ebook img

Yet another Hopf Invariant PDF

0.18 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Yet another Hopf Invariant

YET ANOTHER HOPF INVARIANT 5 1 JEAN-PAULDOERAENEANDMOHAMMEDELHAOUARI 0 2 g u A Abstract. TheclassicalHopfinvariantisdefinedforamapf:Sr →X. Here wedefine‘hcat’whichissomekindofHopfinvariantbuiltwithaconstruction 7 in Ganea’s style, valid for maps not only on spheres but more generally on 2 a ‘relative suspension’ f: ΣAW → X. We study the relation between this invariant and the sectional category and the relative category of a map. In ] particular,forιX:A→Xbeingthe‘restriction’offonA,wehaverelcatιX 6 T hcatf 6relcatιX+1andrelcatf 6hcatf. A Our aim here is to make clearer the link between the Lusternik-Schnirelmann . category (cat), more generally the ‘relative category’ (relcat), closely related to h t James’sectional category (secat), and the Hopf invariants. In order to do this, we a introduce a new integer, namely hcat, that combines the Iwaze’s version of Hopf m invariant [3], based on the difference up to homotopy between two maps defined [ for a given section of a Ganea fibration, and the framework of the sectional and 4 relativecategories,searchingforthe least integer suchthatthe Ganeafibrationhas v a section, possibly with additional conditions. To do this combination, we simply 2 define our invariant hcat, as the least integer such that the Ganea fibration has a 1 sectionσ with additionalconditionthat the correspondingtwo maps(f◦σ andω 7 n in this paper) are homotopic. 3 0 It appears that for f: Sr → X or even for f: ΣW → X, we obtain an integer . that can be either cat(X), or cat(X)+1. More generally, for any f: Σ W →X, 1 A 0 wehaverelcat(f◦θ)6hcat(f)6relcat(f◦θ)+1,whereθ: A→ΣAW isthe map 5 arising in the construction of ΣAW. 1 Insection2,westudythe influenceofhcatinahomotopypushout. Insection3, : v weintroducethe‘strong’versionofourinvariant,andweobtainanotherimportant Xi inequality: for any f: ΣAW → X, we have relcat(f) 6 hcat(f). In section 4, we give alternative equivalent conditions to get hcat. Applications and examples are r given. a 1. The Hopf category We work in the category of pointed topological spaces. All constructions are made up to homotopy. A ‘homotopy commutative diagram’ has to be understood in the sense of [4]. Recall the following construction: 2010 Mathematics Subject Classification. 55M30. Key words and phrases. Ganeafibration,sectionalcategory, Hopfinvariant. 1 2 JEAN-PAULDOERAENEANDMOHAMMEDELHAOUARI Definition 1. For any map ι : A → X, the Ganea construction of ι is the X X following sequence of homotopy commutative diagrams (i>0): >>A❊ ηi ⑥⑥⑥⑥ ❊❊❊❊ ιX ⑥ ❊ Fi❅⑥β❅⑥❅i⑥❅❅❅❅❅ α③i③+γ③i1③③③③❊③❊G<<"" i+1 gi gi+1 77//''X G i where the outside square is a homotopy pullback, the inside square is a homotopy pushoutandthemapgi+1 =(gi,ιX): Gi+1 →X isthewhiskermapinducedbythis homotopy pushout. The iteration starts with g0 =ιX: A→X. We set α0 =idA. For any i > 0, there is a whisker map θ = (id ,α ): A → F induced by i A i i the homotopy pullback. Thus we have the sequence of maps A θi //Fi ηi //A and θi is a homotopy section of ηi. Moreover we have γi ◦αi ≃ αi+1, thus also αi+1 ≃γi◦γi−1◦···◦γ0. We denote by γi,j: Gi →Gj the composite γj−1◦···◦γi+1◦γi (for i< j) and set γ =id . i,i Gi Of course, everything in the Ganea construction depends on ι . We sometimes X denote G by G (ι ) to avoid ambiguity. i i X Definition 2. Let ι : A→X be any map. X 1)Thesectionalcategory ofι istheleastintegernsuchthatthemapg : G (ι )→ X n n X X has a homotopy section, i.e. there exists a map σ: X → G (ι ) such that n X g ◦σ ≃id . n X 2)Therelativecategory ofι istheleastintegernsuchthatthemapg : G (ι )→ X n n X X has a homotopy section σ and σ◦ι ≃α . X n 3)The relative category of order k ofι is the leastintegern suchthatthe map X g : G (ι )→X has a homotopy section σ and σ◦g ≃γ . n n X k k,n Wedenotethesectionalcategorybysecat(ι ),therelativecategorybyrelcat(ι ), X X and the relative category of order k by relcat (ι ). If A = ∗, secat(ι ) = k X X relcat(ι ) and is denoted simply by cat(X); this is the ‘normalized’ version of X the Lusternik-Schnirelmann category. Clearly, secat(ιX) 6 relcat(ιX). We have also relcat(ιX) 6 relcat1(ιX), see Proposition 6 below. In the sequel, we will consider a given homotopy pushout: W η //A β θ (cid:15)(cid:15) (cid:15)(cid:15) A //Σ W A θ Inother words,the map θ is a mapsuchthat Pushcatθ ≤1 in the sense of[2]. We call this homotopy pushout a ‘relative suspension’ because in some sense, A plays the role of the point in the ordinary suspension. We also consider any map f: Σ W →X, and set ι =f ◦θ. A X YET ANOTHER HOPF INVARIANT 3 We don’t assume η ≃ β in general. This is true, however, if θ is a homotopy monomorphism,andin this casewe can ‘think’ ofι asthe ‘restriction’of f onA. X For n>1, consider the following homotopy commutative diagram: W β //A Ayyrrrηrrrr ✤✤ θ // Σ W vv♥θ♥♥♥♥♥♥♥ ❖❖❖❖❖θ❖❖''Σ W ✤ A A (†) (cid:15)(cid:15)✤ ωn (cid:15)(cid:15) αn−1 Fn−1(ιX) //Gn−1(ιX) f Azztttttt //G ((cid:15)(cid:15)ι )ww♦♦♦♦γ♦n♦−1 ◆◆◆g◆n◆−◆◆1//&&X(cid:15)(cid:15) αn n X gn where the mapW →Fn−1 is induced by the bottom outer homotopypullback and the mapω : Σ W →G is induced by the top inner homotopypushout. We have n A n f ≃g ◦ω by the ‘Whiskersmapsinside a cube’lemma (see [2], Lemma49). Also n n notice that αn ≃ωn◦θ ≃γn−1◦αn−1; so ωn ≃(αn,αn) is the whisker map of two copies of α induced by the homotopy pushout Σ W. Finally, for all k > 1, we n A can see that ω ≃γ ◦ω . n k,n k Definition3. TheHopfcategoryoff istheleastintegern>1suchthatg : G (ι )→ n n X X has a homotopy section σ: X →G (ι ) such that σ◦f ≃ω . n X n We denote this integer by hcat(f). Actually, speaking of ‘Hopf category of f’ is a misuse of language. We should speak of ‘Hopf category of the datas η, β and f’. Example 4. Let X =Σ W and f ≃id . Then, asmight be expected, hcat(f)= A X 1. Indeed, in this case, as g1 ◦ω1 ≃ f ≃ idX, ω1 is a homotopy section of g1. Moreover, ω1◦f ≃ω1◦idX ≃ω1, so hcat(f)=1. Example 5. Let X 6≃ ∗ and W = A∨A, β ≃ pr : A∨A → A and η ≃ pr : 1 2 A∨A → A the obvious maps. Then Σ W ≃ ∗ and we have no choice for f that A must be the null map f: ∗ → X. In this case the condition σ◦f ≃ ω is always n satisfied, so hcat(f)=secat(ι )=cat(X). X Notice that relcat is a particularcaseofhcat: WhenW =A, η ≃β ≃id , then A ιX ≃ f, ωn ≃ αn and hcat(f) = relcat(ιX). Also relcat1 is a particular case of hcat: When W = F0, then ΣAW ≃ G1, θ ≃ γ0 ≃ α1, and if, moreover, f ≃ g1, then ωn ≃γ1,n and hcat(f)=relcat1(ιX). Thefollowingpropositionshowsthattheseparticularcasesareinfactlowerand upper bounds for hcat(f). Proposition 6. Whatever can be f (and ι =f ◦θ), we have X secat(f)6relcat(ιX)6hcat(f)6relcat1(ιX)6relcat(ιX)+1. Proof. Consider the following homotopy commutative diagram (n>1): αn 00:: Gn A θ //ΣAW✈✈✈ω✈✈n gn ιX ■■■f■■..$$ X(cid:10)(cid:10) 4 JEAN-PAULDOERAENEANDMOHAMMEDELHAOUARI We see that if there is a map σ: X → G such that ω ≃ σ◦f then α ≃ σ◦ι n n n X and this proves the second inequality. Now consider the following homotopy commutative diagram (n>1): ωn 00==Gn ΣAW ω1 //G1③③③γ③1,n gn ❉❉g1 ❉❉"" (cid:10)(cid:10) f ..X We see that if there is a map σ: X →Gn such that γ1,n ≃σ◦g1 then ωn ≃ σ◦f and this proves the third inequality. The first inequality comes from secat(f) 6 secat(ι ) 6 relcat(ι ), the first of X X these two inequalities comes from [2], Proposition 29. Finally, the fourth inequality is proved in [1]. (cid:3) So hcat(f) establishes a ‘dichotomy’ between maps f: Σ W →X: A – Either hcat(f)=relcat(ι ) andwe haveaσ suchthat f◦σ ≃ω already X n for n=secat(ι ); X – either hcat(f)=relcat(ι )+1 andwe haveaσ suchthatf◦σ≃ω only X n for n>secat(ι ) X Our last example of the section shows that the inequalities of Proposition6 can be strict, and even that two may be strict at the same time: Example 7. Let X = ∗, A 6≃ ∗ and consider ι∗: A → ∗. We have Gi(ι∗) ≃ A ⊲⊳ ... ⊲⊳ A, the join of i+1 copies of A. For any k, γ ≃ id, so it cannot factorize k,k through ∗; but γk,k+1 is homotopic to the null map, so relcatk(ι∗) = k+1. Now consider f ≃ g1(ι∗): A ⊲⊳ A → ∗. As said before, in this case we have hcat(f) = relcat1(ιX). So we get secat(f)=0<relcat(ι∗)=1<hcat(f)=relcat1(ι∗)=2. 2. Hopf invariant and homotopy pushout Let us consider any homotopy commutative square: Σ W ρ // B A (‡) f κY (cid:15)(cid:15) (cid:15)(cid:15) X //Y χ Proposition 8. The homotopy commutative square above can be splitted into the following homotopy commutative diagram: f ΣAW //G1(ιX) // Gn(ιX) ++//X ρ χ (cid:15)(cid:15) (cid:15)(cid:15) (cid:15)(cid:15) (cid:15)(cid:15) B //G1(κY) // Gn(κY) 33//Y κY YET ANOTHER HOPF INVARIANT 5 Proof. Setφ=ρ◦θ. Since θ◦η ≃θ◦β, alsoφ◦η ≃φ◦β. Firstnotice thatwe can insert the original homotopy square inside the following homotopy commutative diagram: W η //A β~~⑥⑥⑥⑥⑥ {{✇✇✇✇✇ ❅❅❅❅ιX A // Σ W // X A f φ (cid:15)(cid:15) ρ (cid:15)(cid:15) φ B B χ B(cid:15)(cid:15) ⑤⑤⑤⑤⑤⑤⑤⑤ B(cid:15)(cid:15) ✈✈✈✈✈✈✈✈✈✈ ❆❆❆❆// Y(cid:15)(cid:15) κY Byinductiononn>1,startingfromthe outsidecube oftheabovediagramand φ0 =φ, we can build a homotopy diagram: W ❑❑❑❑❑❑%% vv♥♥♥♥♥//♥Fn−1✤✤(ιX) {{✈✈✈✈✈✈//A❂❂❂❂❂ιX(cid:30)(cid:30) Gn−1(ιX) ✤ //Gn(ιX) //X (cid:15)(cid:15) φn−1 (cid:15)(cid:15)✤ φn (cid:15)(cid:15)φ B ❑❑❑❑❑❑%% (cid:15)(cid:15) vv♥♥♥♥♥//♥Fn−1(κY) (cid:15)(cid:15) {{✈✈✈✈✈✈//BκY❂❂❂❂❂(cid:30)(cid:30) (cid:15)(cid:15)χ Gn−1(κY) //Gn(κY) gn // Y wherethedashedanddottedmapsareinducedbythehomotopypullbackFn−1(κY) and the homotopy pushout G (ι ) respectively. n X So we obtain a homotopy commutative diagram: W // A ~~⑦⑦⑦⑦⑦ zz✉✉✉✉✉ ❄❄❄❄ιX(cid:31)(cid:31) A // G (ι ) //X n X gn (cid:15)(cid:15) (cid:15)(cid:15) B B χ (cid:15)(cid:15) ⑦⑦⑦⑦⑦⑦⑦⑦⑦⑦ (cid:15)(cid:15) zz✉✉✉✉✉ ❄❄❄❄(cid:31)(cid:31) (cid:15)(cid:15) B //G (κ ) //Y n Y gn Finally take the homotopy pushout inside the upper and lower lefter squares to get the homotopy commutative diagram: Σ W // G (ι ) //X A ωn n X ρ χ (cid:15)(cid:15) (cid:15)(cid:15) (cid:15)(cid:15) B //G (κ ) //Y n Y and this gives the required splitting of the original square. (cid:3) Proposition 9. If the square ‡ is a homotopy pushout, then relcat(κ )6hcat(f). Y Asaparticularcase,whenB ≃∗,Y isthehomotopycofibreoff,andrelcat(κ )= Y cat(Y). So the Proposition asserts that hcat(f)>cat(Y). 6 JEAN-PAULDOERAENEANDMOHAMMEDELHAOUARI Proof. Let hcat(f) 6 n, so we have a homotopy section σ of g (ι ) such that n X σ◦f ≃ω . First apply the ‘Whisker maps inside a cube’ lemma to the outer part n of the following homotopy commutative diagram: Σ W //B ωxxrrnrrr A a(cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0) ■■■■α■n$$ G (ι ) // S //G (κ ) n X n Y b c X(cid:15)(cid:15) xxqqqqqqqΣAW // Y(cid:15)(cid:15) (cid:127)(cid:127)⑦⑦⑦⑦// B ❏❏❏❏❏❏$$Y(cid:15)(cid:15)gn where the inner horizontal squares are homotopy pushouts, and c and b are the whisker maps induced by the homotopy pushout S. Next build the following ho- motopy commutative diagram: Σ W //B A X xxqqqfqqqq //YκY(cid:127)(cid:127)⑦⑦⑦⑦ d σ (cid:15)(cid:15) ωxxrrnrrrΣAW (cid:15)(cid:15) (cid:0)(cid:0)(cid:0)(cid:0)(cid:0)(cid:0)a(cid:0)// B ■■■■α■n$$ G (ι ) // S // G (κ ) n X n Y b where d is the whisker map induced by the homotopy pushout Y. Let σ′ = b◦d. We haveg ◦σ′ ≃g ◦b◦d≃c◦d≃id and σ′◦κ ≃b◦d◦κ ≃b◦a≃α . (cid:3) n n C Y Y n Corollary 10. In the diagram ‡, if relcat(κ ) = relcat(ι )+1, then hcat(f) = Y X relcat(ι )+1. X Proof. ByProposition9,thehypothesisimpliesthathcat(f)>relcat(ι )+1. But X by Proposition 6, we have hcat(f)6relcat(ι )+1. So we have the equality. (cid:3) X It is now easy to exhibit examples of maps f with hcat(f) = relcat(ι ) + X 1. Indeed there are plenty examples of homotopy pushouts where relcat(κ ) = Y relcat(ι )+1: X Example 11. Let A=B =∗ andf: Sr →Sn be any ofthe Hopf maps S3 →S2, S7 →S4 or S15 →S8. So here relcat(ι )=cat(Sn)=1. On the other hand it is X well known that those maps have a homotopy cofibre Sn/Sr of category 2, so here relcat(κ )=cat(Sn/Sr)=2. By Corollary 10, we have hcat(f)=2. Y Example 12. Let f be the map u in the homotopy cofibration Z ⊲⊳Z //ΣZ∨ΣZ // ΣZ×ΣZ u t1 where Z ⊲⊳Z ≃Σ(Z ∧Z) is the join of two copies of Z and is also the suspension of the smash product of two copies of Z. Let A = B = ∗, ΣZ 6≃ ∗. We have relcat(ι ) = cat(ΣZ ∨ ΣZ) = 1 and relcat(κ ) = cat(ΣZ × ΣZ) = 2, so by X Y Corollary 10 again, we have hcat(u)=2. YET ANOTHER HOPF INVARIANT 7 Example 13. For i>1, let f be the map β in the Ganea construction: i A θi // F ηi // A ❅ i ❅ ❅ ❅ αi❅❅❅❅(cid:31)(cid:31) (cid:15)(cid:15)βi (cid:15)(cid:15)αi+1 Gi γi // Gi+1 Actually Fi is a join over A of i+1 copies of F0, and also a relative suspension Σ W where W is a relative smash product. For any i 6 relcat(ι ), we have A X relcat(α )=i, see [2], Proposition 23. So by Corollary 10 again, if i<relcat(ι ), i X we have hcat(β )=relcat(α )+1=i+1. i i 3. The Strong Hopf category In[2],weintroducedthestrongversionofrelcat,namelyRelcat. Inthissection, we introduce the strong version of hcat, namely Hcat. This gives an alternative way, sometimes usefull, to see if a map has a Hopf category less or equal to n. Also this will lead to a new inequality: hcat(f) > relcat(f). Consequently, if relcat(f)>relcat(ι ), then hcat(f)=relcat(ι )+1. X X Definition 14. The strong Hopf category of a map f: Σ W → X is the least A integer n>1 such that: – there are maps ι0: A → X0 and a homotopy inverse λ: X0 → A, i.e. ι0◦λ≃idX0 and λ◦ι0 ≃idA; – for each i, 06i<n, there is a homotopy commutative cube: W β //A η~~⑥⑥⑥⑥ zz✈✈✈✈✈ A // ΣAW ιi (♮) Z(cid:15)(cid:15) zi ζi+1 // X(cid:15)(cid:15) i i A(cid:127)(cid:127)⑧⑧⑧⑧ιi+1 //Xi(cid:15)(cid:15)+1{{✇✇✇χ✇✇i where the bottom square is a homotopy pushout. – X =X and ζ ≃f. n n We denote this integer by Hcat(f). Noticethatιi+1 ≃ζi+1◦θ ≃χi◦ιi. Inparticular,thismeansthatPushcat(ιi)6i in the sense of [2], Definition 3. For 0 6 i 6 n, define the sequence of maps ξ : X → X with the relation i i ξi = ξi+1 ◦ χi (when i < n), starting with ξn = idX. We have ξn ◦ ιn ≃ ιX and ξi ◦ ιi = ξi+1 ◦ χi ◦ ιi ≃ ξi+1 ◦ ιi+1 ≃ ιX by decreasing induction. Also ιX◦λ≃ξ0◦ι0◦λ≃ξ0. Moreover,for 0<i6n wehavewe haveξi◦ζi ≃f by the 8 JEAN-PAULDOERAENEANDMOHAMMEDELHAOUARI ‘Whiskermaps inside acube lemma’. Sowe havethe followinghomotopydiagram: W η //A }}④β④④④④ {{✇✇✇✇✇ ●●●●θ●## A // Σ W Σ W A A (cid:15)(cid:15) ζi+1 ιi (cid:15)(cid:15) ~~⑤⑤⑤⑤Zi (cid:15)(cid:15) ||①①①ι①i①+//1A❋❋❋ι❋X❋❋"" (cid:15)(cid:15)f Xi χi //Xi+1 ξi+1 // X We say that a map g: B → Y is ‘relatively dominated’ by a map f: B → X if there is a map ϕ: X →Y with a homotopy section σ: Y →X such that ϕ◦f ≃g and σ◦g ≃f. Proposition 15. A map g: Σ W → Y has hcat(g) 6 n iff g est relatively domi- A nated by a map f: Σ W →X with Hcat(f)6n. A Proof. Consider the map ω : Σ W → G (ι ) as in diagram † and notice that n A n Y Hcat(ω )6n. If hcat(f)6n, then f is relatively dominated by ω . n n Forthereversedirection,byhypothesis,wehaveamapϕandahomotopysection σ such that ϕ◦f ≃ g and σ◦g ≃ f; composing with θ, we have also ϕ◦ι ≃ ι X Y andσ◦ι ≃ι . Fromthe hypothesisHcat(f)6n,wegetasequenceofhomotopy Y X commutative diagrams, for 0 6 i < n, which gives the top part of the following diagram. Weshowbyinductionthatthemapϕ◦ξ : X →Y factorsthroughg : G (ι )→ i i i i Y Y up to homotopy. This is true for i = 0 since we have ξ0 ≃ ιX ◦λ, so ϕ◦ξ0 ≃ ϕ◦ιX ◦λ ≃ ιY ◦λ = g0◦λ. Suppose now that we have a map λi: Xi → Gi(ιY) such that g ◦λ ≃ϕ◦ξ . Then we construct a homotopy commutative diagram i i i Z //A zz✈z✈i✈✈✈✈ ✤✤i ιiyyt+t1tttt ❃❃❃❃(cid:31)(cid:31) Xi ✤ //Xi+1 //X ξi+1 (cid:15)(cid:15)✤ λi+1 λi (cid:15)(cid:15) {{✈✈✈✈✈Fi (cid:15)(cid:15) yytttαtit+t1//A❃❃❃❃❃(cid:30)(cid:30) (cid:15)(cid:15)ϕ Gi(ιY) // Gi+1(ιY) gi+1 //Y where Z ❴ ❴ ❴//F is the whisker map induced by the bottom homotopy pullback i i and λi+1: Xi+1 //Gi+1(ιY) is the whisker map induced by the top homotopy pushout. The composite gi+1◦λi+1 is homotopic to ϕ◦ξi+1. Hence the inductive step is proven. Atthe endofthe induction, wehaveg ◦λ ≃ϕ◦ξ =ϕ◦id =ϕ. Aswe have n n n X a homotopy section σ: Y →X =X of ϕ, we get a homotopy section λ ◦σ of g . n n n Moreover,we have (λ ◦σ)◦g ≃λ ◦f ≃λ ◦ζ ≃ω . (cid:3) n n n n n Example 16. If we consider any relative suspension Σ f: Σ W → Σ Z (and A A A in particular, of course, when A = ∗, any suspension Σf: ΣW → ΣZ), we have Hcat(Σ f) = 1. And so, any map g that is relatively dominated by a (relative) A suspension has hcat(g)=1. YET ANOTHER HOPF INVARIANT 9 In fact, by definition, a map g has Hcat(g) = 1 if and only if g is a (relative) suspension. There are maps forwhich the strong Hopfcategoryis greaterthanthe Hopfcategory: Forinstance,considerthe nullmapf: ∗→X ofExample5; ifX is a space with cat(X)=1 that is not a suspension, then f cannot be a suspension, so Hcat(f)>hcat(f)=1. Proposition 17. In the diagram ♮, we have Relcat(ζ )6i i As ω is a particular case of ζ , this implies Relcat(ω )6i. i i i Proof. Fori>0,letbuildthefollowinghomotopydiagramwherethethreesquares are homotopy pushouts: η W //Zi−1 ((//A β zi−1 θ (cid:15)(cid:15) (cid:15)(cid:15) (cid:15)(cid:15) A //Ci−1 //ΣAW ιi ci−1 ζi ιi−1 '' (cid:15)(cid:15) (cid:3)(cid:3) (cid:15)(cid:15) (cid:3)(cid:3) Xi−1 //Xi andwherethemapci−1 =(ιi−1,zi−1)isthewhiskermapinducedbythehomotopy pushout. Wehavesecat(ιi−1)6Pushcat(ιi−1)6i−1by[2],Theorem18. Sosecat(ci−1)6 i−1 by [2], Proposition 29. So Relcat(ci−1)6(i−1)+1=i by [2], Theorem 18. And this implies Relcat(ζ )6i by [2], Lemma 11. (cid:3) i Theorem 18. For any f: Σ W →X, we have A Relcat(f)6Hcat(f) and relcat(f)6hcat(f) Proof. If Hcat(f) = n, then we have f ≃ ζ in ♮. So Relcat(f) = Relcat(ζ ) 6 n n n by Proposition 17. If hcat(f) = n, then f is relatively dominated by ω . As Relcat(ω ) 6 n, we n n have relcat(f)6n by [2], Proposition 10. (cid:3) As a corollary, we get an indirect proof of Proposition 9 because relcat(κ ) 6 Y relcat(f)by[2],Lemma 11,thatassertsthata homotopypushoutdoesn’tincrease the relative category. It is not difficult to find an example where these inequalities are strict: Example 19. Let f be the map t1 in the homotopy cofibration Z ⊲⊳Z // ΣZ∨ΣZ // ΣZ×ΣZ u t1 LetA=∗,ΣZ 6≃∗. Ast1 isahomotopycofibre,wehaverelcat(t1)6Relcat(t1)6 1, see [2], Proposition 9. On the other hand, we have Hcat(t1) > hcat(t1) > relcat(ι )=cat(ΣZ×ΣZ)=2 by Proposition 6. X 10 JEAN-PAULDOERAENEANDMOHAMMEDELHAOUARI 4. Equivalent conditions to get the Hopf category Let be given any map f: Σ W → X with secat(ι ) 6 n and any homotopy A X section σ: X →G of g : G →X. Consider the following homotopy pullbacks: n n n Q π //Σ W π′ (cid:15)(cid:15) θnWA(cid:15)(cid:15) ❍❍❍❍❍❍❍❍❍❍❍❍❍❍❍❍❍❍ Σ W //H // Σ W A σ¯ n ηW A n f fn f (cid:15)(cid:15) (cid:15)(cid:15) (cid:15)(cid:15) X // G //X σ n gn where θnW = (ωn,idΣAX) is the whisker map induced by the homotopy pullback H . By the ‘Prism lemma’ (see [2], Lemma 46, for instance ), we know that the n hnoomticoetotphyatpπul≃lbaπc′ksionfcσe πan≃dηfWn i◦sθiWnd◦eeπd≃ΣAηWW,◦aσ¯n◦dπt′h≃atπη′nW. ◦σ¯ ≃ idΣAW. Also n n n Proposition20. Letbegiven anymap f: Σ W →X with secat(ι )6n andany A X homotopy section σ: X →G (ι ) of g : G (ι )→X. With the same definitions n X n n X and notations as above, the following conditions are equivalent: (i) σ◦f ≃ω . n (ii) π has a homotopy section. (iii) π is a homotopy epimorphism. (iv) θW ≃σ¯. n Proof. We have the following sequence of implications: (i) =⇒ (ii): Since σ ◦f ≃ ωn ≃ fn ◦θnW ◦idΣAW, we have a whisker map (f,idΣAW): ΣAW →Q induced by the homotopy pullback Q which is a homotopy section of π. (ii) =⇒ (iii): Obvious. (iii) =⇒ (iv): We have θW ◦π ≃σ¯◦π since π ≃π′. Thus θW ≃σ¯ since π is a n n homotopy epimorphism. (iv) =⇒ (i): We have σ◦f ≃f ◦σ¯ ≃f ◦θW ≃ω . (cid:3) n n n n Theorem 21. Let be a (q−1)-connected map ι : A → X with secatι 6 n. If X X Σ W is a CW-complex with dimΣ W < (n+1)q −1 then σ ◦f ≃ ω for any A A n homotopy section σ of g . n Proof. Recall that g is the (i+1)-fold join of ι . Thus by [4], Theorem 47, we i X obtain that, for each i > 0, g : G → X is (i+1)q−1-connected. As g and ηW i i i i have the same homotopy fibre, the Five lemma implies that ηW: H → Σ W is i i A (i+1)q −1-connected, too. By [5], Theorem IV.7.16, this means that for every CW-complexK withdimK <(i+1)q−1,ηW inducesaone-to-onecorrespondence i [K,H ] → [K,Σ W]. Apply this to K = Σ W and i = n: Since θW and σ¯ are i A A n both homotopy sections of ηW, we obtain θW ≃σ¯, and Proposition 20 implies the n n desired result. (cid:3) Example 22. Let A = ∗ and W = Sr−1, so Σ W = Sr, and X = Sm. In this A case secatι = catSm = 1. Hence Theorem 21 means that if r < 2m−1, we X have σ◦f ≃ω1, whatever can be f and σ: X →G1(ιX), so hcatf =1 and we get

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.