19 December 2003, University of Paris 13. THE EXPONENT PROBLEM IN HOMOTOPY THEORY JIE WU Department of Mathematics National University of Singapore Singapore 117543 Republic of Singapore matwuj@nus.edu.sg http://www.math.nus.edu.sg/∼matwujie This talk will consist of: 1. The Exponent Problem. 2. The Cohen Groups. 3. Shuffle Relations. 4. Configuration Spaces †Supported in part by the Academic Research Fund of the National University of Singapore. 1 2 JIE WU 1. The Exponent Problem It is not clear at least to me whether the answer to the Barratt conjecture is yes or no, but I will talk some progress along this direction. These progress seem interesting by themselves. Barratt Conjecture. Let X be a path-connected co-Hspace such that the degree map [pr ] : X → X is null homotopic. Then pr+1π∗(X) = 0. Barratt Conjecture for Maps. Let X be a path-connected co-H-space and let f : X → Y be a map such that [f ] is of order pr in [X, Y ]. Then pr+1 Im(f∗ : π∗(X) → π∗(Y )) = 0. The first statement follows from the second statement because one can choose Y = X and f = idX . So far essentially only knew that the Barratt conjecture holds for the cases when X is a mod pr Moore space for p > 2. For p = 2 and r > 1, it was known by F. Cohen that π∗(P n(2r )) has a bounded exponent. For mod 2 Moore spaces, it is even not clear where π∗(P n(2)) has a bounded exponent. It was THE EXPONENT PROBLEM IN HOMOTOPY THEORY 3 known that Z/8 occurs as a summand of π∗(P n(2)) for infinite times in different dimensions for n ≥ 3. • For technical reasons, we assume that X is a double (or triple or even higher) suspension in case. • From the cofibre sequence Sn [pr ] - Sn j - P n+1(pr ), the Barratt conjecture (for double suspensions) is equivalent to that pr+1 : Im((idY ∧j)∗ : π∗(Y ∧ S n) → π∗(Y ∧ P n+1(pr )) for n ≥ 2 and any Y because a f : Y ∧ S n → Z is of order pr in [Y ∧ S n, Z] if and only if it factors through Y ∧ P n+1(pr ), and idY ∧j : Y ∧ S n → Y ∧ P n+1(pr ) is of order pr . 4 JIE WU F Basic Ideas (due to F. Cohen): Assume that X = ΣX 0 is a suspension. Given a map f : ΣX 0 → Y , let f 0 : X 0 → ΩY be the adjoint map of f . Observe that Ωf : ΩX = ΩΣ0X → ΩY is the (unique up to homotopy) H-map such that Ωf |X 0 ' f 0. • Step 1. Construct certain group H (we will go through how to construct H) with a group homomorphism θ : H → [ΩΣX 0, ΩY ]. The group H is noncommutative and the image of θ contains [Ωf ] and many other homotopy classes given by composite of Hopf invariants and Whitehead products. Roughly speaking H should be big enough such that we can do something for making relations between the powers of [Ωf ] and other types of maps. • Step 2. Consider the commutative diagram H θ - [ΩΣX 0, ΩY ] Ω : [g] 7→ [Ωg] ? H̃ θ̃- 2 ? 0 [Ω ΣX , Ω2Y ], THE EXPONENT PROBLEM IN HOMOTOPY THEORY 5 where H̃ is certain extension of certain quotient of H by adding possible relations (for instance since [Ω2Σ0X, Ω2Y ] is abelian, the composite Ω ◦ θ factors through the abelianization of H) and, if possible, adding new elements. Assume that the answer to the Barratt conjecture would be positive. We wish that H̃ has a bounded exponent by adding enough relations to H. (Certainly one could consider triple loop spaces if there are good relations occur after looping twice or more.) • I have no ideas what H̃ looks like. But I will talk certain relations to H. These relations do not seem good enough for answering the Barratt conjecture, but they are canonical in some sense. 6 JIE WU 2. The Cohen Groups Let X be a path-connected space. Recall that the James construction J(X) ' ΩΣX. From the James filtration Xn qn ? ? X = J1(X) ⊆ · · · ⊆Jn(X)⊆ · · · ⊆ J(X), there is a tower of groups [J(X), ΩY ] · · · [Jn(X), ΩY ] · · · [X, ΩY ] ∩ qn∗ n ? [X , ΩY ] by suspension splitting theorem, that is, [J(X), ΩY ] = lim[Jn(X), ΩY ] n with representation qn∗ : [Jn(X), ΩY ] ⊂ - [X n, ΩY ]. The coordinate inclusions and projections di : X n+1 → X n di : X n → X n+1 (x0 , x1 , . . . , xn ) 7→ (x0 , . . . , xi−1 , xi+1 , . . . , xn ) (x0 , x1 , . . . , xn−1 ) 7→ (x0 , . . . , xi−1 , ∗, xi , . . . , xn−1 ) THE EXPONENT PROBLEM IN HOMOTOPY THEORY 7 induce faces di = di∗ and cofaces di = d∗i on G = {[X n+1, ΩY ]} such that G is a ∆-group and a co-∆-group with the relation di−1 dj for j < i, i dj d = id for j = i, i d dj−1 for j > i. This relation is different from the third identity of simplicial sets, and we call this kind of objects bi-∆-groups. (I do not know any references for studying these objects.) • The faces are used by obtaining [Jn(X), ΩY ] by the following lemma. F Lemma. Let X and Z be path-connected spaces. Suppose that X is a co-H-space or Z is an H-space. Then [Jn+1(X), Z] is the equalizer of the faces di = di∗ : [X n+1, Z] → [X n, Z] for 0 ≤ i ≤ n. • The cofaces can be used as operations for constructing Hopf invariants combinatorially. It turns out that many classical results such as the distributivity law hold for any bi-∆-group. • There are two ways to obtain the Cohen group H. The first way was given by F. Cohen. Given a map f : X → ΩY , he 8 JIE WU constructed a group homomorphism from the free group θ : Fn+1 → [X n+1, ΩY ] by sending the generators to X n+1 coordinate projections - X f - ΩY. When X is a suspension, he discovered certain commutator relations (the iterated commutators on generators such that one of the generators occurs at least twice) and so the map θ factors through a quotient group Kn+1 of Fn+1. Then he took the equalizer of faces to obtain a group Hn with a group homomorphism θ : Hn → [Jn(X), ΩY ] for any suspension X and any given map f : X → ΩY . By taking limits, then he obtained a group homomorphism θ : H = lim Hn → [J(X), ΩY ] n for any suspension X and any given map f : X → ΩY . One important property of Hn is that: THE EXPONENT PROBLEM IN HOMOTOPY THEORY 9 FCohen Theorem. H is a progroup and the kernel of Hn Hn−1 is isomorphic to Lie(n), where Lie(n) over a ring R (in this case R = Z) is given as follows: Let V be a free n-dimensional R-module with a basis {x1, . . . , xn}. Then Lie(n) is the R-submodule of V ⊗n spanned by Lie elements [[xσ(1), xσ(2)], . . . , xσ(n)] for σ ∈ Sn. The symmetric group Sn acts on Lie(n) by permuting letters {x1, . . . , xn}. • The elements in H are given by (possibly infinite) products of composites J(X) Hn - J(X (n)) J(α) - J(X (n)) ΩWn - J(X) f˜ - ΩY, where Hn is the James-Hopf map, Wn is the Whitehead product, f˜ is the H-map such that f˜|X ' f and α is a linear combination of permutations. 10 JIE WU 3. Shuffle Relations • Observe that Ω : [X, Y ] → [ΩX, ΩY ] = [ΣΩX, Y ] is induced by the evaluation map σ : ΣΩX → X. Recall that there is a homotopy pull-back diagram σ ΣΩX - X ∆ ? ? X ∨X ⊂ - X ×X ¯:X → for any space X. In particular, the reduced diagonal ∆ X ∧ X is null homotopic after looping. The shuffle relations are essentially obtained by considering elements in H that occur as composite J(X) ¯ ∆ - J(X) ∧ J(X) any map - ΩY. ¯ ' ∗. These maps are null homotopic after looping because Ω∆ ¯ : J(X) → J(X) ∧ F Lemma. If X is a co-H-space, then ∆ J(X) preserves the filtration up to homotopy, where the filtration on J(X) ∧ J(X) is given by the product filtration. The proof is almost straightforward. Let µ0 : X → X ∨ X be the comultiplication. Then the diagonal ∆ : X n → X n ×X n = X 2n preserves the filtration up to homotopy, namely (up to THE EXPONENT PROBLEM IN HOMOTOPY THEORY 11 homotopy) ∆ maps into the subspace of X 2n that has at most n coordinates different from the basepoint. By taking quotients, one get that (up to homotopy) ∆ : Jn(X) → Jn(X) × Jn(X) S maps into the subspace Ji(X) × Jj (X), and so does the i+j≤n reduced diagonal. • Let ψ̄ : J(X) → J(X) ∧ J(X) be the filtered map with ψ̄ ' ¯ Then there is a commutative diagram of cofibre sequences ∆. Jn−1(X) ⊂ - ψ̄ Jn(X) - ψ̄ X (n) shuffles ψ̄ ? ? [ Ji(X) ∧ Jj (X) n−1 _ ? ⊂ - i+j≤n−1 [ Ji(X) ∧ Jj (X) i+j≤n - X (i) ∧ X (n−i) i=1 F Lemma. Let X be a path-connected co-H-space. Then the composite Jn(X) → X (n) n−1 _ ψ̄ - X (i) ∧ X (n−i) any map - ΩY i=1 is null homotopic after looping. • Let R be the quotient group of H by adding the shuffle relations. Then the composite H → [J(X), ΩY ] → [ΩJ(X), Ω2Y ] factors through R. The group R has the following properties: • R is abelian. 12 JIE WU • There is commutative diagram of the tower of groups H · · · Hn ? ? R · · · Rn Hn−1 · · · H1 ? ? Rn−1 · · · R1 Theorem. Let Kern be the kernel of Rn → Rn−1. Then there is an exact sequence of groups 0 → Ext−1 R(Sn ) (Lie(n), Lie(n)) → Kern → EndR(Sn ) (Lie(n)), where the ground ring R = Z, Z(p) (if X is a p-local space) or Z/pr (if f : X → ΩY is of order pr ). • The composition operation on [J(X), J(X)] induces an operation on H such that H is a semi-ring and θ : H → [J(X), J(X)] corresponding the f = idΣX is the morphism of semi-rings. The abelian group R has the better property that R is a ring and the quotient H → R is a morphism of semi-rings. • Paul Selick and I have been to introduce a functor Amin(X) which roughly speaking is the functorial smallest retract of J(X) that contains the bottom cell. There is a subring Rmin of R such that Rmin is a local ring with a representation Rmin → [ΩAmin(X), ΩAmin(X)]. THE EXPONENT PROBLEM IN HOMOTOPY THEORY 13 4. Configuration Spaces Now we are going to give some remarks on using configuration spaces for understanding the evaluation map σ : ΣΩ2Σ2X → ΩΣ2X. Let (M, M0) be a pair of manifolds. Recall that the configuration space C(M, M0; X) = ∞ a F (M, n) ×Sn X n/ ≈ n=1 filtered by configuration length with Dn(M, M0; X) = Cn/Cn−1 = F (M, n)/F (M |M0, n)∧Sn X (n), where F (M |M0, n) is the subspace of F (M, n) consisting of all configurations with at least one coordinate in M0. We use the following basic property of configuration spaces: F Let M be a smooth compact manifold and let M0 and N be the smooth compact submanifolds of M with codimN = 0. If N/M0 ∩ N or X is path connected, then C(N, N ∩ M0; X) → C(M, M0; X) → C(M, N ∪ M0; X) is a quasifibration. 14 JIE WU • Let E be the pull-back of the diagram - C(M × (I, I0 ∪ I1 ); X) E q ? ? C(M × (I, I0 ∪ I0.5 ); X) ∨ C(M × (I, I0.5 ∪ I1 ); X) ⊂ - C(M × (I, I0 ∪ I0.5 ∪ I1 ); X), where I = [0, 1] and It is a small neighborhood of t for t ∈ I. Since q is a quasi-fibration and q is homotopic to the diagonal map, the space E ' ΣΩC(M × (I, I0 ∪ I1); X) ' ΣC(M × I; X) and the map σ : E → C(M × (I, I0 ∪ I1); X) is an evaluation map. It follows that the evaluation map σ : ΣC(M × I; X) → C(M × (I, ∂I); X) ' C(M ; ΣX) preserves that configuration filtration and so ΣCn−1(M × I; X) - ΣCn(M × I; X) σn−1 ΣDn(M × I; X) σ̄n σn ? Cn−1(M ; ΣX) - ? - Cn(M ; ΣX) ? - Dn(M ; ΣX) THE EXPONENT PROBLEM IN HOMOTOPY THEORY 15 F Theorem. The map σ̄n : ΣDn(M × I; X) - Dn(M ; ΣX) is homotopic to the suspension of the quotient map F (M × I; n)+ ∧Sn X (n) - F (M × I; n)/An ∧Sn X (n), where An is the subspace of F (M × I; n) consisting of all configurations with the property that at least one coordinate lies in M × I0 and at least one coordinate lies in M × I1. • Roughly speaking the Sn-equivariant map (F (M × I, n), ∅) - (F (M × I, n), An) gives the evaluation map in certain sense. • For the case M = I, the composite ΣDn(R2; X) σ̄n - (ΣX)(n) n−1 _ ψ̄ (ΣX)(i) ∧ (ΣX)(n−i) - i=1 is null homotopic. 16 JIE WU • For the Barratt conjecture, one might consider ΣCn(R2; X) - ΣDn(R2; X) σ̄n σn ? ? Jn(ΣX) ? cofibre - certain maps (ΣX) (n) - Σ2Cn−1(R2; X) certain maps certain maps ? ΩY w w w w w w w w w w - ΩY -