Skip to content

Delta Functor

A graded family of additive functors whose natural connecting maps turn short exact sequences into long exact sequences.

Version
v1 · 2026-10-03 · History
Domain-specific #
13131
Domain group
Formal Sciences
Origin domain
Mathematics
Subdomains
Homological Algebra, Category Theory → Mathematics
Aliases
Delta-functor, Δ-functor

Core Idea

A delta functor packages a graded family of additive functors with connecting morphisms that are coherent across short exact sequences. In the cohomological convention used by the Stacks Project, for abelian categories \(\mathcal A\) and \(\mathcal B\) there are covariant functors \(F^n:\mathcal A\to\mathcal B\) for \(n\geq0\). Every short exact sequence \(0\to A\to B\to C\to0\) supplies boundary maps \(\delta^n:F^n(C)\to F^{n+1}(A)\); the three terms at each degree, joined by those boundaries, form one long exact sequence. The boundary maps must also be natural when one short exact sequence maps to another.[1]

This is an axiom system, not a specific resolution algorithm. Sheaf cohomology \(H^n(X,-)\) and, with a fixed first module argument, \(\operatorname{Ext}^n_R(M,-)\) realize it in different categories. Resolutions are standard ways to construct these examples, but the delta-functor identity is the graded exactness-and-naturality relation itself. The degree-zero functor in this convention is necessarily left exact; adding arbitrary higher functors to a non-left-exact degree-zero component cannot satisfy the stated opening of the long exact sequence.[1][2][3]

Universal is a stronger modifier, not a synonym for delta functor. For a universal family, any natural transformation from \(F^0\) to the degree-zero member of another delta functor extends uniquely through all degrees. A sufficient condition is positive-degree effaceability by suitable monomorphisms. When a universal extension of a given degree-zero functor exists, it is unique up to unique isomorphism of delta functors. These claims have explicit hypotheses and do not say that every delta functor is universal.[1]

Structural Signature

Sig role-phrases: typed abelian categories and variance → graded additive functors → every short exact input → oriented connecting morphisms → natural long exact output → optional universal extension property.

  • Typed categories and variance. State \(\mathcal A\), \(\mathcal B\) and whether the varying argument is covariant or requires an opposite category. Fixing \(M\) in \(\operatorname{Ext}^n_R(M,-)\) gives the covariant second-variable example; varying the first argument without retyping it would reverse arrows.[1][3]
  • Graded family. The data are all \(F^n\) for \(n\geq0\), not one functor named \(F\). Each is additive and acts coherently on morphisms. A single global-sections or Hom functor supplies only degree zero.[1]
  • Short exact input. The rule applies to every \(0\to A\to B\to C\to0\) in the source abelian category, not to an illustrative sequence chosen after the fact.[1]
  • Oriented boundaries. In the cohomological form, \(\delta^n:F^n(C)\to F^{n+1}(A)\) crosses from quotient-side output to subobject-side output one degree higher. The homological counterpart uses a degree-lowering boundary; the two conventions cannot be superimposed without changing the formula.[1][4]
  • Long exactness and naturality. The concatenation \(0\to F^0(A)\to F^0(B)\to F^0(C)\xrightarrow{\delta^0}F^1(A)\to\cdots\) is exact, and every morphism of short exact inputs makes the boundary square commute. Pointwise exact lists with arbitrary, nonnatural joining arrows fail this test.[1]
  • Optional universality. The degree-zero extension property and a sufficient effaceability condition distinguish universal members of the class; they are not needed to recognize a basic delta functor.[1]

What It Is Not

It is not a lone functor. Each \(F^n\) is a functor, but the identity here is the family plus maps \(\delta^n\) and axioms coordinating those maps across exact inputs. This is why the live Functor node is a structural prerequisite, not a strict genus asserted merely from the shared word.[1]

It is not an isolated long exact sequence. One such sequence may arise from a particular input, while a delta functor provides a natural assignment for every short exact input. The live Exact Sequence node names the necessary exactness format, not the whole rule that generates it.[1]

It is not identical to Derived Functor. Right-derived functors of a suitable left-exact functor canonically form a cohomological delta functor and may be universal under the stated acyclic-embedding hypothesis. But “derived” describes a construction using homological resolutions or a derived category; the delta axioms say what graded response and boundary maps are required. The construction and its formal output should not be collapsed.[5]

It is not automatically universal, and Tor is not a cohomological instance under the displayed formula. Tor is left-derived from tensor product and carries a homological degree-lowering long-exact orientation. Calling it a related delta-functor pattern is reasonable only after declaring that dual convention, not by writing \(\operatorname{Tor}_n(C,M)\to\operatorname{Tor}_{n+1}(A,M)\) as if it were the cohomological boundary.[1][6][4]

Scope of Application

The exact formal scope is abelian categories, additive graded functors and natural boundary maps for short exact sequences. The source and target categories may be sheaves and abelian groups, modules and abelian groups, or another appropriate abelian pair; the same symbols do not license leaving this categorical setting. In Stacks's cohomological convention the degree-zero component must be left exact, as follows from the first three terms of its long exact output.[1]

For a fixed topological space \(X\), the category \(\mathrm{Ab}(X)\) of abelian sheaves has enough injectives. Stacks defines \(H^n(X,-)\) from an injective resolution of a sheaf and states that the resulting family is a universal delta functor \(\mathrm{Ab}(X)\to\mathrm{Ab}\). A short exact sequence of sheaves therefore yields natural connecting maps from \(H^n(X,\mathcal F'')\) to \(H^{n+1}(X,\mathcal F')\).[2]

For a fixed \(R\)-module \(M\), \(\operatorname{Ext}^n_R(M,-)\) is covariant in its second argument, with degree zero \(\operatorname{Hom}_R(M,-)\). Short exact sequences in that second variable yield long exact Ext sequences. Computing with injective resolutions supplies a right-derived realization and the usual universal case. Reversing which argument varies changes variance and must be explicitly retyped.[3][5]

Clarity

Three adjectives do important work. Cohomological means the displayed boundary raises the superscripted degree from \(F^n(C)\) to \(F^{n+1}(A)\). Natural means a morphism between two short exact sequences makes the corresponding boundary square commute. Universal means a degree-zero transformation has a unique compatible graded extension. None of those adjectives can be inferred solely from seeing a row of groups with arrows.[1]

The frozen seed conflated the broader pattern with its common examples. In particular, it listed Ext and Tor under the same covariant cohomological formula. Fixed-first-argument Ext fits that formula; Tor\(_n(-,M)\) is the left-derived tensor family and its boundary lowers homological degree. Both illustrate related long-exact organization, but an accurate entry must preserve their orientation rather than treating index placement as decorative.[3][6][4]

Manages Complexity

The delta-functor language lets one prove a result once about how a whole graded theory responds to an extension, instead of reconstructing a connecting map separately for every sheaf or module calculation. If the input sequence and naturality data are in place, the long exact output organizes kernels, cokernels and obstruction groups in a predictable order. One can ask which degree-zero operation is being extended and which higher group records the failure of exactness, without forgetting the precise connecting arrow.[1][2][3]

Universality gives a second compression. When its hypotheses hold, one natural map at degree zero determines a unique morphism of the full delta-functor systems; two universal systems extending the same degree-zero functor agree up to unique isomorphism. Effaceability is a sufficient test that justifies this compression, not an incantation that works for arbitrary graded families.[1]

Abstract Reasoning

Start by typing \(F^n:\mathcal A\to\mathcal B\) and verifying additivity. Give, for each short exact \(0\to A\to B\to C\to0\), a map \(\delta^n:F^n(C)\to F^{n+1}(A)\). Check exactness not just at one degree but at every term in the assembled long sequence. Then check that maps of short exact sequences induce commutative boundary squares. Only after those tests is “cohomological delta functor” warranted.[1]

If universality is claimed, inspect the separate extension condition. Stacks's effaceability lemma assumes that for every object \(A\) and each \(n>0\), there is a monomorphism \(u:A\to B\)—which may depend on both \(A\) and \(n\)—such that \(F^n(u)\) is zero. This gives the sufficient mechanism for extending a degree-zero transformation inductively through cokernels. It is not the same as assuming every object has one globally chosen injective resolution or that every positive-degree group vanishes on every arbitrary object.[1]

For a homological candidate, redo the typing before transferring the argument. A short exact sequence of chain complexes gives \(H_i(C)\to H_{i-1}(A)\); a short exact sequence of cochain complexes gives \(H^i(C)\to H^{i+1}(A)\). That opposing index direction is the quick diagnostic for whether a Tor claim has been slipped into a cohomological formula.[4][6]

Knowledge Transfer

The sheaf case transfers a useful recognition protocol to module Ext: fix the varying input category, identify degree zero, follow a short exact input into a natural long exact output, and ask whether effaceability or a derived-functor theorem warrants universality. The groups themselves and their computations differ; the form of reasoning does not. Conversely, the module example warns sheaf-cohomology users to state variance: \(\operatorname{Hom}_R(M,-)\) and \(\operatorname{Hom}_R(-,N)\) are not interchangeable functors.[1][2][3]

The same protocol identifies the limit of transfer to Tor. Tensor product's left-derived groups have analogous exact-sequence power but a homological boundary lowering degree. The abstract lesson is therefore to preserve typed orientation, exactness and naturality, not merely to spot a familiar Greek letter or a long row of groups.[6][4]

Examples

Sheaf cohomology over a fixed space. Let \(X\) be fixed and \(\mathcal F\) vary among abelian sheaves on \(X\). The host categories are \(\mathrm{Ab}(X)\) and \(\mathrm{Ab}\); \(F^0=\Gamma(X,-)\) takes global sections and \(F^n=H^n(X,-)\) for higher \(n\). Given \(0\to\mathcal F'\to\mathcal F\to\mathcal F''\to0\), the boundary sends \(H^n(X,\mathcal F'')\) to \(H^{n+1}(X,\mathcal F')\). Naturality and long exactness are part of the cohomology construction, and Stacks identifies this as a universal delta functor after constructing cohomology by injective resolutions.[2][1]

Mapped back: The abelian categories, graded family, arbitrary short exact sheaf input, degree-raising natural boundary, exact output and universal subtype all have explicit occupants; the geometric space and sheaf-computation method are accents.

Module Ext with one argument fixed. Fix a ring \(R\) and an \(R\)-module \(M\), and vary a second module \(N\). Now \(F^0(N)=\operatorname{Hom}_R(M,N)\) and \(F^n(N)=\operatorname{Ext}^n_R(M,N)\). Each \(0\to N'\to N\to N''\to0\) gives a boundary \(\operatorname{Ext}^n_R(M,N'')\to\operatorname{Ext}^{n+1}_R(M,N')\) inside a natural long exact sequence. Injective resolutions in the varying second argument compute this right-derived family; positive Ext vanishes on injective targets, supporting the universal extension criterion.[3][5][1]

Mapped back: The module category replaces the sheaf category, Hom replaces global sections and Ext replaces sheaf cohomology, but the same graded/natural/long-exact cohomological structure remains. Fixing the other argument would reverse variance and must not be silently counted as the same typed example.

Structural Tensions

  • Axiomatic portability versus constructive computability. The abstract delta axioms let one compare or uniquely identify suitable graded theories across sheaves and modules without selecting a resolution; by themselves they do not compute a particular \(F^n(A)\). A concrete injective resolution gives computational access but introduces existence and resolution-management obligations specific to the category. Diagnostic: Is the task to prove a comparison from the axioms or calculate a group, and which construction hypotheses have actually been met?[1][2][3]
  • Broad membership versus degree-zero determination. The basic axioms admit more graded families; the universal extension property narrows the cases but grants uniqueness from degree zero. Requiring the latter when effaceability cannot be established loses legitimate nonuniversal cases; omitting it loses the strong comparison theorem. Diagnostic: Has this family passed an effaceability or direct extension test, or should the conclusion remain only “delta functor”?[1]

Structural–Framed Character

Evaluative weight. “Universal” is a mathematical property with a proof obligation, not praise for a useful theory. Human-practice dependence. A mathematician chooses categories, variance and indexing, but exactness and naturality are formal conditions once chosen. Institutional origin. The concept belongs to homological-algebra practice rather than being constituted by an institution's rule. Vocabulary travel. The word “delta” and exact-sequence language travel across sheaves, modules and derived settings, but only the complete typed pattern licenses the name. Import versus recognition. It is legitimate to recognize an unfamiliar graded theory as a delta functor after checking its boundaries and naturality; importing the label to any collection of long exact rows is not enough.[1]

Its character: near the structural end of the structural–framed spectrum within its typed mathematical setting: the exactness and naturality axioms decide membership independently of institutional approval or evaluative preference. It is nonetheless domain-specific because abelian categories, additive functors and degree-oriented boundaries remain constitutive rather than incidental vocabulary. Its portable “graded response with natural correction maps” skeleton might be a future-prime question, but the present identity remains tied to those mathematical conditions.

Structural Core vs. Domain Accent

The core is a typed nonnegative graded additive family, short exact inputs, oriented natural connecting maps and long exact outputs. “Universal” adds a genuine subtype-level property. Sheaves versus modules, global sections versus Hom, and particular resolution machinery are accents of the instances, not constitutive of the basic definition.[1][2][3]

The portable skeleton—coherent corrections linking levels of a graded theory when an input decomposes—could invite a future-prime inquiry, but no existing prime parent is asserted from that analogy. Tor demonstrates why even within mathematics the index orientation must stay explicit: similarity of use is not identity of formula.[6][4]

This entry presupposes Exact sequence and presupposes Functor.

The proposed DAG records composition/presupposes relations to live Functor and Exact Sequence, not subsumption. Each \(F^n\) is a functor, and the input/output exactness conditions require exact sequences, but the delta functor is neither one component functor nor one sequence. Derived Functor is a related construction whose suitable right-derived output carries this structure; its direction is not a strict parent of the axiom system.[1][5]

Relationships to Other Abstractions

Local relationship map for Delta FunctorParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Delta FunctorDOMAINDomain-specific abstraction: Exact sequence — presupposesExact sequenceDOMAINDomain-specific abstraction: Functor — presupposesFunctorDOMAIN

Current abstraction Delta Functor Domain-specific

Parents (2) — more general patterns this builds on

  • Delta Functor presupposes Exact sequence Domain-specific

    The structure responds to each short exact sequence by a long exact sequence.

  • Delta Functor presupposes Functor Domain-specific

    Each graded component is an additive functor and the boundary maps are natural in exact-sequence inputs.

Hierarchy paths (5) — routes to 5 parentless roots

Neighborhood in Abstraction Space

Delta Functor sits in a sparse region of the domain-specific corpus (75th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Categories, Sheaves & Homotopy (21 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With

  • Derived functor: a resolution/derived-category construction; its output may instantiate a delta functor, but the construction is not the axiom system.[5]
  • Universal delta functor: the stronger degree-zero extension property, not a requirement of the basic identity.[1]
  • Tor under cohomological notation: a left-derived, homologically indexed counterpart with degree-lowering boundary.[6][4]
  • Contravariant Ext in the first argument: a legitimate separate variance convention; \(\operatorname{Ext}^n_R(M,-)\) here varies the second argument.[3]
  • One natural long exact sequence: an output for one input, not the family-wide assignment and naturality axiom.[1]

References

[1] The Stacks Project, “Cohomological delta-functors,” Homological Algebra §12.12, tag 010P, especially Definition 12.12.1 (axioms), Definition 12.12.3 (universality), and Lemmas 12.12.4–5 (effaceability and uniqueness). registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x ↩y ↩z ↩27 ↩28

[2] The Stacks Project, “Cohomology of sheaves,” §20.2, tag 01DZ, formulas 20.2.0.1–0.4 and accompanying universal-delta-functor statements. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[3] The Stacks Project, “Ext groups,” Derived Categories §13.27, tag 06XP, especially long exact sequences in each argument and injective/projective resolution computation. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j

[4] The Stacks Project, “Complexes,” Homological Algebra §12.13, tag 010V, Lemmas 12.13.6 and 12.13.12 for degree-lowering homology and degree-raising cohomology boundaries. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g

[5] The Stacks Project, Lemma 13.16.6, tag 05TE, canonical delta-functor structure on right derived functors and its conditional universality. registry ↩a ↩b ↩c ↩d ↩e

[6] The Stacks Project, “Computing Tor,” More on Algebra §15.58, tag 064F, Tor as the left-derived tensor family. registry ↩a ↩b ↩c ↩d ↩e ↩f