Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Key lemma and universal localization

Vladimir Hinich Department of Mathematics, University of Haifa, Mount Carmel, Haifa 3498838, Israel hinich@math.haifa.ac.il
Abstract.

We study localizations of \infty-categories which remain localizations after any base change.

1. Key lemma

Recall the following Key lemma presented in  [H.L], 1.3.6. The Key lemma served a convenient tool in proving the equivalence of various descriptions of the \infty-category underlying a model category, see [H.L], 1.3.4, 1.3.5, 1.3.7.

1.1 Lemma.

Let f:𝒞𝒟:𝑓𝒞𝒟f:\mathcal{C}\to\mathcal{D}italic_f : caligraphic_C → caligraphic_D be a functor between conventional categories satisfying the following property (P).

  • (P)

    For any s:[n]𝒟:𝑠delimited-[]𝑛𝒟s:[n]\to\mathcal{D}italic_s : [ italic_n ] → caligraphic_D the category of sections 𝒞s=Fun𝒟([n],𝒞)subscript𝒞𝑠subscriptFun𝒟delimited-[]𝑛𝒞\mathcal{C}_{s}=\operatorname{Fun}_{\mathcal{D}}([n],\mathcal{C})caligraphic_C start_POSTSUBSCRIPT italic_s end_POSTSUBSCRIPT = roman_Fun start_POSTSUBSCRIPT caligraphic_D end_POSTSUBSCRIPT ( [ italic_n ] , caligraphic_C ) is weakly contractible.

Then f𝑓fitalic_f induces an equivalence

(𝒞,f1(𝒟𝑒𝑞))𝒟,𝒞superscript𝑓1superscript𝒟𝑒𝑞𝒟\mathcal{L}(\mathcal{C},f^{-1}(\mathcal{D}^{\mathit{eq}}))\to\mathcal{D},caligraphic_L ( caligraphic_C , italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( caligraphic_D start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT ) ) → caligraphic_D ,

where \mathcal{L}caligraphic_L denotes the (\infty-categorical) localization.

In this note we generalize this result in several directions.

  • We remove the assumption of 𝒞𝒞\mathcal{C}caligraphic_C and 𝒟𝒟\mathcal{D}caligraphic_D, see 1.5.

  • We weaken the assumptions so that (P) will be required to hold only for simplices of dimension 1absent1\leq 1≤ 1, see 1.6.

  • We provide a version of a converse statement, see 2.2.

1.2. Localization

In this note we will use the word “category” to denote \infty-categories; the categories in the traditional sense are called the conventional categories. We denote by 𝒮𝒮\mathcal{S}caligraphic_S the category of spaces and by 𝙲𝚊𝚝𝙲𝚊𝚝\mathtt{Cat}typewriter_Cat the category of categories. Recall that the full embedding 𝒮𝙲𝚊𝚝𝒮𝙲𝚊𝚝\mathcal{S}\to\mathtt{Cat}caligraphic_S → typewriter_Cat admits a left adjoint

:𝙲𝚊𝚝𝒮:𝙲𝚊𝚝𝒮\mathcal{L}:\mathtt{Cat}\to\mathcal{S}caligraphic_L : typewriter_Cat → caligraphic_S

called the full localization. A functor f:𝒞𝒟:𝑓𝒞𝒟f:\mathcal{C}\to\mathcal{D}italic_f : caligraphic_C → caligraphic_D is called a weak equivalence if (f)𝑓\mathcal{L}(f)caligraphic_L ( italic_f ) is an equivalence. A category 𝒞𝒞\mathcal{C}caligraphic_C is weakly contractible if (𝒞)𝒞\mathcal{L}(\mathcal{C})caligraphic_L ( caligraphic_C ) is contractible.

More generally, one defines the category of marked categories 𝙲𝚊𝚝+superscript𝙲𝚊𝚝\mathtt{Cat}^{+}typewriter_Cat start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT consisting of pairs (𝒞,W)𝒞𝑊(\mathcal{C},W)( caligraphic_C , italic_W ) where W𝒞𝑒𝑞superscript𝒞𝑒𝑞𝑊W\supset\mathcal{C}^{\mathit{eq}}italic_W ⊃ caligraphic_C start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT is a subcategory of 𝒞𝒞\mathcal{C}caligraphic_C (its arrows are called marked arrows). Then localization

:𝙲𝚊𝚝+𝙲𝚊𝚝:superscript𝙲𝚊𝚝𝙲𝚊𝚝\mathcal{L}:\mathtt{Cat}^{+}\to\mathtt{Cat}caligraphic_L : typewriter_Cat start_POSTSUPERSCRIPT + end_POSTSUPERSCRIPT → typewriter_Cat

is the functor left adjoint to the obvious functor 𝒞(𝒞,𝒞𝑒𝑞)maps-to𝒞𝒞superscript𝒞𝑒𝑞\mathcal{C}\mapsto(\mathcal{C},\mathcal{C}^{\mathit{eq}})caligraphic_C ↦ ( caligraphic_C , caligraphic_C start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT ).

We start with a very special case of the result.

1.3 Lemma.

Let f:𝒞𝒟:𝑓𝒞𝒟f:\mathcal{C}\to\mathcal{D}italic_f : caligraphic_C → caligraphic_D be a cocartesian fibration. The fibers of f𝑓fitalic_f are weakly contractible if and only if f𝑓fitalic_f induces an equivalence

(𝒞,f1(𝒟𝑒𝑞))𝒟.𝒞superscript𝑓1superscript𝒟𝑒𝑞𝒟\mathcal{L}(\mathcal{C},f^{-1}(\mathcal{D}^{\mathit{eq}}))\to\mathcal{D}.caligraphic_L ( caligraphic_C , italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( caligraphic_D start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT ) ) → caligraphic_D .
Proof.

Let us say that an arrow in 𝒞𝒞\mathcal{C}caligraphic_C is marked if f𝑓fitalic_f carries it to an equivalence in 𝒟𝒟\mathcal{D}caligraphic_D. With this definition f𝑓fitalic_f becomes a marked cocartesian fibration in the sense of [H.L], 2.1.1, so that Proposition 2.1.4 of [H.L] can be applied. This yields a left fibration

(f):(C,f1(𝒟𝑒𝑞))𝒟.:𝑓𝐶superscript𝑓1superscript𝒟𝑒𝑞𝒟\mathcal{L}(f):\mathcal{L}(C,f^{-1}(\mathcal{D}^{\mathit{eq}}))\to\mathcal{D}.caligraphic_L ( italic_f ) : caligraphic_L ( italic_C , italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( caligraphic_D start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT ) ) → caligraphic_D .

It is an equivalence if and only if its fibers are contractible. ∎

1.4. Rezk nerve

Let (𝒞,W)𝒞𝑊(\mathcal{C},W)( caligraphic_C , italic_W ) be a marked category. The Rezk nerve NR(𝒞,W)superscript𝑁𝑅𝒞𝑊N^{R}(\mathcal{C},W)italic_N start_POSTSUPERSCRIPT italic_R end_POSTSUPERSCRIPT ( caligraphic_C , italic_W ), see [MG], 3.1, is the simplicial space defined by the formula

NR(𝒞,W)n=(Fun([n],𝒞)W)N^{R}(\mathcal{C},W)_{n}=\mathcal{L}(\operatorname{Fun}([n],\mathcal{C})^{W})italic_N start_POSTSUPERSCRIPT italic_R end_POSTSUPERSCRIPT ( caligraphic_C , italic_W ) start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT = caligraphic_L ( roman_Fun ( [ italic_n ] , caligraphic_C ) start_POSTSUPERSCRIPT italic_W end_POSTSUPERSCRIPT )

where Fun([n],𝒞)W\operatorname{Fun}([n],\mathcal{C})^{W}roman_Fun ( [ italic_n ] , caligraphic_C ) start_POSTSUPERSCRIPT italic_W end_POSTSUPERSCRIPT is the subcategory of Fun([n],𝒞)Fundelimited-[]𝑛𝒞\operatorname{Fun}([n],\mathcal{C})roman_Fun ( [ italic_n ] , caligraphic_C ) having the same objects and spanned by the arrows that are pointwise in W𝑊Witalic_W. By Theorem 3.8 of [MG] the complete Segal replacement of NR(𝒞,W)superscript𝑁𝑅𝒞𝑊N^{R}(\mathcal{C},W)italic_N start_POSTSUPERSCRIPT italic_R end_POSTSUPERSCRIPT ( caligraphic_C , italic_W ) represents the localization (𝒞,W)𝙲𝚊𝚝𝒞𝑊𝙲𝚊𝚝\mathcal{L}(\mathcal{C},W)\in\mathtt{Cat}caligraphic_L ( caligraphic_C , italic_W ) ∈ typewriter_Cat.

Let now f:𝒞𝒟:𝑓𝒞𝒟f:\mathcal{C}\to\mathcal{D}italic_f : caligraphic_C → caligraphic_D be a functor. We denote by W=f1(𝒟𝑒𝑞)𝑊superscript𝑓1superscript𝒟𝑒𝑞W=f^{-1}(\mathcal{D}^{\mathit{eq}})italic_W = italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( caligraphic_D start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT ) the subcategory of 𝒞𝒞\mathcal{C}caligraphic_C spanned by the arrows sent by f𝑓fitalic_f to equivalences.

The category Fun([n],𝒞)W\operatorname{Fun}([n],\mathcal{C})^{W}roman_Fun ( [ italic_n ] , caligraphic_C ) start_POSTSUPERSCRIPT italic_W end_POSTSUPERSCRIPT is the base change of fn:Fun([n],𝒞)Fun([n],𝒟):subscript𝑓𝑛Fundelimited-[]𝑛𝒞Fundelimited-[]𝑛𝒟f_{n}:\operatorname{Fun}([n],\mathcal{C})\to\operatorname{Fun}([n],\mathcal{D})italic_f start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT : roman_Fun ( [ italic_n ] , caligraphic_C ) → roman_Fun ( [ italic_n ] , caligraphic_D ) with respect to the embedding Fun([n],𝒟)𝑒𝑞Fun([n],𝒟)\operatorname{Fun}([n],\mathcal{D})^{\mathit{eq}}\to\operatorname{Fun}([n],% \mathcal{D})roman_Fun ( [ italic_n ] , caligraphic_D ) start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT → roman_Fun ( [ italic_n ] , caligraphic_D ).

Applying Lemma 1.3 to the functor Fun([n],𝒞)WFun([n],𝒟)𝑒𝑞\operatorname{Fun}([n],\mathcal{C})^{W}\to\operatorname{Fun}([n],\mathcal{D})^% {\mathit{eq}}roman_Fun ( [ italic_n ] , caligraphic_C ) start_POSTSUPERSCRIPT italic_W end_POSTSUPERSCRIPT → roman_Fun ( [ italic_n ] , caligraphic_D ) start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT, we deduce the following more general version of Key lemma.

1.5 Lemma.

Let f:𝒞𝒟:𝑓𝒞𝒟f:\mathcal{C}\to\mathcal{D}italic_f : caligraphic_C → caligraphic_D be a functor satisfying the above mentioned property (P) for any s:[n]𝒟:𝑠delimited-[]𝑛𝒟s:[n]\to\mathcal{D}italic_s : [ italic_n ] → caligraphic_D. Then f𝑓fitalic_f induces an equivalence

(𝒞,f1(𝒟𝑒𝑞))𝒟.𝒞superscript𝑓1superscript𝒟𝑒𝑞𝒟\mathcal{L}(\mathcal{C},f^{-1}(\mathcal{D}^{\mathit{eq}}))\to\mathcal{D}.caligraphic_L ( caligraphic_C , italic_f start_POSTSUPERSCRIPT - 1 end_POSTSUPERSCRIPT ( caligraphic_D start_POSTSUPERSCRIPT italic_eq end_POSTSUPERSCRIPT ) ) → caligraphic_D .

Our next step is to show that the property (P) is equivalent to a weaker property

  • (P1)subscript𝑃1(P_{1})( italic_P start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT )

    For all s:[n]𝒟:𝑠delimited-[]𝑛𝒟s:[n]\to\mathcal{D}italic_s : [ italic_n ] → caligraphic_D with n1𝑛1n\leq 1italic_n ≤ 1 the category 𝒞s=Fun𝒟([n],𝒞)subscript𝒞𝑠subscriptFun𝒟delimited-[]𝑛𝒞\mathcal{C}_{s}=\operatorname{Fun}_{\mathcal{D}}([n],\mathcal{C})caligraphic_C start_POSTSUBSCRIPT italic_s end_POSTSUBSCRIPT = roman_Fun start_POSTSUBSCRIPT caligraphic_D end_POSTSUBSCRIPT ( [ italic_n ] , caligraphic_C ) is weakly contractible.

1.6 Proposition.

The properties (P) and (P1)subscript𝑃1(P_{1})( italic_P start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ) for a functor f:𝒞𝒟:𝑓𝒞𝒟f:\mathcal{C}\to\mathcal{D}italic_f : caligraphic_C → caligraphic_D are equivalent.

Proof.

Let s:[n]𝒟:𝑠delimited-[]𝑛𝒟s:[n]\to\mathcal{D}italic_s : [ italic_n ] → caligraphic_D be given (up to equivalence) by a sequence

x0s1x1s2snxn.superscriptsubscript𝑠1subscript𝑥0subscript𝑥1superscriptsubscript𝑠2superscriptsubscript𝑠𝑛subscript𝑥𝑛x_{0}\stackrel{{\scriptstyle s_{1}}}{{\to}}x_{1}\stackrel{{\scriptstyle s_{2}}%