Categorical, Homological, and Local-to-Global Language

Modern QFT sources often specify more than fields, equations, and symmetry groups. They also specify which mathematical objects are allowed, which maps compare them, which constructions must be preserved, which maps count as equivalences, and how a stated local-to-global condition reconstructs data on larger regions from data on smaller ones. This chapter is a bounded vocabulary bridge for reading those specifications accurately.

Begin with categories, functors, natural transformations, and universal properties. Then choose the branch that matches the language in front of you: monoidal language for a coherent tensor-like operation, duals, and exchange; homological language for complexes and weak equivalences; or sheaf and cosheaf language for local-to-global reconstruction. A monoidal product acquires a physical interpretation as parallel or disjoint composition only with additional input. Read all three branches before using the final boundary map to compare derived, higher-categorical, and factorization frameworks.

The entire chapter is optional advanced preparation. It does not gate the earlier Mathematical Methods chapters, and it does not attempt to develop a theorem-first formulation of QFT.

The links below describe the intended curriculum routes and destinations. A link identifies where a fuller treatment belongs; it does not by itself assert that the destination currently contains developed material.

Check your preparation · Review the chapter

Enter this chapter

The overview itself has no hard prerequisite. Use the following checks to decide where to begin.

Check	Ready evidence	Where the fuller treatment belongs
Type a composite	From $X\xrightarrow{f}Y\xrightarrow{g}Z$, write $g\circ f:X\to Z$ and state the identity and associativity laws.	Categories, Functors, Natural Transformations, and Universal Properties
Test naturality	Given $F,G:\mathcal C\to\mathcal D$ and maps $\eta_X:F(X)\to G(X)$, write the square that must commute for every $f:X\to Y$.	Categories, Functors, Natural Transformations, and Universal Properties
Separate two operations	Explain why morphism composition $g\circ f$ and a monoidal product $X\otimes Y$ have different input types.	Begin with the category page, then use Monoidal, Rigid, and Braided Language.
Read a complex	For a short chain complex, identify cycles, boundaries, homology, and exactness at the middle term.	Chains, Homology, Cohomology, and Exact Sequences
Follow an inclusion of opens	For $U\subseteq V$, decide whether the induced arrow is a restriction $F(V)\to F(U)$ or an assembly map $G(U)\to G(V)$.	Begin with the category page, then use Presheaves, Sheaves, Cosheaves, and Čech Descent.

Failure on one row identifies a local gap; it is not a reason to take the whole chapter as a prerequisite.

If a check is uncertain, try it once on the smallest nontrivial example before consulting the indicated fuller-treatment topic. Being able to repeat a definition without applying it is not yet ready evidence.

Choose a route

In this table, an arrow means a hard dependency. A plus sign means that both inputs are required. The three branches after the category foundation are parallel: their order relative to one another is a choice, not a prerequisite claim.

Need or signal in a QFT source	Route through this chapter	Mathematical QFT continuation
The categorical vocabulary itself is unfamiliar	Categories and functors	Continue only when a physical framework requires it.
Globally hyperbolic spacetimes and admissible embeddings are assigned algebras covariantly	Categories and functors	Locally Covariant QFT as a Functor
Tensor categories, dual objects, intertwiners, or braidings appear	Categories and functors → monoidal language	Endomorphisms, Intertwiners, and Tensor Products
Complexes, resolutions, ghosts, quasi-isomorphisms, or derived spaces appear	Chains and cochains + categories and functors → homotopy and quasi-isomorphisms	Derived Critical Loci and Derived Gauge Quotients
Observables are attached to open regions and a restriction, assembly, or descent condition is claimed	Categories and functors → sheaves, cosheaves, and Čech descent; chains and cochains are recommended	Weiss Descent and Local-to-Global Observables
Several modern QFT frameworks must be compared without conflating them	Categories and functors → monoidal language + homological language + local-to-global language → boundary map	QFT Frameworks, Object Classes, and Typed Maps

These links identify the first specialist continuation for each route; they do not assert that this chapter supplies every prerequisite of the destination. The chapter navigation order used in the guide below is a useful complete reading order, not one undifferentiated prerequisite chain.

How the five pages fit together

Each page answers a different structural question.

Axis	Question	Data to keep visible	Characteristic check
Typed composition	What are the objects and composable maps?	Domains, codomains, identities, composition, functoriality, naturality	Does every displayed composite type-check?
Monoidal structure	What coherent tensor-like operation is supplied?	Tensor product, unit, associators, unitors, duals, braiding	Is the claimed structure monoidal, rigid, braided, or symmetric?
Homological comparison	When do two complexes encode the same derived information?	Differential, chain map, homotopy, homology, quasi-isomorphism	Which comparison relation has actually been proved?
Local-to-global flow	How does information move along inclusions and covers?	Variance, restriction or assembly maps, overlaps, limits or colimits	Is the claim presheaf, sheaf, cosheaf, Čech, or stronger descent?
Framework selection	Which specialist language is genuinely required?	Objects, maps, weak equivalences, monoidal data, descent rule	Which downstream theorem framework provides the specialist treatment?

The hard dependency graph is therefore:

• the category foundation precedes each of the monoidal, homological, and local-to-global branches;
• the homological branch also requires the earlier chains-and-cochains page;
• the local-to-global branch recommends, but does not require, that page;
• all three branches precede the final boundary map.

Shared vocabulary and invariant checks

Typed composition and naturality

A category $\mathcal C$ has objects, morphisms, identity morphisms, and associative composition. For composable maps

$$X\xrightarrow{f}Y\xrightarrow{g}Z,$$

$g\circ f$ means apply $f$ first and then $g$. A functor $F:\mathcal C\to\mathcal D$ preserves identities and composition:

$$F(\operatorname{id}_X)=\operatorname{id}_{F(X)}, \qquad F(g\circ f)=F(g)\circ F(f).$$

A natural transformation $\eta:F\Rightarrow G$ is not an arbitrary objectwise family. Its components $\eta_X:F(X)\to G(X)$ must satisfy, for every $f:X\to Y$,

$$G(f)\circ\eta_X=\eta_Y\circ F(f).$$

A universal property is likewise a typed statement, normally expressed by existence and uniqueness of a factorization. It characterizes an object only up to the appropriate unique isomorphism compatible with the universal data; it does not make two presentations literally equal.

These category, functor, naturality, and universal-property conventions follow Leinster 2014, Introduction, pp. 1–8, and Chapter 1, §§1.1–1.3, pp. 9–40.

Monoidal, rigid, braided, and symmetric are different

A monoidal category adds a bifunctor $\otimes:\mathcal C\times\mathcal C\to\mathcal C$, a unit object $\mathbf 1$, and coherent associativity and unit isomorphisms. Rigidity adds left and right duals with evaluation and coevaluation maps satisfying the zig-zag, or snake, identities. A braiding adds coherent natural isomorphisms

$$c_{X,Y}:X\otimes Y\longrightarrow Y\otimes X.$$

Symmetry is stronger: it requires $c_{Y,X}\circ c_{X,Y}=\operatorname{id}_{X\otimes Y}$. Thus a braiding is neither literal equality nor automatically a symmetry. None of this structure alone proves a physical tensor-factorization, locality, statistics, or unitarity statement.

For the monoidal, duality, rigidity, and braiding distinctions, see Etingof, Gelaki, Nikshych, and Ostrik 2015, §§2.1, 2.10, and 8.1, PDF.

Chain homotopy and quasi-isomorphism are different

For chain complexes, this volume uses differentials $\partial_n:C_n\to C_{n-1}$ with $\partial_n\partial_{n+1}=0$. For cochain complexes it uses $\delta^n:C^n\to C^{n+1}$. If $f,g:C_\bullet\to D_\bullet$ are chain maps, a chain homotopy consists of maps $h_n:C_n\to D_{n+1}$ such that

$$(f-g)_n = \partial_{D,n+1}h_n+h_{n-1}\partial_{C,n}.$$

Chain-homotopic maps induce the same map on homology. A quasi-isomorphism is a chain map inducing isomorphisms on every homology group. A chain-homotopy equivalence is therefore a quasi-isomorphism, but the converse is false in general: a quasi-isomorphism need not have a chain-level inverse, even up to homotopy. Derived language keeps track of complexes and formally treats selected weak equivalences as invertible; it is not shorthand for taking cohomology and discarding the complex.

The chain-complex, homotopy, quasi-isomorphism, and cone comparison is treated in Weibel 1994, Chapter 1, §§1.1 and 1.4–1.5, pp. 2–23, PDF.

Presheaves and cosheaves reverse the flow

Let $\operatorname{Open}(X)$ be the category of open subsets of $X$, with inclusions as arrows. A presheaf is contravariant,

$$F:\operatorname{Open}(X)^{\mathrm{op}}\longrightarrow\mathcal C,$$

so $U\subseteq V$ gives a restriction map $F(V)\to F(U)$. A precosheaf is covariant, so the same inclusion gives an assembly map $G(U)\to G(V)$. A sheaf asks compatible local data to glue uniquely, while a cosheaf expresses global data as a suitable colimit of local data. These conditions require the relevant limits or colimits in the target category; compatible local data do not glue merely because a cover has been written down.

Čech constructions organize intersections of a cover and the compatibility maps among them. Ordinary sheaf or cosheaf conditions, Čech calculations, Weiss descent, and homotopy-coherent descent are related but not interchangeable claims.

The variance, Čech, and sheaf/cosheaf distinctions used here are developed in Curry 2014, Chapter 2, §§2.1–2.3, and §7.2.

Derived, higher, and factorization frameworks are not synonyms

“Derived” signals that homological or homotopical information and weak equivalences are retained. “Higher-categorical” signals morphisms between morphisms, and further coherently composable levels. The higher-categorical stopping boundary is summarized in Riehl 2012, pp. 1–6, PDF.

A prefactorization algebra assigns data $\mathcal F(U)$ to open regions together with products

$$\mathcal F(U_1)\otimes\cdots\otimes\mathcal F(U_n) \longrightarrow \mathcal F(V)$$

for pairwise disjoint $U_i\subseteq V$. In the vector-space-valued convention used here, a factorization algebra additionally requires the canonical map

$$\mathcal F(U_1)\otimes\mathcal F(U_2) \longrightarrow \mathcal F(U_1\sqcup U_2)$$

to be an isomorphism for disjoint opens and imposes the cosheaf condition for Weiss covers. It is therefore not merely an arbitrary cosheaf or a sheaf of algebras. When the values are cochain complexes, the downstream definition compares $\mathcal F(U)$ with the Čech double complex of a Weiss cover by a quasi-isomorphism; more general homotopical targets require their corresponding notion of equivalence.

For the prefactorization products, Weiss covers, and factorization descent used in this comparison, see Costello and Gwilliam 2023, §§1 and 3.

Recognizing these requirements is the stopping point here. Constructing higher categories, proving Weiss descent, building factorization algebras, and establishing physical equivalence or classification theorems belong downstream.

Exact chapter guide

1. Categories, Functors, Natural Transformations, and Universal Properties

The category foundation asks how mathematical structures and structure-preserving maps can be typed uniformly. Its scope is objects and morphisms, functors, natural transformations, commutative diagrams, and universal properties. There is no hard prerequisite.

The target capability is modest but indispensable: read a categorical statement without erasing domains, codomains, variance, or naturality. The corresponding QFT destination is Locally Covariant QFT as a Functor.

2. Monoidal, Rigid, and Braided Language

The monoidal branch asks what extra data encode a tensor-like operation, units, duals, and coherent exchange. It hard-requires the category foundation; it distinguishes monoidal product from ordinary composition, rigidity from strictness, and braiding from symmetry.

The target capability is to identify exactly which tensor-categorical structure a claim uses without importing a physical interpretation that has not been proved. The corresponding QFT destination is Endomorphisms, Intertwiners, and Tensor Products.

3. Chain Homotopy, Quasi-Isomorphisms, and Derived Vocabulary

The homological branch is intended to ask how complexes can be compared more flexibly than by strict isomorphism. It hard-requires both the category foundation and Chains, Homology, Cohomology, and Exact Sequences.

The target capability is to distinguish equality, isomorphism, chain homotopy, chain-homotopy equivalence, and quasi-isomorphism, and to understand why derived constructions specify which maps are treated as weak equivalences. The corresponding QFT destination is Derived Critical Loci and Derived Gauge Quotients.

4. Presheaves, Sheaves, Cosheaves, and Čech Descent

The local-to-global branch is intended to ask how data vary over open regions and when local information reconstructs global information. It hard-requires the category foundation; the earlier chains-and-cochains page is recommended preparation for Čech complexes and derived variants.

The target capability is to track variance, state a gluing or assembly condition with the needed target-category structure, and separate ordinary Čech language from stronger descent claims. The corresponding QFT destination is Weiss Descent and Local-to-Global Observables.

5. Derived, Higher, and Factorization Frameworks: a Boundary Map

The final boundary topic is intended to ask which modern framework is actually being invoked by a QFT claim. It hard-requires the monoidal, homological, and local-to-global branches; none is a substitute for another.

The target capability is framework triage: identify objects, maps, weak equivalences, monoidal data, and the local-to-global rule, then stop before specialist theorems. The corresponding QFT destination is QFT Frameworks, Object Classes, and Typed Maps.

Chapter synthesis and stopping boundary

When a modern QFT framework is proposed, the following five questions provide a comparison checklist:

1. What are its objects, and which morphisms are allowed?
2. Which assignments are functorial, and what do natural transformations compare?
3. Which maps are isomorphisms, equivalences, chain-homotopy equivalences, or quasi-isomorphisms?
4. What monoidal or disjoint-union structure is present, and are duals, braidings, or symmetries part of the data?
5. How does information move along inclusions, and which cover or descent condition reconstructs larger regions?

Before transferring a claim, make explicit every axis that the claim actually uses; a framework need not carry all five structures. If a used axis is left unspecified, the claim is not yet precise enough to transfer between frameworks. In particular, category vocabulary alone does not establish a physical equivalence, a monoidal structure does not prove locality, and ordinary Čech gluing does not prove Weiss or homotopy-coherent descent.

This chapter stops before derived QFT, higher-category constructions, Weiss-descent theorems, factorization-algebra constructions, and physical existence, equivalence, or classification results. Those topics require their own hypotheses and belong in Mathematical QFT.

Review the chapter

Retrieval. Type a concrete category with two or three objects, a functor, a natural transformation, and a universal-property statement. Success means that domains, codomains, identity and composition laws, a naturality square, and an existence-and-uniqueness clause are all explicit. Relevant repair topic: category foundation.

Explanation. Explain why a universal property characterizes an object up to a unique compatible isomorphism rather than describing “a construction that works everywhere.” Relevant repair topic: category foundation.

Derivation check. Starting from $(f-g)_n=\partial_{D,n+1}h_n+h_{n-1}\partial_{C,n}$, show that chain-homotopic maps induce the same map on homology. Success requires the grading, differential direction, and use of cycles and boundaries to be correct. Relevant repair topics: chain homotopy and quasi-isomorphisms and the chains-and-cochains prerequisite.

Representation change. For $U\subseteq V\subseteq W$, translate the inclusions into presheaf restriction maps and cosheaf assembly maps, then verify their composition laws. Relevant repair topic: presheaves, sheaves, and cosheaves.

Comparison. Distinguish morphism composition, monoidal tensor product, braiding, and symmetry. Success means giving the source and target of each relevant map and applying the double-braid test for symmetry. Relevant repair topic: monoidal, rigid, and braided language.

Transfer. For a two-open cover $U\cup V$, state the sheaf and cosheaf tests separately. The sheaf comparison is

$$F(U\cup V) \longrightarrow F(U)\times_{F(U\cap V)}F(V),$$

whereas the cosheaf comparison is

$$G(U)\sqcup_{G(U\cap V)}G(V) \longrightarrow G(U\cup V).$$

State when each map should be an isomorphism, and do not assume that the target category has the required pullback or pushout. Relevant repair topic: presheaves, sheaves, and cosheaves.

Failure diagnosis. Reject the claim “every quasi-isomorphism has a chain-level inverse.” State what a quasi-isomorphism actually guarantees and what stronger relation would supply an inverse up to homotopy. Relevant repair topic: chain homotopy and quasi-isomorphisms.

Synthesis. Profile a proposed QFT framework by listing its objects, morphisms, weak equivalences, monoidal or disjoint-union structure, and local-to-global rule. Then select the Mathematical QFT destination designated for the missing specialist theorem. Relevant repair topics: the boundary map and whichever branch supplied an incomplete axis.

Continue into Mathematical QFT

Each link below designates the intended specialist continuation. It does not assert that the destination currently contains developed material or that this chapter alone satisfies all of its prerequisites.

If the next claim concerns…	Continue with…
functorial assignments from spacetimes and embeddings	Locally Covariant QFT as a Functor
tensor products, intertwiners, duality, and sector structure	Endomorphisms, Intertwiners, and Tensor Products
derived critical loci and gauge quotients	Derived Critical Loci and Derived Gauge Quotients
Weiss descent and observables reconstructed from regions	Weiss Descent and Local-to-Global Observables
comparing theorem frameworks by their object and map types	QFT Frameworks, Object Classes, and Typed Maps

For a broader route, return to the Mathematical Methods overview.

References

• Kevin Costello and Owen Gwilliam, “Factorization algebra” (2023), survey prepared for the Encyclopedia of Mathematical Physics, 2nd ed., arXiv:2310.06137v2, §§1 and 3 — prefactorization products, Weiss covers, and factorization descent.

• Justin Curry, Sheaves, Cosheaves and Applications, PhD dissertation, University of Pennsylvania (2014), arXiv:1303.3255, Chapter 2, §§2.1–2.3, and §7.2 — sheaf and cosheaf variance, Čech constructions, and the distinction between quasi-isomorphism and chain-homotopy equivalence.

• Pavel Etingof, Shlomo Gelaki, Dmitri Nikshych, and Victor Ostrik, Tensor Categories, Mathematical Surveys and Monographs 205, American Mathematical Society (2015), author-hosted final manuscript, PDF, §§2.1, 2.10, and 8.1 — monoidal structure, duals, rigidity, and braiding.

• Tom Leinster, Basic Category Theory, Cambridge Studies in Advanced Mathematics 143, Cambridge University Press (2014), arXiv:1612.09375, Introduction, pp. 1–8, and Chapter 1, §§1.1–1.3, pp. 9–40 — categories, functors, natural transformations, and universal properties.

• Emily Riehl, “The Algebra and Geometry of ∞-Categories” (2012), author-hosted note, PDF, pp. 1–6 — the higher-categorical stopping boundary.

• Charles A. Weibel, An Introduction to Homological Algebra, Cambridge Studies in Advanced Mathematics 38, Cambridge University Press (1994), Chapter 1, “Chain Complexes”, PDF, §§1.1 and 1.4–1.5, pp. 2–23 — complexes, quasi-isomorphisms, and chain homotopies.