Loading field map
Homological algebra is the branch of algebra that studies mathematical structures by attaching to them sequences of abelian groups or modules—called homology groups—that measure the obstruction to solving certain equations. Its central insight is that many qualitative questions about objects (Does this system of equations have a solution? Can this map be inverted? Is this object built from simpler pieces?) can be translated into quantitative questions about the size and shape of these attached groups. The subject provides a common language and toolkit for algebra, topology, and geometry, and it has reshaped how mathematicians think about structure and invariance.
The subject begins with a simple observation about exact sequences. A sequence of abelian groups and homomorphisms
\[ \cdots \to A{n+1} \xrightarrow{d{n+1}} An \xrightarrow{dn} A_{n-1} \to \cdots \]
is called a chain complex if the composition of any two consecutive maps is zero: \(dn \circ d{n+1} = 0\). This condition says that the image of each map is contained in the kernel of the next. When the image actually equals the kernel, the sequence is exact at that spot, meaning that the structure is "perfectly understood" there: everything that should be killed is killed, and nothing more.
The failure of exactness is measured by the homology at each spot:
\[ Hn(A\bullet) = \frac{\ker(dn)}{\operatorname{im}(d{n+1})}. \]
This quotient is an abelian group (or module) that records the extent to which the image falls short of the kernel. If the homology is zero everywhere, the complex is exact. Nontrivial homology indicates the presence of "holes" or "obstructions" that prevent the sequence from being exact.
The power of this definition comes from its flexibility. By choosing the right chain complex, one can encode a wide variety of mathematical problems. For example, given a module \(M\) over a ring \(R\), one can construct a free resolution—an exact sequence of free modules ending at \(M\). Applying a functor (such as \(\operatorname{Hom}R(-, N)\) or \(-\otimesR N\)) to this resolution typically destroys exactness; the homology of the resulting complex defines the Ext and Tor groups. These groups measure, respectively, the failure of \(\operatorname{Hom}\) to be exact (which relates to extension problems) and the failure of the tensor product to be exact (which relates to torsion and flatness). The remarkable fact is that these groups are independent of the chosen resolution, making them genuine invariants of the original module.
The roots of homological algebra lie in topology. In the late nineteenth and early twentieth centuries, mathematicians such as Henri Poincaré and L. E. J. Brouwer studied topological spaces by associating to them sequences of abelian groups—the homology groups of a space—that count holes of various dimensions. The key insight was that these groups are topological invariants: if two spaces are homeomorphic, their homology groups agree. This provided a powerful tool for distinguishing spaces.
In the 1920s and 1930s, Emmy Noether and her school recognized that the algebraic machinery underlying these topological invariants could be studied on its own terms. The notion of an exact sequence, formalized by Witold Hurewicz and others, became central. The subject crystallized as an independent discipline in the 1940s and 1950s through the work of Samuel Eilenberg and Saunders Mac Lane, who introduced the axiomatic treatment of homology theories, and of Henri Cartan and Jean-Pierre Serre, who developed the spectral sequence machinery. The publication of Homological Algebra by Cartan and Eilenberg in 1956 marked the subject's maturity as a self-contained algebraic theory, no longer dependent on topological motivation.
A second major wave came in the late 1950s with Alexander Grothendieck's reformulation of the subject in relative terms. Grothendieck introduced derived categories and derived functors as a unified framework, showing that many seemingly distinct constructions—Ext, Tor, sheaf cohomology, group cohomology—were instances of a single general mechanism. This perspective emphasized that homological algebra is not merely a collection of computational tools but a way of organizing mathematical information that respects the structure of the underlying category.
While homological algebra is a single discipline, two broad approaches have shaped its practice, and they coexist productively.
The older and more concrete tradition treats homological algebra as a set of computational techniques. The practitioner's goal is to compute specific homology groups, Ext groups, or Tor groups for particular rings, modules, or spaces. This tradition emphasizes explicit constructions: free resolutions, projective resolutions, injective resolutions, and the manipulation of chain complexes. It is the tradition of the original Cartan–Eilenberg book and of much of the classical literature on group cohomology and sheaf cohomology.
The computational tradition is characterized by its concreteness. One builds a resolution, applies a functor, and computes the homology. The tools are often elaborate: spectral sequences, which compute the homology of a double complex by successive approximation; the Künneth formula, which relates the homology of a tensor product of complexes to the tensor product of their homologies; and the universal coefficient theorem, which relates homology with coefficients in an arbitrary abelian group to homology with integer coefficients.
This tradition remains essential because it produces the actual numbers and groups that answer concrete questions. It is also the tradition that connects homological algebra to its topological origins, where computing the homology of a specific space is a meaningful and often difficult task.
The second tradition, associated primarily with Grothendieck and his school, treats homological algebra as a branch of category theory. The focus shifts from computing specific groups to understanding the formal properties of the constructions themselves. The central objects are not individual modules but abelian categories—categories that behave like the category of modules over a ring, with kernels, cokernels, and exact sequences—and the functors between them.
In this tradition, the fundamental question is: given a functor between abelian categories that is not exact, what is the minimal way to repair its failure of exactness? The answer is the theory of derived functors. A left exact functor (one that preserves kernels but not cokernels) has right derived functors, which measure the failure of exactness on the right; a right exact functor has left derived functors. The Ext and Tor groups are the derived functors of \(\operatorname{Hom}\) and \(\otimes\), respectively.
The categorical tradition reached its fullest expression in the theory of derived categories, introduced by Grothendieck and developed by Jean-Louis Verdier in his 1967 thesis. A derived category is obtained from the category of chain complexes by formally inverting quasi-isomorphisms—maps between complexes that induce isomorphisms on homology. In a derived category, two complexes that have the same homology become isomorphic, even if they are not isomorphic as complexes. This allows one to work with complexes up to homological equivalence, which is often the natural level of precision.
The derived category perspective has several advantages. It makes the independence of derived functors from the choice of resolution automatic, since different resolutions become isomorphic in the derived category. It also allows for a more flexible notion of composition of functors, since derived functors can be composed without worrying about the order of operations. This has proven essential in modern algebraic geometry, where derived categories of sheaves have become a central tool.
Several concepts recur throughout homological algebra, regardless of which tradition one works in.
Resolutions are exact sequences that replace a given object by a sequence of nicer objects. A projective resolution replaces a module by a sequence of projective modules (modules that behave like free modules with respect to lifting maps); an injective resolution replaces it by a sequence of injective modules (the dual notion). The existence of enough projectives or injectives in a category is a fundamental structural property. The category of modules over a ring always has both, but many natural categories have only one or the other.
Derived functors are the primary output of homological algebra. Given a functor \(F\) and an object \(M\), one applies \(F\) to a resolution of \(M\) and takes homology. The result is independent of the resolution, giving a well-defined sequence of groups \(R^i F(M)\). The zeroth derived functor recovers the original functor applied to \(M\); the higher derived functors measure the failure of exactness.
Spectral sequences are computational devices for calculating homology in complicated situations. When one has a filtered complex or a double complex, the homology is often difficult to compute directly. A spectral sequence provides a systematic way to approximate the answer: one starts with the homology of the "first page," then successively corrects it through a sequence of pages, each computed from the previous one. In favorable cases, the spectral sequence converges to the desired homology, and the computation reduces to understanding the pages and the differentials between them.
Long exact sequences are the fundamental structural result of homological algebra. Given a short exact sequence of complexes, one obtains a long exact sequence of homology groups, connecting the homology of the three complexes in an infinite alternating pattern. This result is the workhorse of the subject: it allows one to compute unknown homology groups from known ones, and it is the source of many of the invariants' most useful properties.
Contemporary homological algebra is characterized by the dominance of the derived category perspective, particularly in algebraic geometry and representation theory. The language of derived categories, triangulated categories, and their refinements (such as dg-categories and ∞-categories) has become standard in many areas of mathematics. This shift has been driven by the recognition that many natural constructions—such as the pushforward of sheaves or the tensor product of complexes—are only well-behaved at the level of derived categories, not at the level of individual objects.
At the same time, the computational tradition has not disappeared. Explicit computations of Ext and Tor groups remain essential in commutative algebra, where they provide invariants of rings and modules that are used to study singularities, dimension, and depth. The theory of local cohomology, which computes the cohomology of sheaves supported at a point, is a powerful tool in algebraic geometry that relies on homological methods.
A notable recent development is the rise of derived algebraic geometry, which uses derived categories and their higher-categorical analogues as the foundation for a new approach to geometry. In this framework, spaces are not just sets of points but carry additional homological information that records the "derived" structure of functions and sheaves on them. This has led to new insights in intersection theory, deformation theory, and the study of moduli spaces.
Another active area is homological mirror symmetry, a conjecture from mathematical physics that relates the derived category of coherent sheaves on one algebraic variety to the derived category of the Fukaya category (a category built from Lagrangian submanifolds) on a mirror variety. This conjecture, due to Maxim Kontsevich, has stimulated an enormous amount of work connecting homological algebra to symplectic geometry and string theory.
The subject also continues to interact with its topological origins. The development of stable homotopy theory has led to the notion of a triangulated category as an abstraction of the homotopy category of spectra, and the two subjects have influenced each other deeply. The modern theory of ∞-categories, developed by Jacob Lurie and others, provides a unified framework that encompasses both topological and algebraic homological algebra, and it has become the language of choice for many researchers.
Homological algebra is not a finished edifice. Several fundamental issues remain active areas of research. The construction of derived categories involves formal inverses of quasi-isomorphisms, which can be technically delicate; the theory of model categories and ∞-categories provides rigorous foundations, but the choice of framework can affect the results. The question of when two triangulated categories are equivalent is often difficult to answer, and the classification of such categories in specific contexts remains open.
The computational limits of the subject are also real. While homology groups are often computable in principle, the actual computation can be intractable for large or complicated objects. Spectral sequences, while powerful, often require deep insight to use effectively, and there is no general algorithm for determining when a spectral sequence collapses.
Finally, the relationship between the different traditions is not always smooth. The categorical tradition's emphasis on formal properties can obscure the concrete meaning of the invariants, while the computational tradition's focus on specific examples can miss the general patterns. The most successful practitioners move fluidly between the two, using categorical insight to guide computation and computational experience to inform categorical constructions. This interplay is likely to remain the source of the subject's vitality.