This book aims to provide an accessible introduction to the field of stable homotopy theory using the language of \(\infty \)-categories. Our primary goal is to develop the technical foundations of the subject, including stability and the \(\infty \)-category of spectra, the theory of \(\infty \)-operads, and some basic theory of ring spectra.
Historical overview
To put into perspective the modern viewpoint on stable homotopy theory, let us take a step back and consider some of the historical developments leading up to it.
In 19th-century topology, mathematicians like Bernhard Riemann, Enrico Betti and Henri Poincaré studied manifolds using numerical invariants such as Betti numbers and torsion coefficients. In 1925, Emmy Noether emphasized that the underlying homology groups should themselves be regarded as invariants. Their ranks and torsion subgroups recover the numerical data, but the groups provide a more natural algebraic framework. This helped shift attention from numerical invariants to homology groups. Samuel Eilenberg and Norman Steenrod subsequently developed an axiomatic approach to homology and cohomology, presented systematically in their 1952 book.
A central feature of reduced homology is its stability: the reduced homology of a space \(X\) agrees, up to a degree shift, with that of its suspension \(\Sigma X\). Homotopy groups do not generally have this stability property, but the Freudenthal suspension theorem, originating in work of Hans Freudenthal in 1938 [Freudenthal (1938)], shows that they acquire it after sufficiently many suspensions. In its general form, it states that if \(X\) is an \(n\)-connected pointed CW complex, with \(n \geq 0\), then the canonical map \(X \to \Omega \Sigma X\) induces an isomorphism on \(\pi _k\) for \(k \leq 2n\) and a surjection for \(k = 2n+1\). Applying this to successive suspensions \(\Sigma X\), \(\Sigma ^2X\), etcetera, one finds that the sequence of abelian groups1 \[ \pi _k(X) \to \pi _{k}(\Omega \Sigma X) \to \pi _k(\Omega ^2 \Sigma ^2 X) \to \, \ldots \] eventually stabilizes for every \(k \geq 0\). Its colimit is called the \(k\)-th stable homotopy group of \(X\), denoted \(\pi _k^{\st }(X)\). In contrast to their unstable analogues, the stable homotopy groups form a (reduced) generalized homology theory, satisfying the EilenbergâSteenrod axioms except for the dimension axiom. Other important examples of generalized (co)homology theories include cobordism theory, arising from RenĂ© Thomâs work on manifolds with tangential structures modulo bordism [Thom (1954)], and topological K-theory, developed by Atiyah and Hirzebruch from vector bundles over a space [Atiyah and Hirzebruch (1959)].
The success of homology theory in topology inspired mathematicians to develop analogous theories in other fields, leading for example to the theories of group cohomology, Lie algebra cohomology, and associative algebra cohomology. A common framework for these developments was established by Henri Cartan and Samuel Eilenberg in their monumental 1956 book Homological Algebra, which gave the field its name. Its central notion is that of a derived functor, which in the case of the tensor product and the hom-functor yields \(\Tor \)-groups and \(\Ext \)-groups, respectively. Many previously ad hoc cohomology theories could now be described as instances of \(\Ext \)-groups.
Extending the stabilization of homotopy groups to maps between spaces, Spanier and Whitehead introduced the set of stable maps between two pointed CW-complexes \(X\) and \(Y\): \[ [X,Y]^{\st }_* \quad := \quad \colim _n [\Sigma ^nX, \Sigma ^n Y]_*. \] Finite pointed CW-complexes and stable maps assemble into a category, which Spanier and Whitehead used to investigate duality phenomena of spaces [Spanier and Whitehead (1953)]. Enlarging this category by formal desuspensions makes suspension invertible and yields what is now commonly called the SpanierâWhitehead category.
Building on Spanier and Whiteheadâs work, Spanierâs student Lima [Lima (1959)] introduced the first definition of a spectrum, and used it to extend their duality theory to a more general class of spaces. Several constructions followed, culminating in Boardmanâs construction of the stable homotopy category; a published treatment is given by Adams [Adams (1974)]. One convenient presentation of spectra is the following:2
Definition. A spectrum \(X\) is a sequence of pointed spaces \(\{X_n\}_{n \geq 0}\) equipped with pointed homotopy equivalences \(X_n \simeq \Omega X_{n+1}\), where \(\Omega (-)\) denotes the loop space; if in addition \(\pi _k(X_n) = 0\) for \(0 \leq k < n\), we say that \(X\) is connective.
The homotopy category of spectra contains the SpanierâWhitehead category as a full subcategory, represents all generalized cohomology theories on CW complexes via Brownâs representability theorem, and admits an associative, commutative and unital smash product. It provides a common framework for computational tools such as the Adams spectral sequence [Adams (1958)], Bousfieldâs localizations at homology theories [Bousfield (1979)], and the methods of chromatic homotopy theory [Ravenel (1984); Ravenel (1992)].
The years that followed saw a surge in the study of âbrave new algebraâ [May (1977)], i.e. the development of algebraic constructions and methods for spectra. Key technical ingredients included the theory of operads developed by May [May (1972)] and BoardmanâVogt [Boardman and Vogt (1973)], and point-set models of spectra with suitably structured smash products [Lewis et al. (1986); Elmendorf et al. (1997); Hovey et al. (2000)]. Many familiar algebraic constructions on abelian groups have homotopical analogues for spectra:
- There is a direct sum \(X \oplus Y\) of two spectra, which is simultaneously a product and a coproduct;
- There is a tensor product \(X \otimes Y\) of two spectra. This leads to the notion of a ring spectrum: a spectrum \(R\) equipped with a multiplication operation \(m\colon R \otimes R \to R\) that is âcoherentlyâ associative and unital;
- There is a notion of a module over a ring spectrum. Familiar notions for modules over (commutative) rings, like the properties of being projective, perfect, or flat, admit generalizations to the setting of ring spectra;
- Given a spectrum \(X\) and a prime \(p\), we may form its \(p\)-localization \(X_{(p)}\) and its \(p\)-completion \(X^{\wedge }_p\), which have similar behavior as their analogues for abelian groups;
- And so on and so forth...
When formulated intrinsically in the \(\infty \)-category of spectra, these constructions are already derived. For example, unlike the tensor product of abelian groups, the tensor product of spectra is exact in both variables. The exactness that homological algebra obtains by passing to derived functors is thus built into the setting. In homological algebra, this point of view goes back to the derived categories of Grothendieck and Verdier. This also explains why the techniques and concepts from stable homotopy theory have found significant applications in numerous neighboring fields of mathematics, like (derived) algebraic geometry, algebraic K-theory, geometric topology and representation theory.
Higher category theory
To fully appreciate the pervasive role of (stable) homotopy theory in modern mathematics, we must first embrace a new way of thinking about equality:
Fundamental Principle of Homotopy Theory: When expressing that two entities are equal, we must always specify how they are equal by providing an identification between them.
This principle is already familiar to most mathematicians when applied to mathematical objects: for instance, when we want to express that two abelian groups \(A\) and \(B\) are âthe sameâ, we usually mean that there is an isomorphism \(A \cong B\) between them. In homotopy theory, we extend this principle to morphisms between objects; for example, two continuous maps \(f,g\colon X \to Y\) are considered âequalâ when equipped with a homotopy \(f \sim g\) between them. More broadly, in homotopy theory, the notion of âsamenessâ is not a property (a binary yes/no question) but rather structure that must be explicitly provided. We use the word identification for this general notion, which takes the form of an isomorphism for abelian groups or a homotopy for continuous maps. This shift in perspective originated in algebraic topology, but has by now become fundamental in many neighboring fields of mathematics.
While this principle appears straightforward, it rapidly gives rise to intricate âcoherence problemsâ. Consider, for instance, the task of defining a homotopical version of an abelian group \(A\). Focusing first on associativity of addition \(+\colon A \times A \to A\), the Fundamental Principle requires providing an identification between the two maps \(A \times A \times A \to A\) given by \((a,b,c) \mapsto a + (b + c)\) and \((a,b,c) \mapsto (a + b) + c\). Once such an identification is given, one can construct two identifications between the maps \((a,b,c,d) \mapsto a + (b + (c + d))\) and \((a,b,c,d) \mapsto ((a + b) + c) + d\). Identifying these requires an identification between identifications. Continuing with more inputs leads to higher associativity coherences, as studied by Stasheff (1963a). Commutativity and its compatibility with associativity require further coherence data.
The modern solution to such coherence problems lies in the theory of \(\infty \)-categories. Conceptually, an \(\infty \)-category behaves much like a classical 1-category: we may speak of objects and morphisms that can be composed. But in accordance with the Fundamental Principle, we also keep track of identifications between parallel morphisms, identifications between those identifications, and so on. Homotopical versions of familiar categorical constructions are conveniently formulated within this framework: for example, we may define for any \(\infty \)-category \(C\) with finite products a new \(\infty \)-category \(\CGrp (C)\) of commutative groups in \(C\), where all the required higher coherences are built into the objects.
As we will see, there are \(\infty \)-categories \(\Ss \) and \(\Sp \) whose objects are spaces3 and spectra, respectively, and whose identifications encode the familiar homotopical data. The following theorem establishes a crucial relation between these two \(\infty \)-categories:
Recognition Principle for Connective Spectra (BoardmanâVogt, May and Segal). There is a fully faithful functor of \(\infty \)-categories \[ \CGrp (\Ss ) \hookrightarrow \Sp \] whose image consists of the connective spectra.4
In other words, we may think of a connective spectrum as a âhomotopy coherentâ analogue of an abelian group. This result is part of a much broader phenomenon: many algebraic structures have natural âhomotopy coherentâ analogues in the world of spectra (think for example of the ring spectra mentioned before).
While coherence problems in homotopy theory can be (and historically have been) handled without \(\infty \)-categories, the \(\infty \)-categorical framework builds the Fundamental Principle into its foundations, so that many homotopy-theoretic concepts admit native formulations. The methods from higher category theory are rapidly becoming a standard tool among researchers, and they are slowly making their way into more and more graduate courses at universities. It thus seems like the time is ripe for an accessible resource that develops the foundations of stable homotopy theory using the language of \(\infty \)-categories. Providing such a resource is the primary objective of this book.
From abelian categories to stable \(\infty \)-categories
With the concept of an \(\infty \)-category at hand, we can now return to the question of what makes stable homotopy theory such a suitable framework for doing homological algebra. The short answer is that it allows for a richer notion of exact sequences, refining the classical notion of short exact sequences in abelian categories. To see what this means, let us start from the classical situation.
Let us recall how short exact sequences work in an abelian category \(C\). Such a category admits a zero object \(0\), meaning that for every object \(X\) there are unique maps \(X \to 0\) and \(0 \to X\), and every morphism admits a kernel and a cokernel. Moreover, given a diagram \[ X \xhookrightarrow {i} Y \overset {p}{\twoheadrightarrow } Z \] with \(pi = 0\), where \(i\) is a monomorphism and \(p\) is an epimorphism, the map \(i\) exhibits \(X\) as the kernel of \(p\) if and only if \(p\) exhibits \(Z\) as the cokernel of \(i\). When these equivalent conditions hold, we call the diagram a short exact sequence.
The story changes a little if \(C\) is an \(\infty \)-category with a zero object. In this case, the equality \(pi = 0\) from before should be replaced by a choice of nullhomotopy \(pi \simeq 0\). Following the Fundamental Principle, we must take this choice as part of the data; whether the resulting diagram is exact can depend on it. Similarly, we replace the kernel of a map \(g\colon Y \to Z\) by its fiber: the terminal example of an object \(X\) equipped with a map \(f\colon X \to Y\) and a nullhomotopy \(gf \simeq 0\); dually, we replace cokernels by cofibers. These considerations motivate the following central definition in stable homotopy theory:
Definition. An \(\infty \)-category \(C\) is called stable if it has a zero object, if it has fibers and cofibers, and if for every composable pair of maps \(X \xrightarrow {\smash {f}} Y \xrightarrow {\smash {g}} Z\) equipped with a nullhomotopy \(gf \simeq 0\), the map \(f\) and the given nullhomotopy exhibit \(X\) as the fiber of \(g\) if and only if \(g\) and the given nullhomotopy exhibit \(Z\) as the cofiber of \(f\).
We refer to the diagrams satisfying these equivalent conditions as exact sequences. The agreement between fiber and cofiber sequences has powerful consequences. First, every map \(f\colon X \to Y\) fits into an exact sequence \[ X \xrightarrow {f} Y \longrightarrow \cofib (f), \] so exact sequences are no longer restricted to monomorphisms followed by epimorphisms, as they are in abelian categories. More importantly, a functor between stable \(\infty \)-categories is exact as soon as it preserves either finite limits or finite colimits. In particular, left and right adjoints between stable \(\infty \)-categories are exact. Tensoring with a fixed spectrum and taking internal hom from that spectrum form such an adjoint pair, so both functors are already exact.
An important example of a stable \(\infty \)-category is the \(\infty \)-category \(\Sp \) of spectra; in fact, it is a universal example.5 Furthermore, every abelian category \(\Aa \) gives rise to a stable \(\infty \)-category by considering its derived \(\infty \)-category \(\D (\Aa )\), which comes equipped with a fully faithful inclusion \(\Aa \hookrightarrow \D (\Aa )\) that turns short exact sequences in \(\Aa \) into exact sequences in \(\D (\Aa )\).
As we will explore during the course of this book, most of the usual constructions from homological algebra can be performed directly at the level of stable \(\infty \)-categories, without needing to first perform them at the level of abelian categories and then pass to derived functors. The classical derived functors are then obtained as the homotopy groups of these constructions. For example, every ordinary commutative ring \(R\) and every ordinary \(R\)-module \(M\) determine an EilenbergâMacLane ring spectrum and a module spectrum concentrated in degree \(0\), which we denote by the same letters. For two \(R\)-modules \(M\) and \(N\), their classical Tor- and Ext-groups are recovered as \[ \Tor _n^R(M,N) \cong \pi _n(M \otimes _R N) \qquadtext { and } \Ext _R^{n}(M,N) \cong \pi _{-n}(\hom _R(M,N)), \] for \(n \geq 0\), where \(- \otimes _R -\) and \(\hom _R(-,-)\) denote the tensor product and internal hom in the \(\infty \)-category of \(R\)-module spectra.
The theory of \(\infty \)-operads
The higher coherences resulting from the Fundamental Principle do, however, make it considerably more subtle to define and manipulate algebraic structures in the setting of \(\infty \)-categories than in classical algebra. Writing out all these coherences individually quickly becomes unwieldy, so we need a formalism that organizes them systematically. One such formalism is the theory of \(\infty \)-operads, extending the classical theory of operads developed by May (1972) and Boardman and Vogt (1973). Our approach builds on Lurieâs treatment of this theory [Lurie (2017)]. We devote a substantial part of this book to \(\infty \)-operads, with the aim of providing an accessible introduction to them.
Heuristically, an operad encodes operations and their composition laws. For example, the commutative operad has a single \(n\)-ary operation for every \(n \geq 0\). Interpreted in abelian groups with their tensor product, it describes commutative rings: its operations act by the multiplication maps \(R^{\otimes n} \to R\), with the nullary operation giving the unit map \(\Z \to R\). More generally, an operad \(\Oo \) specifies a set \(\mathcal {O}(n)\) of \(n\)-ary operations for each \(n \geq 0\), together with actions of permutations of the inputs and composition laws. For example, plugging an \(n\)-ary operation into one input of a \(k\)-ary operation yields an \((n+k-1)\)-ary operation. These composition laws must be associative, unital and compatible with permutations.
In the \(\infty \)-categorical setting, the composition laws and their higher coherences must be encoded together. Following Lurieâs approach, we do this through an \(\infty \)-category of operators, which organizes operations and their composition into categorical data. We use an equivalent formulation over the \((2,1)\)-category \(\Span (\Fin )\) of spans of finite sets, in place of Lurieâs category \(\Fin _*\) of finite pointed sets. The comparison between these formulations is established in Section 19.4.
Content
The material is organized into a foundational chapter followed by three parts. Chapter 1 introduces \(\infty \)-categories and animae. Part I develops the foundations of stable homotopy theory: (co)homology theories, spectra and stable \(\infty \)-categories, the recognition principle, homological algebra, and localization and completion, followed by ring spectra, complex K-theory, Thom spectra and duality phenomena. Part II treats \(\infty \)-operads over \(\Span (\Fin )\) and the higher algebra built on them, including the tensor product of spectra and the theory of algebras and modules. Part III gathers the \(\infty \)-categorical background used throughout: adjunctions, presentable \(\infty \)-categories, cocartesian fibrations and straightening, and complete Segal animae. Each part opens with an introduction describing its chapters in more detail.
Prerequisites and how to read this book
The main prerequisites are 1-category theory and algebraic topology at the level of a first graduate course. In particular, the reader should be comfortable with limits and colimits, adjoint functors, CW complexes, homotopy groups, and singular homology and cohomology. Some familiarity with chain complexes and vector bundles is useful, but the relevant material is recalled when it first becomes essential. No prior familiarity with \(\infty \)-categories, spectra, operads, model categories, or higher algebra is assumed.
This book is available in a published and an online version. The online version contains supplementary sections, identified by â(online only)â in their titles, and uses color in some explanatory diagrams. The supplementary sections are placed at the ends of chapters, so all material appearing in both versions has the same chapter, section, and theorem numbering. Page numbers, however, differ between the two versions.
The book is not intended to be read strictly from beginning to end. Chapter 1 provides the common foundation for Parts I and II, while Part III collects technical background that may be consulted when it is cited. Parts I and II have several mutual dependencies, which are summarized in Figure 1. For example, Chapter 8 on ring spectra requires the coherent tensor product of spectra, which is only constructed in Part II. So that readers can learn about ring spectra without first working through the higher algebra of Part II, the chapter begins by collecting the results it needs from Part II in a âcoherent algebra packageâ (Proposition 8.1.5). Part II in turn relies on the theory of spectra from Chapter 4 and, in its application to derived categories, on Chapter 6.
Here are several possible reading routes:
- (1)
-
For a first course in stable homotopy theory, read Chapter 1 followed by Part I in order, treating the forward references from Chapter 8 to Part II as black boxes. The final three chapters may be selected according to interest.
- (2)
-
For higher algebra, begin with Chapter 1 and then read Part II in order. The construction of the tensor product of spectra in Chapter 16 uses Chapter 4, and the derived-category application in Chapter 21 uses Chapter 6; these applications may be skipped if the reader is primarily interested in the abstract operadic theory. The relevant chapters of Part III may be consulted when they are first cited.
- (3)
-
For ring spectra and derived algebra, read Part I through Chapter 7, then Part II through Chapter 18, and then return to Chapter 8. For the derived-category constructions, also read Chapter 21.
- (4)
-
For the categorical foundations, read Chapter 1 followed by Part III.
Acknowledgments
This book has grown out of two graduate lecture courses on stable homotopy theory and higher algebra taught at the University of Regensburg in 2024/2025. In preparing the manuscript I have benefited from many existing sources on (stable) homotopy theory and higher algebra. I would like to acknowledge some of these influences more explicitly.
The \(\infty \)-categorical approach to stable homotopy theory and higher algebra adopted in this book is deeply indebted to Lurieâs foundational text Higher Algebra [Lurie (2017)], which remains one of the central references in the field. Most of the theory of \(\infty \)-operads developed in Part II ultimately goes back to this work, even when the order of exposition or the proofs differ from Lurieâs treatment. The exposition in Part I is strongly influenced by the lecture notes of Denis Nardin, Introduction to stable homotopy theory [Nardin (2021)], in particular in the choice and ordering of topics. For Part II, I have benefited from Haugsengâs introductory note on \(\infty \)-operads [Haugseng (2023a)], and from Gepnerâs survey article on higher algebra [Gepner (2020)]. The brief historical sketch of stable homotopy theory in the introduction draws on the accounts of May and Weibel [May (1999b); Weibel (1999)]. Classical developments of the theory of spectra that have shaped my background, and from which I have borrowed a number of examples and perspectives, may be found in the textbooks of Switzer and Adams [Switzer (1975); Adams (1974)].
Although no single section of the present book follows them closely, Fabian Hebestreitâs lecture course on algebraic and Hermitian K-theory [Hebestreit and Wagner (2021)] was my own first systematic introduction to \(\infty \)-categories and modern stable homotopy theory. It has had a lasting impact on how I think about the subject and, indirectly, on the way the material is presented here.
The model-agnostic approach to \(\infty \)-categories used throughout this book is based on joint work with Denis-Charles Cisinski, Kim Nguyen and Tashi Walde [Cisinski et al. (2026)].
The pictorial illustrations for multimorphisms in \(\infty \)-operads were inspired by [Barkan and Steinebrunner (2022), Figure 1].
I am grateful to many colleagues and students for their help and encouragement. I would like to thank Denis-Charles Cisinski, Daniel Gratzer, Rune Haugseng, Sil Linskens, Maxime Ramzi, Stefan Schwede, Jan Steinebrunner, Tashi Walde, Ferdinand Wagner and Christoph Winges for numerous helpful conversations about the material in this book. I am especially grateful to Marc Hoyois for reading substantial parts of the manuscript and for pointing out several mathematical errors. The participants of the two Regensburg courses on which this book is based provided valuable feedback and many questions that led to clarifications and improvements in the exposition. I also thank Keima Aksaka, Karolis Dembickas, Alissa Doggwiler, Michael J. Glaeser, Tim Henke, Taiga Nakamura, Lorenzo Pascarella, Daniele Velati and Yiming Wang for sending corrections and further suggestions.
Use of generative AI
During the final preparation phase of this textbook6 , I have made substantial use of the AI coding assistants Claude Code and Codex. This includes:
- Editorial assistance, such as typo correction and LaTeX assistance;
- Mathematical review, such as finding errors and suggesting corrections;
- Substantive mathematical contribution, such as drafting proofs, explanations, and new material.
AI involvement was particularly substantial in the following chapters and sections: Chapter 6, Section 8.1, Section 5.5, Chapter 9, Section 10.3, Section 15.3, Section 15.1, Section 16.3, Section 19.5, Section 19.6, Section 20.2, Section 18.2, Chapter 21, and Chapter 23.
All AI-generated material incorporated into the text was reviewed and edited by me. This involved checking not only the mathematical arguments but also the compatibility with the bookâs overall structure and style. I remain solely responsible for the contents of the book and for any errors that remain.
Notes
3For reasons that will be explained, we will write \(\An \) rather than \(\Ss \) for this \(\infty \)-category, and refer to its objects as animae.
4The difference between spectra and connective spectra is analogous to the difference between arbitrary chain complexes and those whose homology vanishes in negative degrees.
5For every stable \(\infty \)-category \(C\) with small colimits, a colimit-preserving functor \(\Sp \to C\) is uniquely determined by its value on the sphere spectrum, and every object of \(C\) occurs this way. See Theorem 4.4.14.