We now use the preceding colimit calculations to prove descent results for cocartesian fibrations and then Quillen’s Theorem B. To formulate the universal cocartesian fibration as an object of a category, we temporarily work in a larger universe.

13.4.1 The universal Grothendieck construction

Straightening identifies a functor \(X\to\Cat\) with its cocartesian unstraightening, and naturality by pullback identifies morphisms over \(\Cat\) with pullback squares of cocartesian fibrations. The following construction packages this correspondence as an equivalence of categories.

Consider a bigger universe \(\Cat'\) so that \(\Cat\) is \(\Cat'\)-small and there is a fully faithful functor \(\Cat \hookrightarrow \Cat'\). Let \(\An'\) denote the corresponding universe of animae. Commutative diagram; see the PDF for its LaTeX source The objects of the category \(\Cat_{/\Cat}\) are pairs \((X,F)\) where \(X\) is a small category and \(F\colon X \to \Cat\) is a functor. The category \(\Fun([1],\Cat')\) has a non-full subcategory \(\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat')\) whose objects are cocartesian fibrations in \(\Cat'\) and whose morphisms are the pullback squares in \(\Cat'\), Commutative diagram; see the PDF for its LaTeX source with \(f\) and \(f'\) cocartesian fibrations. Let \(\univ=(\Cat_\bullet,\pi_{\univ})\), where \(\pi_{\univ}\colon\Cat_\bullet\to\Cat\) is the universal cocartesian fibration with small fibers, seen as an object of \(\Fun([1],\Cat')\).

There is a canonical full embedding of the form \[ \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)\hookrightarrow\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat') \] and we can form the relative slice category \[ \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)_{/\univ}= \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)\times_{\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat')}\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat')_{/\univ}\, . \] We observe that the forgetful functor \[ \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)_{/\univ}\to \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat) \] is an equivalence. Indeed, it suffices to check that it induces an equivalence of animae of the form \[ \Map([1],\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)_{/\univ})\to \Map([1],\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)) \] which amounts to see that any pullback square of small categories Commutative diagram; see the PDF for its LaTeX source extends uniquely into pullback squares of the form Commutative diagram; see the PDF for its LaTeX source Evaluating at \(1\) induces a canonical functor \[ \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)_{/\univ}\to \Cat_{/\Cat}\, . \]

Theorem 13.4.1 (Universal Grothendieck construction). There is a canonical equivalence \[ \Cat_{/\Cat} \iso \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat) \] assigning to each pullback square of the form Commutative diagram; see the PDF for its LaTeX source the functor \(p_F\), seen as a term of \(\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)\).

Proof

We have a functor \[ \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)\iso \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)_{/\univ} \to \Cat_{/\Cat}\, . \] We observe that the induced map \[ \Map([1],\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat))\to \Map([1],\Cat_{/\Cat}) \] is an equivalence: this means that pullback squares of the form Commutative diagram; see the PDF for its LaTeX source with \(Y\) and \(Y'\) small categories are uniquely determined by diagrams of the form Commutative diagram; see the PDF for its LaTeX source The inverse is given by the rule \((X,F)\mapsto p_F\).

13.4.2 Descent for cocartesian fibrations

The preceding theorem provides a convenient way of computing colimits in \(\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)\):

Theorem 13.4.2. The category \[ \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat) \] admits small colimits, and its inclusion into \(\Fun([1],\Cat)\) preserves them.

Proof

Since the functor \(\Cat\hookrightarrow\Cat'\) preserves the formation of small colimits and since the functor \(\Cat'_{/\Cat}\to\Cat'\) commutes with large colimits (hence with small ones as well), the pullback \(\Cat_{/\Cat}\) therefore has small colimits and the forgetful functor \[ \Cat_{/\Cat}\to\Cat \] commutes with colimits. The fact that \(\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)\) admits small colimits therefore follows from Theorem 13.4.1. To see that they are preserved by the inclusion, notice that the functors \(\ev_0,\ev_1\colon \Fun([1],\Cat) \to \Cat\) preserve colimits and are jointly conservative. So it suffices to show that the functors \(\ev_0,\ev_1\colon \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat) \to \Cat\) preserve small colimits. This follows from the observation that the composite \[ \Cat_{/\Cat} \iso \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat) \xrightarrow{\ev_1} \Cat \] agrees with the forgetful functor \((X, F\colon X \to \Cat) \mapsto X\), which preserves colimits, and that the composite \[ \Cat_{/\Cat} \iso \Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat) \xrightarrow{\ev_0} \Cat \] agrees with the functor \((X,F\colon X \to \Cat) \mapsto \int F = X \times_{\Cat} \Cat_{\bullet}\), which also commutes with colimits (since pulling back along an exponentiable functor commutes with colimits).

Corollary 13.4.3. Let \(A\) be a small category and \(X,Y\colon A\to \Cat\) be two functors and \(u\colon X\to Y\) a natural transformation. We assume that the following two conditions are satisfied.

  • (i) For each object \(a : A^\simeq\), the induced functor \(u_a\colon X_a\to Y_a\) is a cocartesian fibration (a cartesian fibration, a right fibration, or a left fibration).

  • (ii) For each morphism \(\alpha\colon a_0\to a_1\) in \(A\), the induced commutative square Commutative diagram; see the PDF for its LaTeX source is a pullback square in \(\Cat\).

Then the induced functor \(\colim_{a : A} X_a\to \colim_{a : A} Y_a\) is a cocartesian fibration (a cartesian fibration, a right fibration, a left fibration, respectively) and the canonical commutative squares Commutative diagram; see the PDF for its LaTeX source are pullback squares for any object \(a : A^\simeq\).

Proof

Let us discuss the case of cocartesian fibrations. The natural transformation \(u\) can be interpreted as an \(A\)-indexed diagram with values in \(\Fun^{\mathrm{coCart}}_{\Cart}([1],\Cat)\). Taking the colimit there and the preceding theorem yields the conclusions of the corollary.

Assuming that all functors \(u_a\) are left fibrations, since there is a full embedding \[ \Fun(A,\An)\hookrightarrow\Fun(A,\Cat) \] whose essential image consists of those functors taking values in \(\An\) object-wise, we can see the natural transformation from \(Y\) to the constant \(A\)-indexed functor with value \(\Cat\) classifying the left fibrations \(u_a\) as natural transformation from \(Y\) to the constant functor with value \(\An\). In other words, we get a commutative diagram of the form Commutative diagram; see the PDF for its LaTeX source in which all squares are pullback squares, proving that the functor induced on colimits is a left fibration.

The remaining cases of cartesian fibrations and of right fibrations are proved similarly by duality, or by applying the preceding cases to the diagram induced by composing with the equivalence \((-)\catop\), and then applying \((-)\catop\) again, using that the latter preserves both pullbacks and colimits.

13.4.3 Descent for animae

Lemma 13.4.4. Let \(A\) be a small category and \(X,Y\colon A\to \An\) be two functors, with \(u\colon X\to Y\) a natural transformation. We assume that, for each morphism \(\alpha\colon a_0\to a_1\) in \(A\), the induced commutative square Commutative diagram; see the PDF for its LaTeX source is a pullback square in \(\An\). Then the canonical commutative squares Commutative diagram; see the PDF for its LaTeX source are pullback squares in \(\An\) for any object \(a : A^\simeq\).

Proof

Since the embedding \(\An\hookrightarrow\Cat\) commutes with colimits, so does the induced embedding of categories of arrows \[ \Fun([1],\An)\hookrightarrow\Fun([1],\Cat)\, . \] We apply Corollary 13.4.3, observing that any functor between animae is both a left and a right fibration.

13.4.4 Quillen’s Theorem B

Theorem 13.4.5 (Quillen’s Theorem B). Let \(u\colon A \to B\) be a functor between small categories. The following conditions are equivalent.

  • (i) For any morphism \(\beta\colon b \to b'\) in \(B\), the induced functor \[ \beta_!\colon A_{/b} \to A_{/b'} \] induces an equivalence \(\Pi_{\infty}(A_{/b}) \iso \Pi_{\infty}(A_{/b'})\).

  • (ii) For any object \(b\) of \(B\), the square Commutative diagram; see the PDF for its LaTeX source is a pullback square in \(\An\).

  • (iii) Any pullback square of the form Commutative diagram; see the PDF for its LaTeX source in which \(g\) is a right fibration induces a pullback square Commutative diagram; see the PDF for its LaTeX source in \(\An\).

  • (iv) Any pullback square of the form Commutative diagram; see the PDF for its LaTeX source in which \(g\) is a smooth functor induces a pullback square Commutative diagram; see the PDF for its LaTeX source in \(\An\).

There is a dual set of equivalent conditions obtained by simultaneously replacing relative slices \(A_{/b}\) and \(B_{/b}\) by relative coslices \(A_{b/}\) and \(B_{b/}\), replacing \(\beta_!\) by \(\beta^*\), replacing right fibrations by left fibrations, and replacing smooth functors by proper functors.

Proof

We prove the displayed formulation in terms of slices, right fibrations, and smooth functors. The implications are organized as \[ (iii)\Rightarrow(ii)\Rightarrow(i)\Rightarrow(ii)\Rightarrow(iii), \qquad (iii)\Longleftrightarrow(iv). \] The dual formulation follows by applying this result to \(u\catop\) and using Corollary 12.4.6.

Condition (ii) follows from condition (iii) since the canonical projection \(B_{/b}\to B\) is a right fibration. Similarly, assuming condition (ii), any morphism \(\beta\colon b\to b'\) induces a pullback square of the form Commutative diagram; see the PDF for its LaTeX source in which the lower horizontal map is an endomorphism of the terminal object, so that the upper horizontal map must be invertible. It remains to prove that condition (i) implies condition (ii).

Let \(e\) be the terminal object of \(\An\). We will also write \(e\) for the constant presheaf with value \(e\). We have \(A = A_{/e}\) and \(A_{/b} = A \times_B B_{/b} = A_{/u^*(\Hom_B(-,b))}\), where Commutative diagram; see the PDF for its LaTeX source Since \(\Pi_{\infty}\) commutes with colimits (being a left adjoint), Lemma 13.3.4 implies that \[ X \mapsto \Pi_{\infty}(A_{/X}) \] preserves colimits. But the functor \(u^*\) also has a left adjoint and thus commutes with colimits. This proves that the assignment \(Y \mapsto \Pi_{\infty}(A_{/u^*Y})\) commutes with colimits. Since we have \[ \colim_B h_B = e\qquad\text{and}\qquad u^*(e)=e\, , \] by Theorem 12.6.3, we see that \(\colim_{b : B} \Pi_{\infty}(A_{/b}) \simeq \Pi_{\infty}(A_{/e}) \simeq \Pi_{\infty}(A)\). Since for every morphism \(\beta\colon b \to b'\) we have a pullback square Commutative diagram; see the PDF for its LaTeX source it follows from Lemma 13.4.4 that the commutative square Commutative diagram; see the PDF for its LaTeX source induced by passing to colimits is a pullback square, as desired.

We will finally prove that condition (ii) implies condition (iii). For any object \(b'\) of \(B'\), if we put \(b=g(b')\), the induced functor \[ B'_{/b'}\to B_{/b} \] is an equivalence over \(B\) (it is a terminal functor between right fibrations). The horizontal arrows in the commutative square Commutative diagram; see the PDF for its LaTeX source are equivalences, and we thus obtain from condition (ii) a pullback square of the form Commutative diagram; see the PDF for its LaTeX source for each object \(b'\) in \(B'\). The transition maps in the two \(B'\)-indexed diagrams are equivalences, and hence their naturality squares are pullbacks. Applying Lemma 13.4.4 and using \[ \colim_{b':B'}\Pi_{\infty}(A'_{/b'})\iso\Pi_{\infty}(A') \quad\text{and}\quad \colim_{b':B'}\Pi_{\infty}(B'_{/b'})\iso\Pi_{\infty}(B') \] shows that condition (iii) follows from condition (ii).

Since every right fibration is smooth, condition (iv) implies condition (iii). For the converse, given a pullback square of the form Commutative diagram; see the PDF for its LaTeX source in which \(g\) is a smooth functor, we choose a factorization of \(g\) into \(g=q\circ j\) with a terminal functor \(j\colon B'\to B''\) and a right fibration \(q\colon B''\to B\). We define \(A''=B''\times_BA\). The functor \(j\) induces a functor \(i\colon A'\to A''\) that is terminal (we see this by applying Corollary 12.4.7 in the case where \(\varphi\) is the identity). If \(p\colon A''\to A\) denotes the base change of \(q\), and \(u''\colon A''\to B''\) the canonical projection, we obtain a commutative diagram of the form Commutative diagram; see the PDF for its LaTeX source in which the two horizontal maps of the square at the left hand side are equivalences. Therefore, since the square at the right hand side is a pullback square, so is the composite.