up to homotopy – and what this amounts to is a 2-isomorphism for a 2-category. If none of this is clear, hopefully more focused future notes will help spell it all out with greater lucidity.


In using the language of higher structures in M-theory, there have been many promising developments. For instance, it can be seen how core structures of string/M-theory emerge as higher structures in super homotopy theory [4, 13], leading to a view of M-theory beginning from the superpoint in super Minkowski spacetime going up to 11-dimensions. An interesting part of this work was the use of Elmendorf’s theorem on equivariant homotopy theory. It has led to exciting new developments in our picture of brane physics, with an updated brane bouquet.

Of course, the higher structures programme is far-reaching. From double and exceptional field theory and the global formulation of such actions to the study of homotopy algebras in string field theory, M-branes, sigma models on gerbes, and even modern views on anomalies in which field theories are treated as functors – this merely scratches the surface. Some nice lecture notes on higher structures in M-theory, focusing for example on M5-brane systems and higher gauge theory were recently offered by Christian Saemann [14]. Hopefully we will be able to cover many of these ideas (and others) moving forward. Additionally, I am currently enjoying reading many older works, such as Duiliu-Emanuel Diaconescu’s paper on enhanced D-brane categories in string field theory [15], and I’ve been working through Eric Sharpe’s 1999 paper [16], which was the first to explicitly draw the correspondence between derived categories and Dp-branes in his study of Grothendieck groups of coherent sheaves. These and others will be fun papers to write about in time.

To conclude, we’ve begun to introduce, even if only schematically, some important ideas at their most basic when it comes to studying higher structures in M-theory. In the next entries, we can deepen our discussion with more detailed notes and definitions, perhaps beginning with a formal discussion on category theory and then homotopy theory, and then a more rigorous treatment of the idea of a higher structure.


[1] William P. Thurston. On Proof and Progress in Mathematics, pages 37–55. Springer New York, New York, NY, 2006.

[2] Emily Riehl. Categorical Homotopy Theory. New Mathematical Monographs. Cambridge University Press, 2014.

[3] Branislav Jurco, Christian Saemann, Urs Schreiber, and Martin Wolf. Higher structures in m-theory, 2019.

[4] Domenico Fiorenza, Hisham Sati, and Urs Schreiber. The rational higher structure of m-theory. Fortschritte der Physik, 67(8-9):1910017, May 2019.

[5] Hisham Sati and Urs Schreiber. Survey of mathematical foundations of qft and perturbative string theory, 2012.

[6] J. Baez and J. Dolan. Higher dimensional algebra and topological quantum field theory. Journal of Mathematical Physics, 36:6073–6105, 1995.

[7] Daniel Marsden. Category theory using string diagrams, 2014.

[8] Birgit Richter. From Categories to Homotopy Theory. Cambridge Studies in Advanced Mathematics. Cambridge University Press, 2020.

[9] Carlos T. Simpson. Homotopy theory of higher categories, 2010.

[10] Bob Coecke and Eric Oliver Paquette. Categories for the practising physicist, 2009.

[11] T. Leinster. Topology and higher-dimensional category theory: the rough idea. arXiv: Category Theory, 2001.

[12] J. Baez. An introduction to n-categories. In Category Theory and Computer Science, 1997.

[13] John Huerta, Hisham Sati, and Urs Schreiber. Real ade-equivariant (co)homotopy and super m-branes. Communications in Mathematical Physics, 371(2):425–524, May 2019.

[14] Christian Saemann. Lectures on higher structures in m-theory, 2016.

[15] Duiliu-Emanuel Diaconescu. Enhanced d-brane categories from string field theory. arXiv: High Energy Physics – Theory, 2001.

[16] E. Sharpe. D-branes, derived categories, and grothendieck groups. Nuclear Physics, 561:433–450, 1999.

*Image: ‘Homotopy theory harnessing higher structures’, Newton Institute.

*Edited for spelling, grammar, and syntax.