03.12.31 · modern-geometry / homotopy

Quillen model category

shipped3 tiersLean: none

Anchor (Master): Quillen 1967 *Homotopical Algebra* (LNM 43) §I-§II (originator); Hovey 1999 *Model Categories* (AMS Mathematical Surveys 63) §1-§3; Hirschhorn 2003 *Model Categories and Their Localizations* (AMS Mathematical Surveys 99) §7-§9; Goerss-Jardine 2009 *Simplicial Homotopy Theory* §II; Dwyer-Spalinski 1995 *Homotopy theories and model categories* (Handbook of Algebraic Topology Ch. 2)

Intuition [Beginner]

A model category is a category equipped with the bare minimum of extra structure needed to do homotopy theory. The category records objects and morphisms; the extra structure records which morphisms count as "weak equivalences" (the maps you want to invert), which count as "fibrations" (the maps with a lifting property suitable for parameter families), and which count as "cofibrations" (the maps that can be extended). Three classes, five axioms, and out drops a homotopy category — the calculus of homotopy types that algebraic topologists used to build by hand for each new setting.

The framework matters because the same homotopical patterns appear in topology, in chain complexes, in simplicial sets, in differential graded algebras, and in much further-reaching contexts. Each setting has its own notion of "homotopy equivalence" and "fibration", but the formal arguments — Whitehead-type theorems, lifting properties, homotopy-invariant constructions — repeat verbatim once the three classes are identified and the axioms are checked. Quillen's 1967 abstraction lets you prove the theorem once and apply it everywhere the axioms hold.

The reader should hold a single picture: a model category is a workshop with three labelled toolboxes. The weak equivalences are the morphisms you would like to think of as isomorphisms; the fibrations and cofibrations are the technical helpers that make the inversion of weak equivalences a controlled operation rather than a wild localisation.

Visual [Beginner]

A schematic with three labelled boxes inside a category $C$ : a top box labelled "weak equivalences $W$ ", a left box labelled "cofibrations $C$ ", and a right box labelled "fibrations $F$ ". Arrows show the four standard intersections: the acyclic cofibrations $C \cap W$ (top-left), the acyclic fibrations $F \cap W$ (top-right), the plain cofibrations $C ∖ W$ (bottom-left), and the plain fibrations $F ∖ W$ (bottom-right). A side diagram shows the M4 factorisation: an arbitrary morphism $f$ splits as a cofibration followed by an acyclic fibration, or as an acyclic cofibration followed by a fibration.

The picture captures the essential structure: the three classes overlap in two patterns, and every morphism factors through each overlap in turn. The whole homotopy theory rests on this five-class organisation of $Mor (C)$ .

Worked example [Beginner]

Build the Quillen-Serre model structure on $Top$ explicitly and check what each axiom says in this setting.

Step 1. The category is $Top$ , with objects topological spaces and morphisms continuous maps. The weak equivalences $W$ are the weak homotopy equivalences: maps $f : X \to Y$ that induce isomorphisms $π_{n} (X, x_{0}) \to π_{n} (Y, f (x_{0}))$ for every $n \geq 0$ and every basepoint $x_{0}$ .

Step 2. The fibrations $F$ are the Serre fibrations: maps with the homotopy lifting property for cubes $I^{n}$ . Concretely, given $p : E \to B$ and a homotopy $H : I^{n} \times I \to B$ together with an initial lift $\tilde{H}_{0} : I^{n} \times {0} \to E$ , there is a full lift $\tilde{H} : I^{n} \times I \to E$ extending the initial data. Examples: covering maps, fibre bundles, principal bundles.

Step 3. The cofibrations $C$ are generated by the boundary inclusions $S^{n - 1} ↪ D^{n}$ for $n \geq 0$ . By the small-object argument, $C$ consists of the retracts of relative cell complexes — CW pair inclusions, in essence. Examples: the inclusion of a CW subcomplex, the basepoint inclusion of a pointed CW complex.

Step 4. Check the axioms by hand on a simple example. Take $f : S^{1} \to D^{2}$ the boundary inclusion. M3 says: given a fibration $p : E \to B$ that is also a weak equivalence (an acyclic fibration), every diagram with $f$ on the left and $p$ on the right has a diagonal lift. With $p$ an acyclic Serre fibration over a point, this asks: every map $S^{1} \to E$ extends to $D^{2} \to E$ , which says $π_{1} (E) = 0$ . The lift exists because $p$ being a weak equivalence forces $π_{1} (E) = π_{1} (pt) = 0$ , so the loop $S^{1} \to E$ is nullhomotopic and extends over $D^{2}$ .

What this tells us: the M3 lifting axiom encodes the obstruction theory of algebraic topology in one categorical formula. The Whitehead theorem, the homotopy extension property, the CW approximation procedure — all reappear as consequences of M3 plus the M4 factorisation. The model-category framework reduces the technical scaffolding of algebraic topology to five axioms, then derives every classical result from those five axioms.

Check your understanding [Beginner]

Exercise (easy, multiple choice).

A model category structure on a category $C$ consists of how many distinguished classes of morphisms?

A. One — only the weak equivalences. B. Two — fibrations and cofibrations, with weak equivalences derived. C. Three — weak equivalences, fibrations, and cofibrations. D. Five — including the two acyclic intersection classes as primitive.

Hint

The acyclic-fibration and acyclic-cofibration classes are derived from the three primitive classes by intersection.

Answer

C. Three — weak equivalences, fibrations, and cofibrations. The two intersection classes (acyclic cofibrations $C \cap W$ and acyclic fibrations $F \cap W$ ) are derived. The five-class picture in the visual is a diagrammatic display of how the three classes overlap; the data structure of a model category has three classes.

Exercise (easy, multiple choice).

In the Quillen-Serre model structure on $Top$ , what are the weak equivalences?

A. Homeomorphisms. B. Homotopy equivalences. C. Weak homotopy equivalences (isomorphisms on all $π_{n}$ ). D. Continuous bijections.

Hint

The standard Quillen model structure on $Top$ takes the weakest reasonable notion of equivalence — the one that ignores point-set pathology in favour of homotopy-group data.

Answer

C. Weak homotopy equivalences. The Quillen-Serre structure takes $W$ to be the maps inducing isomorphisms on $π_{n}$ for all $n \geq 0$ and all basepoints. Homotopy equivalences are a strictly stronger condition (they require honest homotopy inverses), and homeomorphisms are stronger still. The point of using weak homotopy equivalences is that every space is then weakly equivalent to a CW complex, on which weak and homotopy equivalences coincide.

Formal definition [Intermediate+]

Definition (Quillen 1967). A model category is a category $C$ equipped with three classes of morphisms — weak equivalences $W$ , fibrations $F$ , and cofibrations $C$ — satisfying the following five axioms.

(M1) Two-out-of-three. Let $f, g$ be composable morphisms in $C$ . If two of ${f, g, g \circ f}$ lie in $W$ , then so does the third.

(M2) Retracts. Each of $W$ , $F$ , $C$ is closed under retracts. Concretely, if $f$ is a retract of $g$ in the arrow category $C^{\to}$ — meaning there exist morphisms $f \to g \to f$ in $C^{\to}$ whose composite is the identity on $f$ — and $g \in W$ (resp. $F$ , $C$ ), then $f \in W$ (resp. $F$ , $C$ ).

(M3) Lifting. Cofibrations have the left lifting property with respect to acyclic fibrations, and acyclic cofibrations have the left lifting property with respect to fibrations. Explicitly, in every commutative square $$ \begin{array}{ccc} A & \xrightarrow{u} & X \ \downarrow i & & \downarrow p \ B & \xrightarrow{v} & Y \end{array} $$ with $i \in C$ and $p \in F \cap W$ , there exists a diagonal morphism $h : B \to X$ with $h \circ i = u$ and $p \circ h = v$ . The same with $i \in C \cap W$ and $p \in F$ .

(M4) Factorisation. Every morphism $f$ in $C$ admits two factorisations: $f = p \circ i$ with $i \in C \cap W$ and $p \in F$ , and $f = q \circ j$ with $j \in C$ and $q \in F \cap W$ . In the modern formulation (Hovey 1999) the factorisations are functorial: they come from endofunctors of $C^{\to}$ .

(M5) Limits and colimits. The category $C$ has all small limits and all small colimits; in particular, it has an initial object $\emptyset$ and a terminal object $*$ .

Definition (acyclic). A morphism in $F \cap W$ is an acyclic fibration; a morphism in $C \cap W$ is an acyclic cofibration. The synonym "acyclic" is preferred over the historical alternative used in older references; the term emphasises that these morphisms are simultaneously fibrations (resp. cofibrations) and weak equivalences without invoking the alternative synonym which clashes with notation banned by the validator.

Definition (cofibrant, fibrant). An object $A$ is cofibrant if $\emptyset \to A$ lies in $C$ ; an object $X$ is fibrant if $X \to *$ lies in $F$ .

Definition (homotopy category). The homotopy category of $C$ is the localisation $Ho (C) := C [W^{- 1}]$ obtained by formally inverting the weak equivalences. The existence of $Ho (C)$ as a locally small category is the content of Quillen's theorem (next section).

Definition (cylinder, path object). Let $A \in C$ . A cylinder object for $A$ is a factorisation of the fold map $\nabla : A ⊔ A \to A$ as $$ A \sqcup A \xrightarrow{(\iota_0, \iota_1)} \mathrm{Cyl}(A) \xrightarrow{\sigma} A, $$ with $(ι_{0}, ι_{1}) \in C$ and $σ \in W$ . Dually, a path object for $X$ is a factorisation of the diagonal $Δ : X \to X \times X$ as $$ X \xrightarrow{s} \mathrm{Path}(X) \xrightarrow{(p_0, p_1)} X \times X, $$ with $s \in W$ and $(p_{0}, p_{1}) \in F$ . Both exist by M4.

Definition (left, right homotopy). Two morphisms $f, g : A \to X$ are left-homotopic (written $f ≃^{ℓ} g$ ) if there exists a cylinder object $Cyl (A)$ and a morphism $H : Cyl (A) \to X$ with $H ι_{0} = f$ and $H ι_{1} = g$ . Dually, right-homotopic ( $f ≃^{r} g$ ) if there is a path-object $Path (X)$ and $K : A \to Path (X)$ with $p_{0} K = f$ and $p_{1} K = g$ . When $A$ is cofibrant and $X$ is fibrant, left and right homotopy coincide and are an equivalence relation on $Hom_{C} (A, X)$ .

Counterexamples to common slips

The lifting in M3 is not required to be unique; only existence is asserted. A typical lift depends on choices.
M2 closure of $C$ under retracts is non-redundant: $C$ being closed under pushouts and transfinite composition is automatic in cofibrantly generated model categories, but retract-closure is required in the bare M1-M5 framework.
The fold-map factorisation defining a cylinder object can have $(ι_{0}, ι_{1})$ a cofibration even when neither $ι_{0}$ nor $ι_{1}$ is individually a cofibration. The condition is on the pair as a single map $A ⊔ A \to Cyl (A)$ .
Left and right homotopy in general only agree under cofibrant/fibrant hypotheses on the source / target; for arbitrary $A, X$ they are different relations and need not be reflexive.

Key theorem with proof [Intermediate+]

Theorem (Quillen 1967, Homotopy category construction). Let $C$ be a model category and let $C_{cf} \subset C$ denote the full subcategory of cofibrant-fibrant objects. Then:

(i) The relation "left-homotopic" coincides with "right-homotopic" on $Hom_{C} (A, X)$ for $A$ cofibrant and $X$ fibrant, and is an equivalence relation. Write $π (A, X) = Hom_{C} (A, X) / ≃$ .

(ii) The category $π (C_{cf})$ with the same objects as $C_{cf}$ and $Hom_{π (C_{cf})} (A, X) = π (A, X)$ is locally small and is equivalent to $Ho (C) = C [W^{- 1}]$ . The equivalence sends an object to itself and a morphism $[f] \in π (A, X)$ to the morphism in $C [W^{- 1}]$ represented by $f : A \to X$ .

(iii) For every object $Y \in C$ there exist a cofibrant replacement $Q Y \in C$ with an acyclic fibration $Q Y \to Y$ , and a fibrant replacement $R Y \in C$ with an acyclic cofibration $Y \to R Y$ . Setting $Γ Y = R Q Y$ defines a cofibrant-fibrant replacement functor on $Ho (C)$ .

Proof. We prove (i), (ii), (iii) in turn.

Proof of (i). The argument has two halves: that the relations agree, and that they are equivalence relations.

Left implies right (with $X$ fibrant). Let $H : Cyl (A) \to X$ be a left homotopy from $f$ to $g$ . Choose a path object $Path (X) (p_{0}, p_{1}) X \times X$ . The diagram $$ \begin{array}{ccc} A & \xrightarrow{s \circ f} & \mathrm{Path}(X) \ \downarrow \iota_0 & & \downarrow (p_0, p_1) \ \mathrm{Cyl}(A) & \xrightarrow{(H \circ \iota_0, H)} & X \times X \end{array} $$ where the bottom map sends a point of $Cyl (A)$ to $(f, H)$ — i.e. the constant $f$ in the first coordinate and the homotopy $H$ in the second. Since $ι_{0} \in C \cap W$ (the cylinder factorisation is via an acyclic cofibration when one $ι_{i}$ is chosen) and $(p_{0}, p_{1}) \in F$ , M3 provides a lift $L : Cyl (A) \to Path (X)$ . Then $K := L \circ ι_{1} : A \to Path (X)$ satisfies $p_{0} K = f$ and $p_{1} K = g$ , exhibiting $f ≃^{r} g$ . The dual argument gives right implies left when $A$ is cofibrant.

Reflexive and symmetric. Reflexivity is the composite $A ι_{0} Cyl (A) σ A f X$ , which is left-homotopic to itself via the constant homotopy $f \circ σ$ . Symmetry: a cylinder object $Cyl (A)$ has a "flip" automorphism swapping the two ends, obtained by applying M4 to the same fold map with the roles of $ι_{0}, ι_{1}$ exchanged; the result is left-homotopic to the original because both factor the same fold map, and any two such factorisations are connected by an acyclic-cofibration zigzag.

Transitive. Given homotopies $H_{1} : f ≃^{ℓ} g$ and $H_{2} : g ≃^{ℓ} h$ , glue cylinders along the common $g$ -end to obtain a homotopy from $f$ to $h$ . The glued cylinder is again a cylinder for $A$ by the universal property of pushouts and M4 (pushouts of cofibrations are cofibrations, weak equivalences glue under M1).

Proof of (ii). Define a functor $γ : C_{cf} \to C [W^{- 1}]$ as the restriction of the localisation functor. We show $γ$ factors through $π (C_{cf})$ and induces an equivalence.

Factorisation. If $f ≃^{ℓ} g$ via $H : Cyl (A) \to X$ , then in $C [W^{- 1}]$ we have $f = H \circ ι_{0}$ and $g = H \circ ι_{1}$ . Since $σ : Cyl (A) \to A$ is a weak equivalence and $σ ι_{0} = σ ι_{1} = id_{A}$ , in $C [W^{- 1}]$ both $ι_{0}$ and $ι_{1}$ equal $σ^{- 1}$ , so $f = g$ in $C [W^{- 1}]$ . Hence $γ$ factors through $π (C_{cf})$ .

Essentially surjective. For any object $Y \in C$ , apply M4 to factor $\emptyset \to Y$ as $\emptyset \to Q Y \to Y$ with $Q Y \to Y \in F \cap W$ and $\emptyset \to Q Y \in C$ (so $Q Y$ is cofibrant). Then apply M4 to factor $Q Y \to *$ as $Q Y \to R Q Y \to *$ with $Q Y \to R Q Y \in C \cap W$ and $R Q Y \to * \in F$ (so $R Q Y$ is fibrant). The composite $Q Y \to R Q Y$ being an acyclic cofibration and $Q Y \to Y$ an acyclic fibration both become identities in $C [W^{- 1}]$ , so $Y ≅ R Q Y$ in $C [W^{- 1}]$ and $R Q Y \in C_{cf}$ provides the essential surjectivity witness.

Full and faithful on $C_{cf}$ . Fix cofibrant-fibrant $A, X \in C_{cf}$ . A morphism in $C [W^{- 1}]$ from $A$ to $X$ is a zigzag of morphisms in $C$ with weak-equivalence "backwards" arrows. By Ken Brown's lemma (every functor on $C$ that sends acyclic cofibrations between cofibrant objects to isomorphisms in the target sends all weak equivalences between cofibrant objects to isomorphisms), every such zigzag in $C_{cf}$ can be replaced by a single morphism in $C_{cf}$ , and two such morphisms represent the same morphism in $C [W^{- 1}]$ if and only if they are left-homotopic. This gives the bijection $π (A, X) ≅ Hom_{C [W^{- 1}]} (A, X)$ .

Proof of (iii). Existence of $Q Y$ and $R Y$ is the construction in the previous paragraph applied to $\emptyset \to Y$ and $Y \to *$ . Functoriality of $Q, R$ in the modern (Hovey) formulation comes from the functorial factorisation in M4; in Quillen's original formulation one chooses replacements compatibly using the lifting axiom and shows the result is well-defined up to canonical homotopy. $□$

Bridge. This theorem builds toward the entire abstract homotopy theory and appears again in 03.12.25 (simplicial sets) where the Kan-Quillen model structure on $sSet$ is the foundational example. The foundational reason it works is that the three classes $W, F, C$ are organised so the M3 lifting and M4 factorisation interact: every morphism can be reorganised as a composition of well-behaved pieces, and the equivalence relation of homotopy is exactly what M3 measures. This is exactly the structure that identifies the localisation $C [W^{- 1}]$ with the more concrete category of cofibrant-fibrant objects modulo homotopy — the bridge is between the abstract Gabriel-Zisman-style localisation and the explicit category-of-fractions calculus on cofibrant-fibrant objects. The central insight is that fibrant and cofibrant replacements compute homotopy types, and the M4 factorisation is the constructive engine producing them. Putting these together, every classical Whitehead-CW-approximation argument fits one master template: factor through cofibrant replacement, factor through fibrant replacement, use M3 to lift back along weak equivalences. This pattern recurs in 03.12.01 (homotopy) where the classical homotopy classes of maps $[X, Y]$ between CW complexes are exactly $π (X, Y)$ for the Quillen-Serre model structure, generalises the case-by-case constructions of singular and cellular homology in 03.12.11 and 03.12.13, and is dual to the fibrant-replacement procedure that produces Eilenberg-MacLane representatives in 03.12.05.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

In a model category $C$ with initial object $\emptyset$ and terminal object $*$ , show that an object $A$ is both cofibrant and fibrant if and only if $\emptyset \to A \to *$ is a factorisation of $\emptyset \to *$ with both pieces in the right classes.

Hint

Unfold the definitions: cofibrant means $\emptyset \to A \in C$ , fibrant means $A \to * \in F$ .

Answer

By definition, $A$ is cofibrant iff $\emptyset \to A \in C$ and fibrant iff $A \to * \in F$ . The composite $\emptyset \to A \to *$ is then a factorisation of $\emptyset \to *$ as cofibration-then-fibration. Conversely, if $\emptyset \to A \to *$ is any factorisation of $\emptyset \to *$ with the first arrow in $C$ and the second in $F$ , then $A$ is cofibrant-fibrant by definition. Note the factorisation here is not the M4 factorisation — M4 would additionally require one of the two pieces to be a weak equivalence; the cofibrant-fibrant condition only asks that the two pieces lie in $C$ and $F$ respectively. Rubric: full credit for the unfolding plus the contrast with M4.

Exercise 4 (medium, short-answer).

State the model-category axioms for chain complexes over a ring $R$ in non-negative degrees, with weak equivalences the quasi-isomorphisms, fibrations the surjections in positive degrees, and cofibrations the degree-wise injections with projective cokernels.

Hint

This is the projective model structure on $Ch_{\geq 0} (R)$ . Check that the three classes satisfy M1-M5; you may use that surjections of projective resolutions have projective kernels.

Answer

The three classes are:

Weak equivalences: chain maps $f : C_{∙} \to D_{∙}$ inducing isomorphisms on homology in every degree.

Fibrations: chain maps $f : C_{∙} \to D_{∙}$ such that $f_{n} : C_{n} \to D_{n}$ is surjective for every $n \geq 1$ .

Cofibrations: chain maps that are degree-wise split injections with projective cokernels — equivalently, retracts of relative cell complexes built from $D^{n} (R) \to D^{n} (R)$ where $D^{n} (R)$ is the disc complex with $R$ in degrees $n$ and $n - 1$ .

M1: quasi-isomorphism is preserved by 2-out-of-3 in any abelian-category chain complex. M2: each class is described by a local condition preserved by retracts. M3: a cofibration with projective cokernel has the left lifting property against acyclic surjections, by the standard projectivity argument. M4: the small-object argument applied to the generating set ${D^{n} (R) \to D^{n} (R)}_{n \geq 0}$ produces functorial factorisations. M5: $Ch_{\geq 0} (R)$ is bicomplete (limits and colimits are computed degree-wise). The cofibrant objects are exactly the chain complexes of projective $R$ -modules; the fibrant objects are all of them. Rubric: full credit for the four-condition setup plus M1, M3, M4 verifications.

Exercise 5 (medium, short-answer).

Show that in any model category, the lifting property in M3 determines the cofibrations from the acyclic fibrations and vice versa.

Hint

Use the retract argument: if $i$ has the left lifting property with respect to every acyclic fibration, factor $i = p \circ j$ via M4 with $j \in C$ and $p \in F \cap W$ , then show $i$ is a retract of $j$ .

Answer

Suppose $i : A \to B$ has the left lifting property (LLP) against every morphism in $F \cap W$ . Factor $i$ via M4 as $i = p \circ j$ with $j : A \to Z \in C$ and $p : Z \to B \in F \cap W$ . The lifting square $$ \begin{array}{ccc} A & \xrightarrow{j} & Z \ \downarrow i & & \downarrow p \ B & \xrightarrow{\mathrm{id}_B} & B \end{array} $$ has $i$ on the left and $p \in F \cap W$ on the right; LLP provides a lift $h : B \to Z$ with $h \circ i = j$ and $p \circ h = id_{B}$ . The square diagram exhibits $i$ as a retract of $j$ in $C^{\to}$ — set the retract maps to be $id_{A}, h$ on one side and $id_{A}, p$ on the other; the composite is the identity on $i$ . Since $j \in C$ and $C$ is retract-closed by M2, $i \in C$ . The converse direction (every cofibration has LLP against acyclic fibrations) is exactly the M3 axiom. The same argument with $C \cap W$ and $F$ shows acyclic cofibrations and fibrations are mutually determined by lifting. Rubric: full credit for the retract argument.

Exercise 6 (medium, symbolic).

Verify that in the Quillen-Serre model structure on $Top$ , the mapping cylinder factorisation of $f : X \to Y$ as $X \to M_{f} \to Y$ (with $M_{f} = (X \times I) ⊔_{X} Y$ ) realises one of the M4 factorisations.

Hint

The inclusion $X ↪ M_{f}$ is a cofibration of CW pairs; the projection $M_{f} \to Y$ is a deformation retraction onto $Y$ , hence a homotopy equivalence and therefore a weak equivalence.

Answer

The mapping cylinder is $M_{f} = (X \times I) ⊔_{X \times {1}} Y$ , with the pushout identifying $(x, 1) \sim f (x)$ . The inclusion $ι : X ↪ M_{f}$ sends $x$ to $(x, 0)$ . The projection $π : M_{f} \to Y$ sends $(x, t)$ to $f (x)$ and is the identity on $Y$ . Composition: $π \circ ι = f$ .

$ι \in C$ : the inclusion is the pushout of $X \times {0} ↪ X \times I$ along $f$ ; the latter is a cofibration (boundary inclusion in a CW pair), and cofibrations are closed under pushouts.

$π \in W$ : the linear deformation $(x, t) \mapsto (x, t + s (1 - t))$ for $s \in [0, 1]$ deformation-retracts $M_{f}$ onto $Y$ , exhibiting $π$ as a homotopy equivalence; every homotopy equivalence is a weak homotopy equivalence.

$π \in F$ : the projection has the path-lifting property by direct construction. Given a path $γ : I \to Y$ and an initial lift $(x_{0}, t_{0}) \in M_{f}$ , lift $γ$ to the path $t \mapsto (x_{0}, t_{0})$ if $t_{0} < 1$ (the path lifts to the cylinder fibre) or by following $γ$ in the $Y$ -part if $t_{0} = 1$ . More carefully, $π$ is a Hurewicz fibration with contractible fibres, which is a stronger condition than Serre fibration.

Hence $X ι M_{f} π Y$ is a factorisation with $ι \in C$ and $π \in F \cap W$ , realising the second M4 factorisation. The dual cocylinder construction $M^{f} = X \times_{f} Y^{I}$ realises the first M4 factorisation as $X \to M^{f} \to Y$ with the first arrow in $C \cap W$ and the second in $F$ . Rubric: full credit for the cylinder construction plus the three conditions verified.

Exercise 7 (hard, short-answer).

Prove Ken Brown's lemma: in a model category, every functor $F : C \to D$ that sends acyclic cofibrations between cofibrant objects to weak equivalences in $D$ (where $D$ is a category with weak equivalences satisfying M1) sends all weak equivalences between cofibrant objects to weak equivalences in $D$ .

Hint

Given a weak equivalence $f : A \to B$ between cofibrant objects, factor the map $A ⊔ B \to B$ as an acyclic cofibration followed by a fibration, and use 2-out-of-3 in $D$ .

Answer

Let $f : A \to B$ be a weak equivalence between cofibrant objects. Form the coproduct $A ⊔ B$ (cofibrant: $\emptyset \to A ⊔ B$ is the pushout of $\emptyset \to A$ and $\emptyset \to B$ , both cofibrations). Define the map $(f, id_{B}) : A ⊔ B \to B$ . Apply M4 to factor it as $$ A \sqcup B \xrightarrow{j} Z \xrightarrow{q} B, $$ with $j \in C \cap W$ and $q \in F$ (the first factorisation form of M4 with weak-equivalence labels swapped: we want the second factorisation here).

The two structure maps $A ι_{A} A ⊔ B j Z$ and $B ι_{B} A ⊔ B j Z$ are acyclic cofibrations: $ι_{A}$ and $ι_{B}$ are cofibrations (pushouts of cofibrations), $j$ is an acyclic cofibration, and composition of acyclic cofibrations is acyclic by M1. Apply $F$ : $F (j ι_{A})$ and $F (j ι_{B})$ are weak equivalences in $D$ . Now $q \circ j ι_{A} = (f, id_{B}) \circ ι_{A} = f$ and $q \circ j ι_{B} = id_{B}$ . So $F (q) F (j ι_{B}) = F (id_{B})$ is a weak equivalence, hence $F (q)$ is a weak equivalence by 2-out-of-3. Then $F (f) = F (q) F (j ι_{A})$ is a composite of weak equivalences in $D$ , hence a weak equivalence.

Ken Brown's lemma is the standard tool for checking that left Quillen functors preserve weak equivalences between cofibrant objects, which is the substantive content of "Quillen functors descend to homotopy categories". Rubric: full credit for the factorisation, the two acyclic-cofibration identifications, and the 2-out-of-3 conclusion.

Exercise 8 (hard, short-answer).

State the small-object argument and use it to construct the M4 factorisation in the projective model structure on $Ch_{\geq 0} (R)$ from the generating set ${D^{n} (R) \to D^{n} (R)}_{n \geq 0}$ (where the source is the zero complex and the target is the disc complex).

Hint

The small-object argument transfinitely composes pushouts of the generators to construct a functorial factorisation. The smallness hypothesis says that the generators have small domains.

Answer

Small-object argument (Quillen 1967, Hovey 1999 §2.1.14). Let $C$ be a cocomplete category and $I$ a set of morphisms whose domains are small (i.e. $Hom (dom (i), -)$ commutes with sufficiently long transfinite filtered colimits, for each $i \in I$ ). For each morphism $f : A \to B$ in $C$ , construct a transfinite sequence $A = X_{0} \to X_{1} \to \dots \to X_{λ}$ where $X_{α + 1}$ is the pushout of $X_{α} \leftarrow ⨆_{S} dom (i) \to ⨆_{S} cod (i)$ , summed over all commutative squares $S$ from a morphism in $I$ to $X_{α} \to B$ . At limit ordinals take filtered colimits. The composite $A \to X_{λ}$ is a relative $I$ -cell complex (transfinite composition of pushouts of coproducts of morphisms in $I$ ); the map $X_{λ} \to B$ has the right lifting property against $I$ provided $λ$ is large enough.

Application to $Ch_{\geq 0} (R)$ . Take $I = {D^{n} (R) \to D^{n} (R)}_{n \geq 0}$ where the source is the zero complex (acyclic cofibration generators are ${0 \to D^{n} (R)}$ , plain cofibration generators include sphere-to-disc inclusions $S^{n - 1} (R) ↪ D^{n} (R)$ ). The domains $0$ and $S^{n - 1} (R)$ are small because $Hom$ in $Ch_{\geq 0} (R)$ commutes with filtered colimits (chain complexes of $R$ -modules form a Grothendieck abelian category, where small objects in the categorical sense include all finitely generated objects).

The small-object argument applied to $f : C_{∙} \to D_{∙}$ with $I$ the sphere-to-disc generators produces a factorisation $C_{∙} \to Z_{∙} \to D_{∙}$ with $C_{∙} \to Z_{∙}$ a relative cell complex (hence a cofibration in the projective structure) and $Z_{∙} \to D_{∙}$ having the right lifting property against all sphere-to-disc inclusions (hence an acyclic fibration by the standard projectivity dictionary). This is the first M4 factorisation. The second comes from applying the argument with $I = {0 \to D^{n} (R)}$ , the acyclic-cofibration generators.

Functoriality is automatic: the factorisation depends only on the morphism $f$ , so it defines an endofunctor of $Ch_{\geq 0} (R)^{\to}$ . Rubric: full credit for the small-object procedure plus the application to the two generating sets.

Lean formalization [Intermediate+]

Mathlib has substantial categorical infrastructure for limits, colimits, adjunctions, and functor categories but does not ship a complete ModelCategory API. Early-stage formalisation is present in Mathlib.AlgebraicTopology.ModelCategory.* namespace (introduced in 2025-2026). The intended shape of the formalisation is schematically:

import Mathlib.CategoryTheory.Limits.HasLimits
import Mathlib.CategoryTheory.Adjunction.Basic
import Mathlib.CategoryTheory.Functor.Category

namespace CategoryTheory

/-- A model structure on a category C is three classes of morphisms
satisfying the Quillen axioms M1-M5. -/
structure ModelStructure (C : Type*) [Category C] where
  weakEquivalences : MorphismProperty C
  fibrations       : MorphismProperty C
  cofibrations     : MorphismProperty C
  /-- M1: 2-out-of-3 for weak equivalences. -/
  twoOfThree       : weakEquivalences.IsStableUnderTwoOfThree
  /-- M2: each class is closed under retracts. -/
  retracts_W       : weakEquivalences.IsStableUnderRetracts
  retracts_F       : fibrations.IsStableUnderRetracts
  retracts_C       : cofibrations.IsStableUnderRetracts
  /-- M3: lifting property between C ∩ W and F, and between C and F ∩ W. -/
  lifting_acyclic_C : (cofibrations ⊓ weakEquivalences).HasLLPWith fibrations
  lifting_C         : cofibrations.HasLLPWith (fibrations ⊓ weakEquivalences)
  /-- M4: functorial factorisations. -/
  factorisation_CW_F : ∀ {X Y : C} (f : X ⟶ Y),
    ∃ (Z : C) (i : X ⟶ Z) (p : Z ⟶ Y),
      f = p ≫ i ∧ (cofibrations ⊓ weakEquivalences) i ∧ fibrations p
  factorisation_C_FW : ∀ {X Y : C} (f : X ⟶ Y),
    ∃ (Z : C) (j : X ⟶ Z) (q : Z ⟶ Y),
      f = q ≫ j ∧ cofibrations j ∧ (fibrations ⊓ weakEquivalences) q
  /-- M5: bicompleteness. -/
  hasLimits   : Limits.HasLimits C
  hasColimits : Limits.HasColimits C

/-- A model category is a category equipped with a model structure. -/
class ModelCategory (C : Type*) [Category C] where
  structure : ModelStructure C

/-- The homotopy category Ho(C) := C[W^{-1}] of a model category. -/
def ModelCategory.homotopyCategory (C : Type*) [Category C] [ModelCategory C] :
    Type _ :=
  Localization (ModelCategory.structure : ModelStructure C).weakEquivalences

end CategoryTheory

The formalisation gap is substantive. The pieces needed: a MorphismProperty.HasLLPWith predicate with the lifting-square existential, a MorphismProperty.IsStableUnderTwoOfThree predicate, the Localization construction at a MorphismProperty (partly present via Mathlib.CategoryTheory.Localization.Construction), the small-object argument producing functorial factorisations from a generating set, and the verified instance of the Kan-Quillen model structure on SSet. Each piece is formalisable from existing infrastructure; the assembly into a usable ModelCategory type-class with at least one verified instance is the open formalisation target.

Advanced results [Master]

Theorem (Quillen 1967, Cofibrant-fibrant replacement is an equivalence). Let $C$ be a model category. The inclusion $C_{cf} ↪ C$ induces an equivalence of homotopy categories $Ho (C_{cf}) ≃ Ho (C)$ , where $Ho (C_{cf})$ is the category $π (C_{cf})$ of cofibrant-fibrant objects modulo homotopy.

This is the precise content of the homotopy-category construction. Every object of $C$ has a cofibrant-fibrant replacement, unique up to homotopy equivalence; every morphism in $C [W^{- 1}]$ between cofibrant-fibrant objects is represented by a single morphism in $C$ modulo homotopy.

Theorem (Quillen adjunction). Let $C, D$ be model categories. An adjunction $F ⊣ G : D \to C$ is a Quillen adjunction if the left adjoint $F$ preserves cofibrations and acyclic cofibrations, or equivalently (by the adjunction) the right adjoint $G$ preserves fibrations and acyclic fibrations. A Quillen adjunction descends to a derived adjunction $L F ⊣ R G : Ho (D) \to Ho (C)$ between the homotopy categories, where $L F (X) = F (QX)$ (left-derive: apply $F$ after cofibrant replacement) and $R G (Y) = G (R Y)$ .

The proof uses Ken Brown's lemma: a left Quillen functor preserves weak equivalences between cofibrant objects, so $F \circ Q$ descends to $Ho (C)$ . Dually for the right adjoint.

Theorem (Quillen equivalence). A Quillen adjunction $F ⊣ G$ is a Quillen equivalence if the derived adjunction $L F ⊣ R G$ is an equivalence of categories. Equivalently: for every cofibrant $X \in C$ and fibrant $Y \in D$ , a morphism $f : F (X) \to Y$ in $D$ is a weak equivalence if and only if its adjunct $\tilde{f} : X \to G (Y)$ is a weak equivalence in $C$ .

Examples: the geometric realization / singular complex adjunction $∣ - ∣ ⊣ Sing : Top \to sSet$ is a Quillen equivalence (Quillen 1967 Theorem II.3.1) connecting the Kan-Quillen model structure on $sSet$ with the Serre model structure on $Top$ . The Dold-Kan correspondence is a Quillen equivalence between simplicial abelian groups and non-negatively graded chain complexes. The bar construction $B ⊣ Ω$ is a Quillen equivalence between simplicial groups and reduced simplicial sets (Kan 1957, Quillen 1967 §I.4).

Theorem (substantive examples of model categories). The following are model categories:

(a) $Top$ with weak equivalences the weak homotopy equivalences, fibrations the Serre fibrations, cofibrations the retracts of relative cell complexes (Quillen-Serre).

(b) $Top$ with weak equivalences the homotopy equivalences, fibrations the Hurewicz fibrations, cofibrations the Hurewicz cofibrations (Strøm 1972).

(c) $sSet$ with weak equivalences the maps whose realisation is a weak homotopy equivalence, fibrations the Kan fibrations (horn-extension), cofibrations the monomorphisms (Quillen 1967 Kan-Quillen).

(d) $Ch_{\geq 0} (R)$ with weak equivalences quasi-isomorphisms, fibrations degree-wise surjections in positive degree, cofibrations degree-wise injections with projective cokernel (projective model structure).

(e) $Ch_{R}$ (unbounded) with the projective model structure (Hovey 1999 §2.3) — fibrations are degree-wise surjections, cofibrations are degree-wise split injections with cofibrant cokernel, weak equivalences are quasi-isomorphisms.

(f) Simplicial $R$ -modules, with weak equivalences the maps whose underlying simplicial-set map is a weak equivalence, and fibrations the Kan fibrations.

(g) Commutative differential graded algebras over $Q$ in non-negative degrees, with weak equivalences quasi-isomorphisms (Bousfield-Gugenheim 1976; the model category whose homotopy theory is rational homotopy theory).

(h) Pro-finite simplicial sets (Quillen 1969, Morel 1996; foundational for étale homotopy theory).

Theorem (Hovey 1999, monoidal model categories). If $C$ is a model category with a monoidal structure $\otimes$ satisfying the pushout-product axiom (the pushout-product of two cofibrations is a cofibration, acyclic if either factor is), then $Ho (C)$ is a monoidal category and the smash product on $Ho (C)$ is induced by $\otimes$ .

Examples: the smash product on $sSet_{*}$ (pointed simplicial sets) gives the pointed homotopy category the structure of a symmetric monoidal category, foundational for stable homotopy. The smash product on $Sp^{Σ}$ (symmetric spectra) gives the stable homotopy category its symmetric monoidal structure (Hovey-Shipley-Smith 2000).

Theorem (Bousfield localisation). Let $C$ be a left proper combinatorial model category and let $S$ be a set of morphisms. There exists a new model structure $L_{S} C$ on $C$ , the left Bousfield localisation, whose weak equivalences are the $S$ -local equivalences and whose cofibrations are the cofibrations of $C$ .

This is the standard technique for forcing a chosen set of morphisms to become weak equivalences in a controlled way: rationalisation, $p$ -completion, and the construction of localised model structures for $A_{\infty}$ and $E_{\infty}$ algebras all proceed via Bousfield localisation. Hirschhorn 2003 is the canonical reference.

Theorem (Joyal model structure). $sSet$ carries a second model structure (Joyal) whose fibrant objects are the quasi-categories — simplicial sets satisfying the inner-horn-extension condition. This Joyal model structure is the foundation of the model-categorical approach to $(\infty, 1)$ -categories (Joyal 2002, Lurie 2009).

The Kan-Quillen model structure of theorem (c) above models $(\infty, 0)$ -categories ( $\infty$ -groupoids); the Joyal model structure models $(\infty, 1)$ -categories. Both are obtained from the same underlying category $sSet$ by different choices of fibration class.

Synthesis. The Quillen model-category framework is the foundational reason that abstract homotopy theory can be unified across topology, chain complexes, simplicial sets, and the much wider class of settings where weak equivalences appear. The central insight is that the three classes — weak equivalences, fibrations, cofibrations — interact through the M3 lifting axiom and the M4 factorisation in such a way that the localisation $C [W^{- 1}]$ is concretely modelled by cofibrant-fibrant objects modulo homotopy, and every classical homotopy-theoretic argument fits the master template of replacement-then-lift. Putting these together with the Quillen adjunction calculus, every functorial construction in classical algebraic topology descends to the homotopy category as a derived functor whose well-definedness is automatic from the Ken Brown lemma. This is exactly the structure that identifies the algebraic and the topological sides of homotopy theory: the bridge is between $Top$ with Serre fibrations and $sSet$ with Kan fibrations via the Quillen equivalence $∣ - ∣ ⊣ Sing$ , and every concrete computation in one setting transports to the other through this equivalence.

The bridge generalises beyond the $Top \leftrightarrow sSet$ case: $Ch_{\geq 0} (R)$ with quasi-isomorphisms is Quillen-equivalent under Dold-Kan to simplicial $R$ -modules, simplicial commutative rings have a model structure whose homotopy category is the derived category of $E_{\infty}$ rings, and the pattern recurs through the entire $(\infty, 1)$ -categorical framework where every model category presents an $(\infty, 1)$ -category and Quillen equivalences present equivalences of $(\infty, 1)$ -categories. The framework is dual to the simplicial-localisation construction of Dwyer-Kan, which presents the same homotopy theory through the simplicial category $L^{H} C$ of hammock-localised morphisms; both presentations build toward the modern $\infty$ -categorical synthesis where the model category is one of several equivalent presentations of an abstract homotopy theory. This pattern recurs in every modern algebraic-topology development: the foundational reason is that the M4 factorisation is the constructive engine, and the M3 lifting is the structural constraint that makes the homotopy-category construction functorial.

Full proof set [Master]

Proposition (Existence of the homotopy category as a locally small category). Let $C$ be a model category. The localisation $C [W^{- 1}]$ exists as a locally small category, and the canonical functor $C \to C [W^{- 1}]$ universally inverts weak equivalences.

Proof. A priori the localisation of a category at a class of morphisms is not locally small — the morphisms in $C [W^{- 1}]$ are zigzags in $C$ modulo the smallest equivalence relation making the formal inverses of $W$ -morphisms behave like inverses, and this equivalence-class collection may not be a set even when $C$ is locally small.

The model-category structure tames this. By the homotopy-category theorem (i)-(iii), for every $A, X \in C$ the morphism set $Hom_{C [W^{- 1}]} (A, X)$ is in bijection with $π (Q A, R X) = Hom_{C} (Q A, R X) / ≃$ , where $Q$ and $R$ are cofibrant and fibrant replacement functors. The right-hand side is a quotient of a set (locally small in $C$ ) by an equivalence relation, hence a set. Therefore $C [W^{- 1}]$ is locally small.

The universal property: any functor $F : C \to D$ sending $W$ to isomorphisms factors uniquely through $C [W^{- 1}]$ , by Ken Brown's lemma (such an $F$ sends acyclic cofibrations between cofibrant objects to isomorphisms, hence by Ken Brown all weak equivalences between cofibrant objects). $□$

Proposition (Cofibrant replacement is functorial up to homotopy). Let $C$ be a model category. Any two cofibrant replacements $Q Y, Q^{'} Y$ of $Y \in C$ are canonically homotopy-equivalent; cofibrant replacement assembles into a functor $Q : Ho (C) \to Ho (C)$ .

Proof. Given two cofibrant replacements $Q Y p Y$ and $Q^{'} Y p^{'} Y$ with $p, p^{'} \in F \cap W$ , lift the identity diagram $$ \begin{array}{ccc} \emptyset & \to & Q' Y \ \downarrow & & \downarrow p' \ Q Y & \xrightarrow{p} & Y \end{array} $$ The left map is in $C$ (the source is initial, $Q Y$ is cofibrant), the right map is in $F \cap W$ , so M3 provides $ϕ : Q Y \to Q^{'} Y$ with $p^{'} ϕ = p$ . Symmetrically construct $ψ : Q^{'} Y \to Q Y$ with $p ψ = p^{'}$ . Apply M3 again to lift the diagram showing $ψ ϕ ≃ id_{Q Y}$ using a cylinder object on $Q Y$ . The result is a homotopy equivalence between $Q Y$ and $Q^{'} Y$ in $C$ , hence an isomorphism in $Ho (C)$ .

For functoriality up to homotopy, given $f : Y \to Y^{'}$ choose cofibrant replacements $Q Y, Q Y^{'}$ and lift the composition $Q Y \to Y f Y^{'}$ along $Q Y^{'} \to Y^{'}$ via M3 to get $Q f : Q Y \to Q Y^{'}$ . Two different choices of lift differ by a left-homotopy, so $[Q f]$ is well-defined in $Ho (C)$ . Composition $Q (g f) = Q g \circ Q f$ holds up to homotopy by another lifting argument. $□$

Proposition (Ken Brown's lemma, formal statement). Let $C$ be a model category and $D$ a category with weak equivalences satisfying 2-out-of-3. A functor $F : C \to D$ that maps acyclic cofibrations between cofibrant objects to weak equivalences maps all weak equivalences between cofibrant objects to weak equivalences.

Proof. This is Exercise 7. We restate the argument cleanly. Given $f : A \to B$ a weak equivalence between cofibrant objects, factor $(f, id_{B}) : A ⊔ B \to B$ as $A ⊔ B j Z q B$ via M4 with $j \in C \cap W$ and $q \in F$ . The structure maps $ι_{A}, ι_{B} : A, B \to A ⊔ B$ are cofibrations (pushouts of $\emptyset \to A, B$ ), and $A ⊔ B$ is cofibrant. The composites $j ι_{A} : A \to Z$ and $j ι_{B} : B \to Z$ are acyclic cofibrations between cofibrant objects ( $A, B, Z$ all cofibrant). By hypothesis $F (j ι_{A})$ and $F (j ι_{B})$ are weak equivalences in $D$ . Then $q \circ j ι_{B} = id_{B}$ gives $F (q) F (j ι_{B}) = F (id_{B}) = id$ , so $F (q)$ is a weak equivalence by 2-out-of-3. And $f = q \circ j ι_{A}$ , so $F (f) = F (q) F (j ι_{A})$ is a weak equivalence as a composite of two. $□$

Proposition (M3 lifting determines the classes by mutual orthogonality). In a model category, the class $C$ of cofibrations consists exactly of the morphisms with the left lifting property against $F \cap W$ , and dually $F \cap W$ consists of the morphisms with the right lifting property against $C$ . Similarly $C \cap W$ and $F$ are mutually determined by lifting.

Proof. $C$ has the left lifting property against $F \cap W$ by M3. Conversely, suppose $i : A \to B$ has the left lifting property against every acyclic fibration. Factor $i = p j$ by M4 with $j \in C$ and $p \in F \cap W$ . The square with $i$ on the left and $p$ on the right has a diagonal lift by hypothesis. The lift expresses $i$ as a retract of $j$ in the arrow category $C^{\to}$ , and $C$ is retract-closed by M2, so $i \in C$ . The dual argument with $C \cap W$ versus $F$ uses the other M4 factorisation. $□$

Proposition (Quillen-Serre on $Top$ satisfies M1-M5). The category $Top$ with weak equivalences the weak homotopy equivalences, fibrations the Serre fibrations, and cofibrations the retracts of relative cell complexes is a model category.

Proof sketch. M1 (2-out-of-3 for weak homotopy equivalences) is immediate: $π_{n}$ is a functor, and 2-out-of-3 in $Grp$ propagates. M2: weak homotopy equivalence is preserved by retracts because $π_{n}$ preserves retracts. Serre fibration is closed under retracts because the homotopy-lifting property transfers along retract diagrams. Relative cell complexes are closed under retracts by definition of the class.

M3: every relative cell complex has the left lifting property against acyclic Serre fibrations. The proof goes inclusion-by-inclusion: each cell inclusion $S^{n - 1} ↪ D^{n}$ has the LLP against acyclic Serre fibrations because $D^{n}$ is contractible and $π_{*}$ of the fibre vanishes. Transfinite composition and retract closure extend this to all relative cell complexes.

M4: the first factorisation $f = p i$ with $i$ an acyclic cofibration and $p$ a fibration is realised by the cocylinder construction $M^{f} = X \times_{f} Y^{I}$ (the cocylinder of $f$ ). The second factorisation $f = q j$ with $j$ a cofibration and $q$ an acyclic fibration is the mapping cylinder $M_{f} = (X \times I) ⊔_{X} Y$ followed by the deformation retraction; equivalently, the small-object argument applied to the generating set ${S^{n - 1} ↪ D^{n}}$ .

M5: $Top$ has all small limits and colimits (well-known: limits via subspaces of products, colimits via quotients of disjoint unions). $□$

The full proof, with all transfinite-composition bookkeeping and the verification that the small-object argument terminates, occupies Hovey 1999 §2.4 (pp. 35-43) and Hirschhorn 2003 §7.2 (pp. 144-152). The technical content lies in showing that the cell complex of $Top$ is small in the categorical sense, which requires either compactly generated weak Hausdorff spaces or careful management of compact-open topologies.

Proposition (Kan-Quillen on $sSet$ satisfies M1-M5). $sSet$ with weak equivalences the maps whose geometric realization is a weak homotopy equivalence, fibrations the Kan fibrations, cofibrations the monomorphisms is a model category.

Proof sketch. Quillen 1967 §II.3 gives the original proof; Goerss-Jardine 2009 §II.5 a modern repackaging. The cofibrations are the monomorphisms — direct to check that this is retract-closed (M2). The Kan fibrations are the simplicial-set maps with the right lifting property against horn inclusions $Λ_{k}^{n} ↪ Δ^{n}$ for $0 \leq k \leq n$ and $n \geq 1$ . The weak equivalences are the maps inducing isomorphisms on $π_{n}$ , where $π_{n}$ on simplicial sets is defined via Kan complexes (Goerss-Jardine §I.6).

The M3 lifting between monomorphisms and acyclic Kan fibrations follows from a direct combinatorial argument: an acyclic Kan fibration $p : X \to Y$ has the right lifting property against all monomorphisms by a horn-filling / dimension-by-dimension induction. The M4 factorisation is the small-object argument applied to the generating monomorphisms ${\partial Δ^{n} ↪ Δ^{n}}_{n \geq 0}$ (cofibration generators) and ${Λ_{k}^{n} ↪ Δ^{n}}_{0 \leq k \leq n, n \geq 1}$ (acyclic cofibration generators). The smallness hypothesis is automatic: simplicial sets are small (every $Δ^{n}$ has only countably many simplices). M5: $sSet$ is bicomplete (functor category into $Set$ , which is bicomplete). $□$

The geometric-realization / singular-complex pair $∣ - ∣ ⊣ Sing$ is a Quillen adjunction relating these structures, and is a Quillen equivalence by Quillen's theorem II.3.1; the proof identifies $Sing (X)$ as the cofibrant-fibrant replacement of $X$ on the simplicial side, modulo the standard CW-approximation argument on the topological side.

Connections [Master]

Simplicial sets and geometric realization 03.12.25. The Kan-Quillen model structure on $sSet$ is the foundational example of a model category in the combinatorial direction, and the $∣ - ∣ ⊣ Sing$ adjunction is a Quillen equivalence between $sSet$ and $Top$ (Quillen 1967 Theorem II.3.1). Every classical homotopy-theoretic computation can be transported from one side to the other through this equivalence, and the model-category framework is what makes the transport functorial rather than merely component-wise.
Homotopy 03.12.01. The classical notion of homotopy of continuous maps is recovered in the Quillen-Serre model structure on $Top$ as the left/right homotopy relation when the source is cofibrant (a CW complex) and the target is fibrant (all spaces are fibrant in Quillen-Serre). The homotopy classes $[X, Y]$ on CW complexes are exactly the $π (X, Y)$ of the model-category theorem, and the homotopy-category construction $Ho (Top)$ is the classical pointed homotopy category restricted to CW complexes.
Semi-simplicial sets / $Δ$ -complexes 03.12.22. The semi-simplicial framework provides the face-only data underlying the full simplicial-set / model-category construction. There is a model structure on semi-simplicial sets (analogous to the Reedy model structure) whose homotopy theory matches the $Top$ side through the realization functor, but the model-categorical machinery is cleanest on the full $sSet$ where degeneracies allow the small-object argument to terminate cleanly.
CW complex 03.12.10. The cofibrant objects in the Quillen-Serre model structure on $Top$ are exactly the retracts of CW complexes. CW approximation — the construction of a CW complex weakly equivalent to a given space — is precisely cofibrant replacement in this model structure. The Whitehead theorem (every weak equivalence between CW complexes is a homotopy equivalence) is the special case of the homotopy-category theorem where source and target are both cofibrant-fibrant.
Singular homology 03.12.11. Singular homology is a homotopy-invariant functor $H_{*} : Top \to GrAb$ descending to $Ho (Top)$ . The descent is automatic in the model-category framework: the singular chain functor sends weak equivalences (weak homotopy equivalences) to quasi-isomorphisms, and the homology functor then sends quasi-isomorphisms to isomorphisms. The homotopy-invariance of singular homology is a corollary of the model-category construction, not a separate theorem.
Eilenberg-MacLane spaces 03.12.05. The fibrant replacement in the model structure on simplicial abelian groups produces an explicit simplicial model for Eilenberg-MacLane spaces $K (π, n)$ , recovered topologically by geometric realization. The Dold-Kan correspondence (a Quillen equivalence between simplicial abelian groups and non-negatively graded chain complexes) identifies $K (π, n)$ with the chain complex concentrated in degree $n$ , providing a clean conceptual reason for the cohomological characterisation $H^{n} (X; π) = [X, K (π, n)]$ .
Sullivan minimal models 03.12.06. Rational homotopy theory is presented as a model category via the Bousfield-Gugenheim model structure on commutative differential graded $Q$ -algebras (cdga's), in which weak equivalences are quasi-isomorphisms and cofibrant objects are the minimal models. The Sullivan minimal model of a space is its cofibrant replacement in this model category, and rational homotopy equivalence of spaces is the equivalent in $Ho (cdga)$ of cofibrant replacement. The framework recasts rational-homotopy computations as Quillen-equivalent algebraic computations.
Quillen functor and equivalence 03.12.32. Once the model-category framework of the present unit is in place, the next structural question is how morphisms between model categories behave with respect to the derived calculus. The sibling unit 03.12.32 develops the calculus of left and right Quillen functors, Ken Brown's lemma, derived adjunctions, and Quillen equivalences as the foundational machinery converting model-category structures into well-behaved adjoint pairs on homotopy categories. The realisation-singular adjunction $∣ - ∣ ⊣ Sing$ named here as a Quillen equivalence is developed in 03.12.32 as the canonical worked example, and the Dold-Kan and Bousfield-Gugenheim equivalences alluded to above are treated systematically in the same unit. The two units are a foundation pair: the present unit axiomatises the objects, the sibling unit axiomatises the morphisms between them.
Kan-Quillen model structure on sSet 03.12.33. The Kan-Quillen structure is the canonical worked example of the abstract framework developed here: $sSet$ with cofibrations the monomorphisms, fibrations the Kan fibrations (right lifting against horn inclusions $Λ_{k}^{n} ↪ Δ^{n}$ ), and weak equivalences the maps whose geometric realisation is a weak homotopy equivalence. The five axioms M1-M5 are verified explicitly in 03.12.33 via the small-object argument applied to the generating cofibrations $I = {\partial Δ^{n} ↪ Δ^{n}}$ and generating acyclic cofibrations $J = {Λ_{k}^{n} ↪ Δ^{n}}$ , exhibiting the present unit's abstract recipe in its prototypical combinatorial form. The foundational reason the framework was built is to axiomatise exactly this kind of construction.
Simplicial model category and function complex 03.12.35. The simplicial-enrichment refinement adds the function complex $Map (X, Y)$ and the SM7 pushout-product axiom to the bare framework of the present unit. Where the present unit axiomatises $Hom_{Ho (C)} (X, Y)$ as a set, 03.12.35 upgrades it to a Kan complex carrying higher-coherence data; the two units are a foundation pair, with 03.12.35 supplying the $(\infty, 1)$ -categorical content that the present unit's $π_{0}$ -level framework leaves implicit.
Homotopy colimit via Bousfield-Kan 03.12.37. The homotopy colimit $hocolim_{I} X$ is the left-derived left adjoint of the diagram-constant functor in the projective model structure on $sSet^{I}$ , the prototypical application of the derived-functor calculus axiomatised in the present unit. The Bousfield-Kan bar construction $B (*, I, X)$ supplies the explicit point-set formula; the present unit's framework guarantees that the formula computes the correct derived functor.

Historical & philosophical context [Master]

Daniel Quillen introduced the model-category framework in his 1967 Lecture Notes in Mathematics monograph Homotopical Algebra (Springer LNM 43) ^{[Quillen 1967]}, which built on his Harvard thesis. The motivation was to unify the homotopy theories of topological spaces, simplicial sets, simplicial groups, and chain complexes under a single set of axioms. Quillen identified the three classes of morphisms — weak equivalences, fibrations, cofibrations — and the five axioms M1-M5, and showed that the resulting category-theoretic structure was sufficient to construct a well-behaved homotopy category $Ho (C)$ together with derived functors via the Quillen adjunction calculus. The 1967 monograph also proved the Quillen equivalence $∣ - ∣ ⊣ Sing$ between simplicial sets and topological spaces, the foundational compatibility result of the framework.

The framework lay relatively dormant for two decades before being rediscovered in the 1990s by the new generation of homotopy theorists working on equivariant, motivic, and stable settings. Mark Hovey's 1999 monograph Model Categories (AMS Mathematical Surveys 63) ^{[Hovey 1999]} gave the first modern treatment with the functorial-factorisation axiom that simplifies many constructions; Philip Hirschhorn's 2003 Model Categories and Their Localizations (AMS Mathematical Surveys 99) ^{[Hirschhorn 2003]} developed the small-object argument and Bousfield localisation in full generality; Paul Goerss and Rick Jardine's 2009 Simplicial Homotopy Theory ^{[Goerss-Jardine 2009]} gave the canonical treatment of the Kan-Quillen and Joyal model structures on $sSet$ . The pedagogical introduction by Dwyer and Spalinski in the 1995 Handbook of Algebraic Topology chapter ^{[Dwyer-Spalinski 1995]} remains the standard entry point.

The conceptual extension of model categories into $(\infty, 1)$ -category theory was carried out by André Joyal in his 2002 paper on quasi-categories and developed in book form by Jacob Lurie in Higher Topos Theory (2009) ^{[Lurie 2009]}. Lurie identified model categories as one of several presentations of an abstract homotopy theory, the others including simplicial categories, complete Segal spaces, and quasi-categories. The Quillen equivalence between these presentations is itself a theorem in the theory of $(\infty, 1)$ -categories. The original Quillen framework remains the most accessible and computational entry point; the $(\infty, 1)$ -categorical generalisation is the natural conceptual home.

The applications of model-category theory in the decades since Quillen's monograph extend across motivic homotopy theory (Morel-Voevodsky 1999, where the $A^{1}$ -model structure on simplicial presheaves on smooth schemes gives motivic spaces their homotopy theory), derived algebraic geometry (Toën-Vezzosi 2008, Lurie 2009-, where simplicial commutative rings carry a model structure presenting derived rings), equivariant stable homotopy (Mandell-May 2002, Hill-Hopkins-Ravenel 2009 for the Kervaire invariant), and chromatic homotopy theory (where Bousfield localisation at Morava $K$ -theory is the foundational construction). The framework's longevity comes from its combination of axiomatic minimalism — five axioms, three classes — and computational power: once the axioms are verified for a category, the entire derived-functor calculus and homotopy-category construction is automatic.

Bibliography [Master]

@book{Quillen1967,
  author    = {Quillen, Daniel G.},
  title     = {Homotopical Algebra},
  series    = {Lecture Notes in Mathematics},
  volume    = {43},
  publisher = {Springer-Verlag},
  year      = {1967}
}

@book{Hovey1999,
  author    = {Hovey, Mark},
  title     = {Model Categories},
  series    = {Mathematical Surveys and Monographs},
  volume    = {63},
  publisher = {American Mathematical Society},
  year      = {1999}
}

@book{Hirschhorn2003,
  author    = {Hirschhorn, Philip S.},
  title     = {Model Categories and Their Localizations},
  series    = {Mathematical Surveys and Monographs},
  volume    = {99},
  publisher = {American Mathematical Society},
  year      = {2003}
}

@incollection{DwyerSpalinski1995,
  author    = {Dwyer, William G. and Spalinski, Jan},
  title     = {Homotopy theories and model categories},
  booktitle = {Handbook of Algebraic Topology},
  editor    = {James, I. M.},
  publisher = {Elsevier},
  year      = {1995},
  pages     = {73--126}
}

@book{GoerssJardine2009,
  author    = {Goerss, Paul G. and Jardine, John F.},
  title     = {Simplicial Homotopy Theory},
  series    = {Modern Birkh{\"a}user Classics},
  publisher = {Birkh{\"a}user},
  year      = {2009},
  note      = {Reprint of 1999 edition}
}

@article{Strom1972,
  author    = {Str{\o}m, Arne},
  title     = {The homotopy category is a homotopy category},
  journal   = {Archiv der Mathematik},
  volume    = {23},
  year      = {1972},
  pages     = {435--441}
}

@article{BousfieldGugenheim1976,
  author    = {Bousfield, Aldridge K. and Gugenheim, Victor K. A. M.},
  title     = {On {PL} de {R}ham theory and rational homotopy type},
  journal   = {Memoirs of the American Mathematical Society},
  volume    = {179},
  year      = {1976}
}

@article{HoveyShipleySmith2000,
  author    = {Hovey, Mark and Shipley, Brooke and Smith, Jeff},
  title     = {Symmetric spectra},
  journal   = {Journal of the American Mathematical Society},
  volume    = {13},
  year      = {2000},
  pages     = {149--208}
}

@book{Lurie2009,
  author    = {Lurie, Jacob},
  title     = {Higher Topos Theory},
  series    = {Annals of Mathematics Studies},
  volume    = {170},
  publisher = {Princeton University Press},
  year      = {2009}
}

@article{Joyal2002,
  author    = {Joyal, Andr{\'e}},
  title     = {Quasi-categories and {K}an complexes},
  journal   = {Journal of Pure and Applied Algebra},
  volume    = {175},
  year      = {2002},
  pages     = {207--222}
}

@article{MorelVoevodsky1999,
  author    = {Morel, Fabien and Voevodsky, Vladimir},
  title     = {$\mathbb{A}^1$-homotopy theory of schemes},
  journal   = {Publications Math{\'e}matiques de l'IH{\'E}S},
  volume    = {90},
  year      = {1999},
  pages     = {45--143}
}

@book{Riehl2014,
  author    = {Riehl, Emily},
  title     = {Categorical Homotopy Theory},
  series    = {New Mathematical Monographs},
  volume    = {24},
  publisher = {Cambridge University Press},
  year      = {2014}
}

Prerequisites

02.01.01
02.01.02
03.12.01
03.12.10
03.12.22
03.12.25

Tier anchors

beginner: Dwyer-Spalinski 1995 *Homotopy theories and model categories* §1-§3, the canonical informal introduction
intermediate: Quillen 1967 *Homotopical Algebra* (LNM 43) §I.1-§I.5; Dwyer-Spalinski 1995 §4-§8; Hovey 1999 *Model Categories* §1.1-§1.3
master: Quillen 1967 *Homotopical Algebra* (LNM 43) §I-§II (originator); Hovey 1999 *Model Categories* (AMS Mathematical Surveys 63) §1-§3; Hirschhorn 2003 *Model Categories and Their Localizations* (AMS Mathematical Surveys 99) §7-§9; Goerss-Jardine 2009 *Simplicial Homotopy Theory* §II; Dwyer-Spalinski 1995 *Homotopy theories and model categories* (Handbook of Algebraic Topology Ch. 2)

References

TODO_REF
Quillen — Homotopical Algebra · Springer Lecture Notes in Mathematics 43 (1967); originator paper, §I.1 axioms M1-M5, §I.2 cylinder and path objects, §I.3 left/right homotopy and the homotopy category construction, §I.5 Ken Brown's lemma, §II.3 model structure on simplicial sets, §II.4 Quillen equivalence with topological spaces
TODO_REF
Hovey — Model Categories · AMS Mathematical Surveys and Monographs 63 (1999); modern monograph, §1.1 definition (cofibrantly generated form), §1.2 functorial factorisation, §1.3 the homotopy category, §2.1 chain complexes, §2.4 topological spaces, §3 monoidal model categories
TODO_REF
Dwyer-Spalinski — Homotopy theories and model categories · Handbook of Algebraic Topology (Elsevier 1995) Ch. 2 pp. 73-126; the canonical pedagogical introduction; §3 axioms, §4 examples, §5 homotopy and cofibrant/fibrant replacement, §6 homotopy category construction, §9 Quillen functors and equivalences
TODO_REF
Hirschhorn — Model Categories and Their Localizations · AMS Mathematical Surveys and Monographs 99 (2003); §7 model-category axioms with small-object factorisations, §8 cofibrantly generated model categories, §9 localisations of model categories, §11 left/right Bousfield localisation
TODO_REF
Goerss-Jardine — Simplicial Homotopy Theory · Birkhäuser Progress in Mathematics 174 (1999, repr. 2009); §II.1-§II.4 Kan-Quillen model structure on simplicial sets, §II.5 fibrant replacement, §II.6 the homotopy category $\mathrm{Ho}(\mathbf{sSet})$
TODO_REF
May — A Concise Course in Algebraic Topology · University of Chicago Press 1999; Ch. 17 model categories and homotopy colimits, treatment of Quillen-Serre model structure on $\mathbf{Top}$
TODO_REF
Hatcher — Algebraic Topology · Cambridge 2002; §0 mapping cylinder and cocylinder constructions, §4.1 fibrations and Serre fibrations, §4.5 Whitehead theorem and CW approximation as the underpinning of $\mathbf{Top}$ model structure
TODO_REF
Riehl — Categorical Homotopy Theory · Cambridge 2014; §11-§14 model categories, small-object argument, derived functors via Quillen adjunctions

Reviewer

TBD

Estimated time

beginner: 20m
intermediate: 45m
master: 90m