CATEGORY WEIGHT: NEW IDEAS CONCERNING LUSTERNIK-SCHNIRELMANN CATEGORY Yuli B. Rudyak Introduction The concept of category weight was introduced by Fadell-Husseini [FH] and developed by Rudyak and Strom. Here we give a survey, some further development and applications of category weight. The Lusternik-Schnirelmann category of a topological space X, catX, is defined as the minimal number k such that X admits a numerable covering {A1; : :;:Ak+1} where each Ai is contractible in X. Lusternik and Schnirelmann [LS] introduced the invariant catX for manifolds. They proved that, for every connected smooth (=C1 ) closed manifold M, 1 + catM CritM := min{critf|f 2 C1 (M; R)} where critf is the number of critical points of a smooth real-valued function f* * on M. Afterwards Fox [Fox] suggested considering catX as an invariant of a space X. The basic information concerning the Lusternik-Schnirelmann category can be found in [Fox], [Sv], [J]. This (homotopy) invariant is quite far from other invariants like homotopy and homology groups, so it is difficult to compute catX. For example, in [G], Ganea asked whether cat(X x Sn) = catX + 1, n > 0, X connected. The affirmative claim is usually referred to as the Ganea conjecture. Recently it was disproved* * by Iwase [I], but it is still unclear whether it is true for manifolds. Here is a* *nother rather naive question. Let f : M ! N be a map of degree 1 of closed manifolds. Is it true that catM catN? One of the favorite and famous ways to estimate the Lusternik-Schnirelmann ca* *t- egory is a so-called cup-length estimation (Froloff-Elsholz [FE], Eilenberg). N* *amely, if u1. .u.n6= 0 for some ui2 eH*(X) then catX n. The idea of the proof is quite simple: if X = A1[ . .[.An where each Aiis contractible in X then ui|Ai= 0, and so u1. .u.n|(A1 [ . .A.n) = 0. However, the cup-length estimation is not perfec* *t. For example, if L = L2n+1pis the lens space with p an odd prime, then catL = 2n* *+1 (Krasnosel'ski, 1955), while the cup-length estimation gives us catL n + 1 onl* *y. Fadell-Husseini refined the cup-length estimation by suggesting that elements ui be equipped with "weights". Speaking informally, we say that the category weight of u (cwgt u) is k if u|(Ai1[ . .[.Aik) = 0 _____________ 1991 Mathematics Subject Classification. Primary 55M30, secondary 55N20, 55S* *30, 58F05. Typeset by AM S-TEX 1 2 YULI B. RUDYAK forPAi as above. Clearly, catX cwgtu if u 6= 0. Furthermore, cwgt(u1. .u.n) cwgt ui, and, since cwgtui 1, we conclude that X catX cwgt(u1. .u.n) cwgt ui n provided u1. .u.n6= 0. Certainly, this improves the cup-length estimation. For example, this establishes a short calculation of catL2n+1p, see [FH]. So, it se* *ems reasonable to find (indecomposable) elements of high category weight and, more generally, to be able to compute category weight. Unfortunately, category weight is not a homotopy invariant, i.e., cwgt h*u is not necessary equal to cwgt u fo* *r a homotopy equivalence h. This makes category weight difficult for calculations. Because of this, it makes sense to introduce a homotopy invariant version of category weight as the author did in a talk in the AMS Summer Research Institut* *e, Seattle, July 1996 (for the publication see [R2]). This invariant is called st* *rict category weight. A similar concept was also introduced by Strom [S2]. There are many ways to define swgtu; u 2 H*(X). One of them is: swgt u = min{cwgt h*u} where h runs over all homotopy equivalences Y ! X. Clearly, cwgtu swgtu, so we can use swgt to estimate cwgt. Furthermore, swgt has better multiplicative properties than cwgt, see x3. It turns out that swgt* * is quite manageable: for example, it is possible both to find many elements of high swgt(see x4) and to apply swgt to certain geometric problems (see xx6,7). Also, notice that strict category weight yields a decreasing filtration {Fn(X)} of H** *(X), Fn(X) := {u 2 H*(X)| swgtu n}. Here is another description of swgt: we have X ' BX, the classifying space for the loop space of X. Let {BnX} be the Milnor filtration of BX, and let in : BnX BX ' X. (Notice that in is homotopy equivalent to a map (1.5) described in x1.) Then swgtu = sup{k|i*k(u) = 0}: In other words, the above filtration {Fn(X)} is just the filtration Ker {i*n: H*(X) ! H*(BnX)}: This also shows how to apply the Eilenberg-Moore spectral sequence to the study of the Lusternik-Schnirelmann category, cf. [To]. It is clear that category weight can be defined in more general situations. F* *or example, we can consider an arbitrary (extraordinary) cohomology theory instead of H*(-). In this paper we consider even a little bit more general functors th* *en cohomology. Throughout this paper, we reserve the term "map" for continuous functions and the term "inessential map" for homotopy trivial maps (i.e., for maps which are homotopic to constant maps). We use the sign ' for homotopy of maps, and we write f 'o g when there is a pointed homotopy between pointed maps f and g. The reduced cone of a pointed map f is denoted by Cf. "Connected" always means "path connected". CATEGORY WEIGHT 3 Given a pointed topological space X, the (reduced) cohomology group Hk(X; ss) is defined as [X; K(ss; n)] (pointed homotopy classes) where the Eilenberg-Mac Lane space K(ss; n) is assumed to be ( homotopy equivalent to) a CW -space. Two exceptions to this agreement (in 3.10 and 4.2) are mentioned explicitly. Given x 2 Hk(X), the notation x|ssk(X)= 0 means that = 0 for every a 2 Im{h : ssk(X) ! Hk(X)} where h is the Hurewicz homomorphism. The paper is organized as follows. In x1 we consider the Lusternik-Schnirelma* *nn category. In x2 we give a definition and background properties of (strict) cate* *gory weight. In x3 we demonstrate multiplicative properties of (strict) category wei* *ght. In x4 we explain how to find elements of high category weight. In x5 we show how to use swgt for control of the Lusternik-Schnirelmann category and in x6 we app* *ly this technique to manifolds. In particular, we prove special cases of the Ganea conjecture and the degree conjecture. In x7 we apply our technique to the famous Arnold conjecture about symplectic fixed points. In fact, we idicate how to pro* *ve it for closed symplectic manifolds (M; !) with !|ss2(M)= 0 = c1|ss2(M). 1. Lusternik-Schnirelmann category 1.1. Definition. (a) ([Fet], [BG]) Given a map ' : A ! X, we say that cat' k if there is a numerable covering U1; : :;:Uk+1 of A such that '|Ui is inessenti* *al for every i. Then cat' = k if k is minimal with this property. Also, we set cat' = * *-1 if A = ;. (b) If i : A ! X is an inclusion then we set catX A := inf{cati} where i runs over all the inclusions i : U ! X of neighborhoods U of A. (c) ([LS]) We define the Lusternik-Schnirelmann category catX of a space X by setting catX := cat1X = catXX. 1.2. Lemma. Let f : X ! Y be a map with Y connected, and let x0 2 X be an arbitrary point. If catf = k then there is a numerable covering {B1; : :;:Bk+1} such that f|Bi is inessential and x0 2 B for every i. Proof. Let {A1; : :;:Ak+1} be a numerable covering of X with f|Aiinessential, a* *nd let {'1; : :;:'k+1} be a partition of unity dominated by {A1; : :;:Ak+1}. We set Ci:= '-1i[1=(2k + 2); 1] and Bi:= Ci[ {x0}. It is easy to see that {Bi} is the desired covering. 1.3. Proposition ([BG]). (i) For every diagram A -'!Y -f!X we have catf' min{cat'; catf}. In particular, catf min{catX; catY }. (ii) If ' ' : A ! X then cat' = cat . (iii) If h : Y ! X is a homotopy equivalence then cat' = cath' for every ' : A ! X. Notice that, in view of 1.3(iii), catX is a homotopy invariant of X. Also, it* * is easy to see that catX dim X for every connected CW -space X. 1.4. Proposition ([B], [Fox]). For every two connected CW -spaces X; Y we have max {catX; catY } cat(X x Y ) catX + catY: 4 YULI B. RUDYAK Let X be a connected space. Take a point x0 2 X, set fi P X = P (X; x0) = {! 2 XI fi!(0) = x0} and consider the fibration p : P X ! X; p(!) = !(1) with the fiber X. Given a natural number k, we use the short notation (1.5) pk : Pk(X) ! X for the map pX_*X_._.*.XpX_-z______": P_X_*X_._.*.XP-Xz_______"----!X k times k times where *X denotes the fiberwise join over X, see e.g. [J]. In particular, P1(X) * *= P X. 1.6. Theorem ([Sv, Theorems 3 and 190]). Let ' : A ! X be a map with X connected. Then cat' < k iff there is a map : A ! Pk(X) such that pk = '. For future references we fix the following simple information, the proofs can* * be found e.g. in [R2]. 1.7. Proposition. (i) Pk(X) is a fibration over X; (ii) If X has the homotopy type of a CW -space then P - k(X) does; (iii) catPk(X) k - 1. 1.8. Proposition ([Sv, Th. 21]). The map p2 : P2(X) ! X (and so the map w2 : W2(X) ! X) is homotopy equivalent over X to the map " : SX ! X. The concept of Lusternik-Schnirelmann category can be generalized as follows. Let T be the category of pointed spaces and pointed maps. Let E be the category whose objects are pairs (set, subset) and whose morphisms (U; V ) ! (U0; V 0) a* *re functions f : U ! U0 with f(V ) V 0. Given a contravariant functor F : T ! E with F (X) = (U; V ) for some X 2 T , the notation u 2 F (X) means that u 2 U. We say that u 2 F (X) = (U; V ) is trivial if u 2 V . Given u 2 F (Y ) and f : * *X ! Y , we write f*u for F (f)(u). Furthermore, given A X, we denote i*u by u|A. 1.9. Definition. Let F : T ! E be a contravariant functor satisfying the following properties: (1)(weak homotopy property) If f 'o g : X ! Y then f*u is trivial if g*u is. (2)(triviality property) F (pt) has the form (U; U), i.e., every u 2 F (pt) * *is trivial. Let X be a connected pointed space. Given u 2 F (X), we define the Lusternik- Schnirelmann category of u, catu, to be the minimal k with the following proper* *ty: there is a numerable covering {A1; : :;:Ak+1} where each Ai is a pointed subspa* *ce of X and u|Ai is trivial for every i = 1; : :;:k. 1.10. Examples. (a) Given a pointed topological space Y , set F (X) = TY (X) := (U; V ) where U is the set of all pointed maps X ! Y and V is the subset of all inessential (under non-pointed homotopy) maps. (The action of F on morphisms is given by the composition.) Then, the definition above leads to the invariant ca* *tf defined in 1.1. CATEGORY WEIGHT 5 (b) Every pointed set (U; u0) can be considered as the pair (U; {u0}) of sets. Thus, every functor from T to the category of pointed sets can be regarded as a functor T ! E . (c) Every abelian group can be regarded as a pointed set (with the base point 0). So, every functor from T to the category of abelian groups can be treated a* *s a functor T ! E . (d) (Fary [F]) Because of (c), a reduced cohomology theory (not necessarily additive) E*(-) on T yields the functor X 7! F (X) := (Ei(X); {0}). Then, given u 2 Ei(X), we have the invariant catu, the Lusternik-Schnirelmann category of the class u. (e) Given X 2 T , set F (X) = (U; V ) where U is the set of fiberwise homotopy equivalence classes of fibrations over X and V consists of fibrations which adm* *it a section. The action of F on morphism is given by passing to induced fibrations.* * If u 2 F (X) is the equivalence class of a fibration , then catu is the genus of ,* * [Sv]; another name is the sectional category, secat, [J]. Notice that the example (e) generalizes example (a) since catf = secatf* where = {p1 : P Y ! Y } for f : X ! Y . 1.11. Proposition. Let X be a connected pointed space. (i) catf*u min{catf; catu} for every f : X ! Y in T and every u 2 F (Y ). (ii) If f 'o g : X ! Y then catf*u = catg*u for every u 2 F (Y ). (iii) If h : X ! Y is a pointed homotopy equivalence then cath*u = catu for every u 2 F (Y ). 2. Category weight 2.1. Definition. Let F : T ! E be a functor as in 1.9, and let u 2 F (X). We do not require X to be connected. (a) We define the category weight of u (denoted by cwgtu) by setting fi cwgtu = sup{k fiu|A is trivial whenevercatXA < k} where A runs over all pointed closed subsets of X. (b) We define the strict category weight of u (denoted by swgtu) by setting fi swgtu = sup{k fi'*u is trivial for every map ' : A ! X in T withcat' < k}: Notice that cwgtu = 1 = swgtu for every trivial element u. In particular, the (strict) category weight of a map (cf. 1.10(a)) and a coh* *o- mology class (cf.1.10(d)) is defined. Category weight was defined by Fadell and Husseini [FH] (for F = H*(-)). Strict category weight was defined by Rudyak for F = E*(-) as in 1.10(d) and Strom for F as in 1.10(a), cf. [R2], [S2]; Strom ca* *lls it essential category weight. I must also note that Strom prefers to say that * *the (strict) category weight of a trivial element is not defined. 2.2. Theorem (cf. [R2]). (i) swgt u cwgt u, and cwgt u catX provided u is not a trivial element. Furthermore, for every map f : Y ! X in T we have catf swgtu provided f*u is non-trivial. Finally, swgtu 1 for every u. (ii) For every inclusion i : A ! X in T we have cwgti*u cwgtu. (iii) For every map f : Y ! X in T we have swgtf*u swgtu. 6 YULI B. RUDYAK (iv) If f 'o g : X ! Y then cwgtf*u = cwgtg*u and swgtf*u = swgtg*u. (v) swgtu = swgth*u for every pointed homotopy equivalence h. Notice that properties (iii) and (v) of 2.2 are not valid for cwgteven if f =* * H*(-), see [R2], [S2]. In other words, category weight is not a homotopy invariant, a* *nd this is a main motivation for introducing strict category weight. The following proposition gives us another description of strict category weight. 2.3. Proposition. Let X be a connected pointed space. Then for every u 2 F (X) we have: (i) swgtu = sup{k|p*k(u) is trivial}; (ii) If swgtu = k then cwgtp*k+1u = k; (iii) swgtu = minf2F {cwgt f*u} where F is the class of all maps f : Y ! X in T . Proof. (i) See [R2, 1.8(v)]. (ii) Since swgtu = k, we conclude that, by (i), p*k+1u is non-trivial. Now k = swgtu swgtp*k+1u cwgtp*k+1u catPk+1X k: (iii) Clearly, swgt u swgtf*u cwgt f*u. Now, if swgtu = k then, by (ii) swgtu = cwgtp*k+1u. 2.4. Theorem. Let X be a connected pointed space. (i) For every f : X ! Y in T and every u 2 F (Y ) we have catf (swgt u) cat(f*u): In particular, for every diagram X -f!Y -g!Z we have catf (swgt g) cat(gf). (ii) For every u 2 F (X) we have catXA (cwgt u) cat(u|A): In particular, for every map f : X ! Y we have catX A (cwgt f) cat(f|A). Formula (ii) was found by Strom [S2], formula (i) is an obvious analog of (ii* *). This theorem improves properties 1.11(i) and 2.2(i). Notice that the proof does* * not use properties (1) and (2) of 1.9. Proof. We prove only (i). Let catf = n, swgt u = k > 0. We must prove that cat(f*u) p := [n=k]. Let {A1; : :;:An+1} be a numerable covering of X such that f|Ai is inessential for every i. Without loss of generality we can assume * *Y to be connected, and so, by 1.2, we can assume that each Ai contains the base poin* *t. We set B1 := A1 [ . .[.Ak;B2 = Ak+1 [ . .[.A2k; : :;: Bp = A(p-1)k+1[ . .[.Apk;Bp+1 = Apk+1 [ . .[.An+1: Since cat(f|Bj) < k, we conclude that, by 2.2(i), (f|Bj)*u is trivial. So, (f*u* *)|Bj is trivial for every j = 1; : :;:p + 1. Thus, catf*u p. 2.5. Corollary ([S2]). For every f : X ! Y in T with X connected and every u 2 F (Y ) we have cwgt f*u (cwgt f) swgtu; swgt f*u (swgt f) swgtu: 2.6. Proposition ([R2], [S2]). Let X be a CW -space, and let f : X ! Y , X 2 T be an essential map such that ssi(Y ) = 0 for i > m. Then cwgt f m. In particular, if E is spectrum with ssi(E) = 0 for i > m then cwgtu q +m for eve* *ry u 2 Eq(X); u 6= 0. CATEGORY WEIGHT 7 3. Multiplicative properties Let F be a functor as in 1.9. Given a Puppe sequence X -f!Y -! Cf in T , we conclude, by 1.9(1), that the image of the composition F (Cf) -! F (Y ) -! F (X) consists of trivial elements. 3.1. Definition. A functor F as 1.9 is called half-exact if, for every pointed polyhedral pair (X; A), the sequence * i* F (X [ CA) -j!F (X) -! F (A) is "exact", i.e., i*u is trivial iff u = j*v for some v 2 F (X [ CA). Here i : * *A ! X and j : X ! X [ CA are the inclusions. Given (U1; V1); : :;:(Un; Vn) 2 E , we s* *et (U1; V1) x . .x.(Un; Vn) := (U1 x . .U.n; V1 x . .V.n): Let E; F be two half-exact functors, and let G be an arbitrary functor as in 1.* *9. Suppose that, for every two polyhedra X; Y (not necessarily finite), there is a natural transformation (where x is as described in (3.2)) (3.3) m : E(X) x F (Y ) ! G(X ^ Y ): This transformation yields a transformation o : E(X) x F (Y ) -m!G(X ^ Y ) -! G(X x Y ); where shrinks the wedge X _ Y . We set u v := o(u; v) for u 2 E(X); v 2 F (Y * *): If X = Y and : X ! X x X is the diagonal, we set uov := *(u v) 2 G(X). 3.4. Theorem (cf. [FH], [R2]). (i) For every pointed connected polyhedron X and every u 2 E(X); v 2 F (X) we have cwgt(uov) cwgtu + cwgtv: (ii) For every pointed connected CW -space X and every u 2 E(X); v 2 F (X) we have swgt(uov) swgtu + swgtv: Proof. (i) First, assume that cwgt u = k < 1; cwgtv = l < 1. Take a pointed k+l[ closed subspace A of X with catXA < k + l. Then, clearly, A Vi where each i=1 Viis open and contractible in X. Since A is closed, there is a subdivision of X* * with the following property: every simplex e with e \ A 6= ; is contained in some Vi* *, cf. [W, Theorem 35]. We let Ai; i = 1; : :;:rn be the union of all simplexes contai* *ned in Vi. Then A A1 [ . .[.Ak+l where each subpolyhedron Ai is contractible in X. We set K := A1 [ . .[.Ak [ {*}; L := Ak+1 [ . .[.Ak+l[ {*} and let iK : K X; iL : L X; jK : X X [CK; jL : X X [CL; h : X X [C(K [L) 8 YULI B. RUDYAK be the inclusions. Consider the commutative diagram *m E(X [ CK) x F (X [ CL) ----! G(X [ C(K [ L)) ? ? j*Kxj*L?y ?yh* *o E(X) x F (X) ----! G(X); where, for the top row, we have used the canonical homotopy equivalence (X [ CX0) ^ (Y [ CY 0) ' X x Y [ C((X x Y 0) [ (X0x Y )) which holds for any polyhedral pairs (X; X0) and (Y; Y 0). Since both i*Ku and * *i*Lv are trivial, u = j*Ku0; v = j*Lv0 for some u02 E(X [ CK); v02 F (X [ CL). Clear* *ly, uov = h(u0ov0). Thus, (uov)|A is trivial since A K [ L. If, say, cwgt u = 1 then we must prove that uov|A is trivial if catX A < 1. Asserting as above, we conclude that there is a pointed subpolyhedron K X with A K and catX K < 1. Then u|K is trivial, and hence (uv)|K is (take L = * in the above diagram). (ii) Let swgt(uv) = k. Then, by 2.3(ii), cwgt p*k+1(uv) = k. Recall that every CW -space is homotopy equivalent to a polyhedron, and so, by 1.7(ii), there is a homotopy equivalence " : Y ! Pk+1(X) such that Y is a polyhedron. We put f = pk+1" : Y ! X. Then f*(uv) 6= 0, and k = swgt(uv) cwgtf*(uv) catY k: Now swgt(uov)= cwgtf*(uov) cwgtf*u + cwgtf*v swgtf*u + swgtf*v swgtu + swgtv: 3.5. Theorem ([R2, 1.14]). swgt(u v) swgtu + swgtv. Notice that the corresponding inequality for cwgt is wrong, cf. 2.2(iii,v). Certainly, we can consider not only two functors E; F but any finite number of functors equipped with a natural transformation like (3.3). 3.6. Examples. (a) Given n arbitrary spectra E(1); : :;:E(n) and n pointed CW -spaces X1; : :;:Xn, we have the homomorphism : E(1)*(X1) . . .E(n)*(Xn) ! (E(1) ^ . .^.E(n))*(X1 x . .x.Xn): see [Sw]. Furthermore, consider the correspondence (not a homomorphism!) r : E(1)*(X1) x . .x.E(n)*(Xn)! E(1)*(X1) . . .E(n)*(Xn); r(u1; : :;:un)= u1 . . .un: So, we have a natural transformation o := r : E(1)*(X1) x . .x.E(n)*(Xn) ! (E(1) ^ . .^.E(n))*(X1 x . .x.Xn): CATEGORY WEIGHT 9 Now, by 3.5 X swgt(u1 . . .un) swgtui: for every ui 2 E(i)*(Xi). Furthermore, given ui 2 E(i)*(X), we have the element u1o. .o.un 2 (E(1) ^ . .^.E(n))*(X), and X X cwgt(u1o. .o.un) cwgt ui; swgt(u1o. .o.un) swgtui (in the first unequality X is assumed to be a polyhedron). (b) Given a ring spectrum E with a multiplication u : E ^ E ! E, for every n we have the iterated multiplication un : E_^_._.^.E-z____"-!E n times which yields a function o : E*(X1)x. .x.E*(Xn) -! (E ^. .^.E)*(X1x. .x.Xn) -(un)*--!E*(X1x. .x.Xn) where the first arrow is o from (a). In this case the element u1o. .o.un 2 E*(X* *) is usually denoted by u1. .u.n, and we have X (3.7) catX cwgt(u1. .u.n) cwgt ui n if u1. .u.n6= 0. Certainly, this refines the cup-length estimation catX n. 3.8. Example ([RO]). A closed connected symplectic manifold is a pair (M2n; !) where M is a connected closed smooth 2n-dimensional manifold and ! is a closed non-degenerate 2-form. Notice that in this case !n is a volume form for M. In particular, ! yields a non-trivial de Rham cohomology class which we denote also by !. Since !n 6= 0, we conclude that catM n. Rudyak-Oprea [RO] proved that swgt! = 2 provided !|ss2(M)= 0. Thus, if (M2n; !) is a symplectic manifold with !|ss2(M)= 0 then, by (3.7), catM swgt!n n swgt! = 2n; and hence catM = 2n (because catM dim M). 3.9. Examples. (a) ([S2]) Given two pointed spaces A; B, set E(-) = TA (-), F (-) = TB (-) and G(-) = TA^B (-) (see 1.10(a)). We define m : E(X)xF (Y ) ! G(X ^ Y ) by setting m(f; g) = f ^ g, and 3.4 is applicable to this case. (b) According to 1.10(e), the category weight cwgt of a fibration is defined. It turns out to be that, for every fibrations ; j over the same base X, cwgt( *X j) cwgt + cwgtj: This can be deduced from 3.4, but we indicate a direct proof. Let A; K; L be as* * in the proof of 3.4, and let s1 (resp. s2) be a section of over K (resp of j over* * L). Take a function ' : A ! [0; 1] such that '-1(0; 1] 2 K and '-1[0; 1) 2 L. We set s(x) := 2 {the total space of * j}; x 2 A: Then s is a desired section over A. Now, we can formulate (and, probably, explo* *it) the obvious analog of (3.7). The following result shows how Massey products help to estimate catX. Let H*(-; R) denote singular cohomology with coefficients in a commutative ring R. Given a matrix V over H*(X; R), we set cwgtV = min{cwgt v} where v runs over all entries of V . 10 YULI B. RUDYAK 3.10. Theorem ([R2]). Given X 2 T , let V1; : :;:Vn be matrices over H*(X; R). Suppose that the matric Massey product is defined. If 0 =2 then catX mini{cwgt V2i} + mini{cwgt V2i+1}: My feeling is that this result is somehow related to (3.7), but I can't say h* *ow explicitly. 4. Elements of high category weight The results of the previous section show that it makes sense to search (inde- composable) elements of high category weight. The first example of this kind was found by Fadell-Husseini [FH]. Namely, they proved that, for every odd prime p and connected X, (4.1) cwgt fiP nu 2 ifu 2 H2n+1(X; Z=p): (Actually, they proved that swgtfiP nu 2.) 4.2. Theorem. Let X 2 T , and let V1; : :;:Vn be matrices over H*(X; R) (singular cohomology with coefficients in a commutative ring). Suppose that the matric Massey product is defined. Then, for every V 2 and every entry u of V , we have cwgtu 2. This theorem was explicitly formulated in [R2]. Actually, it follows from 1.* *6, 1.8, and the result of Gugenheim-May [GM] that Ker{"* : H*(X) ! H*(SX)} contains all the matric Massey products. Note that 4.2 implies (4.1) since fiP * *nu 2 (p times) for every u 2 H2n+1(X; Z=p), Kraines [K]. Let E; F be two spectra, and let : E ! F be a (stable) cohomology operation. Without loss of generality we can assume that E = {En} and F = {Fn} are - spectra, i.e., that there are weak homotopy equivalences oen : En ! En+1, etc. Then yields a family n : En ! Fn of maps such that the diagram En --n--! Fn ? ? oeEn?y ?yoeFn En+1 -n+1---!Fn+1 commutes up to homotopy. Following Strom [S2], set d = d() = inf{k|k is an essential map} 4.3. Theorem ([S2]). swgt n = 1 for n > d, and swgtd 2. Notice that d() = 2n + 1 for = fiP n. This gives us another proof of (4.1). Indeed, if u 2 H2n+1(X; Z=p) then swgtfiP n(u) = swgt2n+1(u) swgt(2n+1) swgtu 2 swgtu 2: CATEGORY WEIGHT 11 4.4. Theorem ([S3]). Let G be a discrete group, and let E be a spectrum such that ssi(E) = 0 for i < m. Then swgtu k + m for every u 2 Ek(BG). For k = 2 this result goes back to Fadell-Husseini [FH]. Notice that, by 2.6, swgtu = k for every u 2 Hk(BG); u 6= 0. As an application of 4.4, we note the following corollary. 4.5. Corollary. Let M be a smooth manifold such that the structure group of its tangent bundle reduces to a discrete group G. If wI(M) 6= 0 then catM |I|.PHere I = {i1; : :;:ik}, wI is the Stiefel-Whitney class wi1. .w.ikand |I| = ik. Proof. It suffices to prove that swgtwI |I|. But this follows from 4.3 and 2.2* *(iii), since wI is induced from BG. The last example of elements of high category weight is as follows. 4.6. Theorem. Let X be a simply connected rational space, and let (; d) be the minimal Sullivan model for X. If a cohomology class x 2 H*(X; Q) = H*(; d) has the form x = [a1. .a.k] for some ai2 ; dimai 0, then swgtx k. Proof. Notice that the aiare not assumed to be cocycles and, hence, we can't ap* *ply (3.6). This follows from the result of Felix-Halperin [FeH, x3] that p*k: H*(X;* * Q) ! H*(Pk(X); Q) annihilates all elements of the form [a1. .a.k]. 5. Detecting elements 5.1. Definition. Let F be as 1.9. An element u 2 F *(X) is called a detecting element for X if swgtu = catX. We formulate the above definition for the general situation, but really we wi* *ll apply it to the case when F is a cohomology theory, as in 1.9(d). 5.2. Theorem ([R2]). Let X; Y be two connected p[ointed CW -spaces and let E; G be two spectra. Suppose that there are detecting elements u 2 E*(X); v 2 G*(Y ). If 0 6= u v 2 (E ^ G)*(X x Y ) then cat(X x Y ) = catX + catY , and u v is a detecting element for X x Y . By 2.2(i), 3.5 and 1.4, cat(X x Y ) swgt(u v) swgtu + swgtv = catX + catY cat(X x Y ): 5.3. Corollary. If a pointed CW -space X possesses a detecting element then cat(X x Sn) = catX + 1 for every n 0. In other words, the Ganea conjecture holds for X. Proof. We take a detecting element u 2 E*(X) and apply 5.2 to the case Y = Sn; Gn(Y ) = n(Sn), and v is given by the identity map Sn ! Sn. One can prove that u v 6= 0 (see e.g. [R2]), and the result follows. For example, Strom [S1] proved that a (q - 1)-connected CW -space X possesses a detecting element if catX = [dim X=q], and so such an X satisfies the Ganea conjecture. 12 YULI B. RUDYAK 5.4. Theorem ([R2]). Let R be a ring spectrum, and let E be an arbitrary R- module spectrum. Let Mn ; Nn be two closed connected HZ-orientable PL manifolds, and let f : N ! M be a map of degree 1 and such that N is R-orientable. If M possesses a detecting element u 2 E*(M) then catf = catM. In particular, catN catM. Proof. It is easy to see that f* : E*(M) ! E*(N) is monic. So, f*u 6= 0, and hence, by 2.2(i), catf swgtu = catM. Hence, by 1.3, catf = catM. The results above show that it is useful and important to know whether a space possesses detecting elements. Consider the Puppe sequence Pm (X) -pm-!X -jm-!Cm (X) := C(pm ) where pm : Pm (X) ! X is the fibration (1.5) and C(pm ) is the cone of pm . 5.5. Theorem ([R2]). Let X be a pointed connected CW -space with catX = k < 1. If jk is stably essential (i.e., the stable homotopy class of jk is non-zero* *) then X possesses a detecting element. In fact, the stable homotopy class of jm can be treated as a universal (among cohomology functors) detecting element. Also, we note that, for every X with catX = k, the fibration pk : Pk(X) ! X is a detecting element (see 1.10(e)). However, unfortunately, it is difficult to apply analogs of 5.2-5.4 to this cas* *e. 6. Manifolds Given a PL manifold M, we denote by M the stable normal bundle of M. 6.1. Theorem ([R2]). Let Mn ; n = dim M 4, be a closed (q - 1)-connected PL manifold, q 1. Suppose that there is a natural number m such that M |M(m) is trivial and n min{2q catM - 4; m + q catM - 1}: Then M possesses a detecting element. Putting m = n + 1 and m = 1, we get the following corollary. 6.2. Corollary. (i) Let M be a closed (q - 1)-connected stably parallelizable * *PL manifold, q 1. Suppose that 4 dim M 2(q catM - 2). Then M possesses a detecting element. (ii) Let q 1, and let Mn ; n = dim M 4 be a closed orientable (q - 1)- connected PL manifold such that q catM = n. Then M possesses a detecting element. Moreover, there exists a detecting element u 2 Hn (M; ssn(Cn(M))). 6.3. Corollary. Let M be as in 6:1. Then cat(M x Sm1 x . .x.Smn ) = catM + n for any natural numbers m1; : :;:mn. Other results about the Ganea conjecture for manifolds can be found in [R1], [R* *2], [S1]. Based on 6.2 and 5.4, we get the following theorem. 6.4. Theorem. (i) Let Mn be as in 6.2(i), and let f : Nn ! Mn be a map of degree 1 where N is a stably parallelizable PL manifold. Then catf = catM. In particular, catN catM. (ii) Let f : N ! M be a map of degree 1 of closed HZ-orientable PL manifolds. If catM = dim M then catf = catM = catN. CATEGORY WEIGHT 13 6.5. Corollary. Let M be an oriented PL manifold with catM = dim M, and let f : X ! M be a map of an arbitrary topological space such that f* : Hn (M; ssn(Cn(M))) ! Hn (M; ssn(Cn(M))) is a monomorphism. Then catf = catM. Proof. By 6.2, there is a detecting element u 2 Hn (M; ssn(Cn(M))), and f*u 6= * *0. Thus, cat f swgtu = catM: But, by 1.3(i), catf catM. 7. Applications to the Arnold conjecture In [A, Appendix 9] Arnold proposed a beautiful conjecture concerning the rela- tion between the number of fixed points of certain self-diffeomorphisms of a cl* *osed symplectic manifold (M; !) and the minimum number of critical points of any smooth (= C1 ) function on M. Let (M2n; !) be a closed symplectic manifold. A symplectomorphism OE : M ! M (i.e., a diffeomorphism with OE*! = !) is called Hamiltonian (or exact) if it belongs to the flow of a time-dependent Hamiltonian vector field on M. See [HZ]* * or [MS] for details. We define Arn(M; !) to be the minimum number of fixed points for any Hamiltonian symplectomorphism of M. The Arnold conjecture claims that the following inequality holds for every closed symplectic manifold (M; !): Arn(M; !) CritM: The conjecture, usually (but not universally) weakened by replacing CritM by the cup-length of M, has been proved under various hypotheses for various class* *es of manifolds ([CZ], [H], [Fl1], [Fl2]). Here the following theorem (formulated* * ex- plicitely in [R2] and based on Floer's approach) plays the crucial role. 7.1. Theorem. Let (M; !) be a closed connected symplectic manifold such that !|ss2(M)= 0 = c1(M)|ss2(M). Then there exists a map f : X ! M with the following properties: (i) X is a compact metric space; (ii) 1 + catf Arn(M; !); (iii) The homomorphism f* : Hn (M; G) ! Hn (X; G) is a monomorphism for every coefficient group G. The following theorem is proved in [R3] and [RO]. 7.2. Theorem. Let (M2n; !) be a closed connected symplectic manifold with !|ss2(M)= 0 = c1|ss2(M): Then Arn(M; !) CritM; i.e., the Arnold conjecture holds for M. Proof. First, note that, by 3.8, catM = 2n = dim M. Hence, 1 + catM CritM 1 + dimM = 1 + catM 14 YULI B. RUDYAK (the last inequality is a theorem of Takens [T]). So, 1 + catM = CritM. Thus, in view of 7.1, it remains to prove that catf catM where f is the map from 7.1. We give two proofs of this inequality. First proof. This follows from 6.5. Second proof. Rudyak-Oprea [RO] proved that swgt! = 2. Thus, since f*!n 6= 0, catf swgt!n 2n = catM: References [A] V.I. Arnold, Mathematical Methods in Classical Mechanics. Springer, Berlin Heidelberg New York 1989 [B] A. Bassi, Su alcuni nuovi invarianti delle varieta topologiche. Annali di Math. 16 (1935), 275-297 [BG] I. Berstein, T. Ganea, The category of a map and of a cohomology class. Fund. Math. 50 (1961/2), 265-279 [CZ] C. Conley, E. Zehnder, The Birkhoff-Lewis Fixed Point Theorem and a Conjecture of V.I. Arnold. Invent. Math. 73 (1983), 33-49 [FH] E. Fadell, S. Husseini, Category weight and Steenrod operations. Boletin de la Sociedad Matematica Mexicana 37 (1992), 151-161 [F] I. Fary, Sur la categorie des classes de homologie d'un espace. Proc. Intern. Congr. Math. 1954, 2, 215, North-Holland, Amsterdam 1957 [FeH] Y. Felix, S. Halperin, Rational L.-S. category and its applications, Tra* *ns. Amer. Math. Soc. 273 (1982), 1-38 [Fet] A. I. Fet, A connection between the topological properties and the number of extremals on a manifold (Russian). Doklady AN SSSR, 88 (1953), 415-417 [Fl1] A. Floer, Cuplength estimates on Lagrangian intersections. Comm. Pure Appl. Math 42 (1989), 335-356 [Fl2] A. Floer, Symplectic fixed points and holomorphic spheres. Commun. Math. Phys. 120 (1989), 575-611 [Fox] R. Fox, On the Lusternik-Schnirelmann category. Ann. of Math. 42 (1941), 333-370 [FE] S. Froloff, L. Elsholz, Limite inferieure pour le nombre des valeurs cri* *tic- ues d'une fonction, donne sur une variete. Math. Sbornik, 42, 5 (1935), 637-643 [G] T. Ganea, Some problems on numerical homotopy invariants. Sym- posium in Algebraic Topology, Seattle 1971, 23-30, Lecture Notes in Mathematics 249, Springer, Berlin 1971 [GM] V. K. A. M. Gugenheim, J. P. May, On the Theory and Applications of Differential Torsion Products. Memoirs Amer. Math. Soc. 142, AMS, Providence, Rhode Island 1974 [H] H. Hofer, Lusternik-Schnirelmann theory for Lagrangian intersections. Annales de l'inst. Henri Poincare- analyse nonlineare, 5 (1988), 465-499 [HZ] H. Hofer, E. Zehnder, Symplectic Invariants and Hamiltonian Dynamics, Birkh"auser, Basel 1994 CATEGORY WEIGHT 15 [I] N. Iwase, Ganea's conjecture on Lusternik-Schnirelmann category. Pre- print 1997 [J] I. M. James, On category, in the sense of Lusternik-Schnirelmann. To- pology 17 (1978), 331-349 [K] D. Kraines, Massey higher products. Trans. Amer. Math. Soc. 124 (1966), 431-449 [LS] L. A. Lusternik, L. G. Schnirelmann, Methodes topologiques dans le problemes variationels. Hermann, Paris 1934 [MS] D. McDuff, D. Salamon, Introduction to Symplectic Topology, Clarendon Press, Oxford 1995 [R1] Yu. B. Rudyak, On the Ganea conjecture for manifolds. Proc. Amer. Math. Soc. 125, 8, (1997), 2511-2512 [R2] Yu. B. Rudyak, On category weight and its applications. Topology, to appear [R3] Yu. B. Rudyak. On analytical applications of stable homotopy (the Arnold conjecture, critical points). Math. Zeitschrift, to appear [RO] Yu. B. Rudyak, J. Oprea, On the Lusternik-Schnirelmann Category of Symplectic Manifolds and the Arnold Conjecture. Math. Zeitschrift, to appear [S1] J. Strom, Two Special Cases of Ganea's Conjecture. Trans. Amer. Math. Soc., to appear [S2] J. Strom, Essential category weight. Preprint, 1997 [S3] J. Strom, Essential category weight and classifying spaces. Preprint, 1997 [Sv] A. Svarc, The genus of a fiber space. Amer. Math. Soc. Translations 55 (1966), 49-140 [Sw] R.W. Switzer, Algebraic Topology - Homotopy and Homology. Springer, Berlin Heidelberg New York 1975 [T] F. Takens, The minimal number of crirical points of a function on a com- pact manifold and the Lusternik-Schnirelmann category. Invent. Math. 6 (1968), 197-244 [To] G. Toomer, Lusternik-Schnirelmann Category and the Moore Spectral Sequence. Math. Z. 138 (1974), 123-143 [W] J. H. C. Whitehead, Simplicial spaces, nuclei, and m-groups. Proc. London Math. Soc. 45 (1939), 243-327 Mathematisches Institut, Universit"at Heidelberg, Im Neuenheimer Feld 288, D-69120 Heidelberg 1, Germany E-mail address: july@mathi.uni-heidelberg.de