ON CERTAIN GEOMETRIC AND HOMOTOPY PROPERTIES OF CLOSED SYMPLECTIC MANIFOLDS Ra'ul Ib'a~nez, Yuli Rudyak, Aleksy Tralle and Luis Ugarte Abstract. The paper deals with relations between the Hard Lefschetz prope* *rty, (non)vanishing of Massey products and the evenness of odd-degree Betti nu* *mbers of closed symplectic manifolds. It is known that closed symplectic manif* *olds can violate all these properties (in contrast with the case of Kaehler manifo* *lds). However, the relations between such homotopy properties seem to be not analyzed. * *This analysis may shed a new light on topology of symplectic manifolds. In th* *e paper, we summarize our knowledge in tables (different in the simply-connected a* *nd in symplectically aspherical cases). Also, we discuss the variation of symp* *lectically harmonic Betti numbers on some 6-dimensional manifolds. 1. Introduction Homotopy properties of closed symplectic manifolds attract the attention of * *ge- ometers since the classical papers of Sullivan [S] and Thurston [Th]. On one ha* *nd, ös ft" homotopy techniques help in the solution of many äh rd" problems in sym- plectic geometry, cf. [G1, McD, RT1, TO]. On the other hand, it is still unknown if there are specific homotopy properties of closed manifolds dependent on the * *ex- istence of symplectic structures on them. It turns out that symplectic manifol* *ds violate many specific homotopy conditions shared by the Kähler manifolds (which form a subclass of symplectic manifolds). In particular, if M is a closed Kähl* *er manifold then the following holds: (1) all the odd-degree Betti numbers b2i+1(M) are even; (2) M has the Hard Lefschetz property; (3) all Massey products (of all orders) in M vanish. It is well known (and we shall see it below) that closed symplectic manifolds v* *iolate all the homotopy properties (1) - (3). However, it is not clear whether propert* *ies (1) - (3) are independent or not, in case of closed symplectic manifolds or cer* *tain classes of these ones. In other words, can a combination of the type (1) - (2) - non-(3) be realized by a closed symplectic manifold (possibly, with prescribed properti* *es). The knowledge of an answer to this question might shed a new light on the whole understanding of closed symplectic manifolds. In Theorem 3.1 we have summarized our knowledge by writing down the corre- sponding tables. We have considered two classes of symplectic manifolds: the cl* *ass ______________ 1991 Mathematics Subject Classification. 53C15, 57R19. Typeset by AM S-T* *EX 1 2 RA'UL IB'A~NEZ, YULI RUDYAK, ALEKSY TRALLE AND LUIS UGARTE of symplectically aspherical symplectic manifolds and the class of simply-conne* *cted symplectic manifolds. Recall that a symplectically aspherical manifold is a sy* *m- plectic manifold (M, !) such that !|ß2(M) = 0, i.e. Z f# ! = 0 S2 for every map f : S2 ! M. In view of the Hurewicz Theorem, a closed symplec- tically aspherical manifold always has a non-trivial fundamental group. It is * *well known that symplectically aspherical manifolds play an important role in geomet* *ry and topology of symplectic manifolds, [F, G2, H, RO, RT2]. The next topic of the paper is about symplectically harmonic forms on closed symplectic manifolds. Brylinski [B] and Libermann (Thesis, see [LM]) have intro- duced the concept of a symplectic star operator * on a symplectic manifold. In a sense, it is a symplectic analog of the Hodge star operator which is defined * *in terms of the given symplectic structure !. Using this operator, one defines a s* *ym- plectic codifferential ffi := (-1)k+1 (*d*), deg ffi = -1. Now we define symple* *ctically harmonic differential forms ff by the condition ffiff = 0, dff = 0. Let *hr(M, !) denote the space of all symplectically harmonic forms on M. Clea* *rly, the space Hkhr(M) := khr=( khr\ Im d) is a subspace of the de Rham cohomology space Hk (M). Here we also have an interesting relation between geometry and homotopy theo* *ry. For example, Mathieu [M] proved that Hkhr(M, !) = Hk (M) if and only if M has the Hard Lefschetz property. We will also see that the Lefschetz map Lk : Hm-k (M) ! Hm+k (M), dim M = 2m (multiplication by [!]k) plays an important role in studying of Hkhr(M, !). We set hk(M, !) = dim Hkhr(M, !). According to Yan [Y], the following questi* *on was posed by Boris Khesin and Dusa McDuff. Question: Are there closed manifolds endowed with a continuous family !t of symplectic structures such that hk(M, !t) varies with respect to t? Yan [Y] constructed a closed 4-dimensional manifold M with varying h3(M). So, he answered affirmatively the above question. (Actually, Proposition 4.1 from [Y] is wrong, the Kodaira-Thurston manifold * *is a counterexample, but its Corollary 4.2 from [Y] is correct because it follows * *from our Lemma 4.4. Hence, the whole construction holds.) However, the Yan's proof was essentially 4-dimensional. Indeed, Yan [Y] wrot* *e: öF r higher dimensional closed symplectic manifolds, it is not clear how to * *answer the question in the beginning of this section", i.e. the above stated question. In this note we prove the following result (Theorem 4.6): There exists at le* *ast one 6-dimensional indecomposable closed symplectic manifold N with varying h5(N). HOMOTOPY PROPERTIES OF SYMPLECTIC MANIFOLDS 3 Moreover, Yan remarked that there is no 4-dimensional closed symplectic nil- manifolds M with varying dim H*hr(M). On the contrary, our example is a certain 6-dimensional nilmanifold. Remark(November 2001). During preparartion the paper for the publication, we made some progress in this area. Namely, now we know that, for 6-dimensional closed manifolds, the numbers hi for i = 3, 4, 5 can vary, and there are precis* *ely ten 6-dimensional nilmanifolds with varying hi, [IRTU1,IRTU2]. 2. Preliminaries and notation Given a topological space X, let (MX , d) be the Sullivan model of X, that i* *s, a certain natural commutative DGA algebra over the field of rational numbers Q which is a homotopy invariant of X, see [DGMS, TO, S] for details. Furthermore, if X is a nilpotent CW -space of finite type then (MX , d) completely determines the rational homotopy type of X. A space X is called formal if there exists a DGA-morphism æ : (MX , d) ! (H*(X; Q), 0) inducing isomorphism on the cohomology level. Recall that every closed Kähler manifold in formal [DGMS]. We refer the reader to [K, Ma, RT1] for the definition of Massey products. I* *t is well known and easy to see that Massey products yield an obstruction to formali* *ty [DGMS, RT1, TO]. In other words, if the space is formal then all Massey products must be trivial. Thus, all the Massey products in every Kähler manifold vanish. We need also the following result of Miller [Mi]: 2.1. Theorem. Every closed simply-connected manifold M of dimension 6 is formal. In particular, all Massey products in M vanish. The next homotopy property related to symplectic (in particular, Kähler) str* *uc- tures is the Hard Lefschetz property. Given a symplectic manifold (M2m , !), we denote by [!] 2 H2(M) the de Rham cohomology class of !. Furthermore, we de- note by L! : k(M) ! k+2 (M) the multiplication by ! and by L[!]: Hk (M) ! Hk+2 (M) the induced homomorphism in the de Rham cohomology H*(M). As usual we write L instead of L! or L[!]if there is no danger of confusion. We say that a symplectic manifold (M2m , !) has the Hard Lefschetz property if, for ev* *ery k, the homomorphism Lk : Hm-k (M) ! Hm+k (M) is surjective. In view of the Poincar'e duality, for closed manifolds M it me* *ans that every Lk is an isomorphism. We need also the following result of Gompf [G1, Theorem 7.1]. 2.2. Theorem. For any even dimension n 6, finitely presented group G and integer b there is a closed symplectic n-manifold M with ß1(M) ~=G and bi(M) b for 2 i n - 2, such that M does not satisfy the Hard Lefschetz condition. Furthermore, if b1(G) is even then all degree-odd Betti numbers of M are even. We denote such manifold M by M(n, G, b). 4 RA'UL IB'A~NEZ, YULI RUDYAK, ALEKSY TRALLE AND LUIS UGARTE 2.3. Remark. Theorem 7.1 in [G1] is formulated in a slightly different way, but the proof is based on constructing of M by some "symplectic summation" in a way to violate the Hard Lefschetz property. In our explicit constructions we will need some particular classes of manifo* *lds, namely, nilmanifolds, resp. solvmanifolds. These are homogeneous spaces of the form G= , were G is a simply connected nilpotent, resp. solvable Lie group and is a co-compact discrete subgroup (i.e. a lattice). The most important informat* *ion for us is the following (see e.g. [TO] for the proofs): 2.4. Recollection. (i) Let g be a nilpotent Lie algebra with structural constan* *ts cijkwith respect to some basis, and let {ff1, ..., ffn} be the dual basis of g** *. Then the differential in the Chevalley-Eilenberg complex ( *g*, d) is given by the formu* *la X ij dffk = - ck ffi^ ffj. 1 i 2 H*(M), ff, fi 2 H*(M), then ]CP m posses* *ses a non-trivial Massey triple product even for m - n = 3. Proof. (i) and (ii) are proved in [McD], (i) and (iii) are proved in [RT1]. HOMOTOPY PROPERTIES OF SYMPLECTIC MANIFOLDS 5 2.6. Theorem. There exists a closed simply-connected symplectic manifold V , dim V = 8 such that b3(V ) = 1 and all the Massey products in V are trivial. Proof. Let (M, !) be a 4-dimensional symplectic manifold with b1(M) = 1. The existence of such manifolds follows from results of Gompf [G1]. Without loss of generality we can assume that the symplectic form on M is integral. We embed M in CP5 symplectically and denote by X the result of the blow up along this embedding. So, we have a commutative diagram fM ----! X ? ? q?y ?yp M ----! CP5 where q : fM ! M is a locally trivial bundle with the fiber CP2 and p : X \ fM ! CP5 \ M is a diffeomorphism. Let CP11be a projective line in CP5 which does not meet M. Then the subman- ifold p-1 (CP11) gives us a class a 2 H2(X). Similarly, the inclusion CP12 CP2 Mf X gives us a class b 2 H2(X). It is well known that {a, b} is a basis of H* *2(X), [MS, TO]. Let æ 2 H2(X) be the cohomology class of p*!0. It is clear that æ(a) 6= 0 and æ(b) = 0. It is well known that X possesses a symplectic form whose cohomology class [ ] is æ + varepsilonoe for varepsilon small enough. It is also easy to s* *ee that b3(X) = b1(M) = 1, see loc. cit. Let be the normal bundle of the inclusion i : fM X, and let U 2 H*(T ) * *be the Thom class of . Consider the Browder-Novikov collapsing map c : X ! T and set oe = c*U. It is easy to see that oe(a) = 0 and oe(b) 6= 0. Furthermor* *e, for every x 2 H*(X) we have (*) oex = c*(U(i*x)). Notice that in H*(X) we have æ5 6= 0 6= æ2oe2. Indeed, æ5 = p*([!0])5, and t* *he restriction of æ2oe2 on fM is (up to a non-zero multiplicative constant) the to* *p class ,q*! 2 H8(Mf ), where , restricts to a non-zero element of the fiber CP2. 2.7. Lemma. In X we have æ3oe = 0, æ4 \ [X] = k1a, æ2oe2 \ X = k2b, k1, k2 6= 0. Proof. The equality æ3oe = 0 follows from the equalities (*) and !3 = 0. Furthermore, æ \ (æ4 \ [X]) = æ5 \ [X] 6= 0, oe \ (æ4 \ [X] = (æ4oe) \ [X] = 0 and hence æ4 \ [X] = k1a, k1 6= 0. Finally, æ \ (æ2oe2 \ [X]) = (æ3oe2) \ [X] = 0 6 RA'UL IB'A~NEZ, YULI RUDYAK, ALEKSY TRALLE AND LUIS UGARTE and hence æ2oe2 \ [X] = k2b, k2 2 R. Thus, k2b 6= 0 since æ2oe2 6= 0. By the routine arguments, we can assume that is an integral form (by choos* *ing a suitable varepsilon). Because of the Donaldson Theorem [[]D, Theorem 1 and Proposition 39, there is a closed symplectic submanifold V of X of codimension* * 2 (i.e. dim V = 8) such that the homology class j*[V ] is dual to N[ ] for N la* *rge enough; here j : V ! X denotes the inclusion. In other words, j*[V ] = (~æ + ~o* *e) \ [X] for some ~, ~ 2 R. Furthermore, j* : ßi(V ) ! ßi(X) is an isomorphism for i 3 and an epimorphism for i = 4. So, according to the Hurewicz-Whitehead theorem, the homomorphism j* : Hi(V ) ! Hi(X) is an isomorphism for i 3 and an epimorphism for i = 4. In particular, b3(V ) = 1. We set u = j*æ and v = j*oe. 2.8. Lemma. The R-vector space H6(V ) has dimension 2 and is generated by u3 and u2v. Proof. We have H6(V ) = H2(V ) = H2(X) = R2. Furthermore, by Lemma 1, j*(u3 \ [V ])= æ3 \ (j*[V ]) = æ3 \ (~æ + ~oe) \ [X] = (~æ4 + ~æ3oe) \ [X] = ~æ4 \ [X] = ~0a, ~06= 0. Similarly, j*(u2v \ [V ])= æ2oe \ (j*[V ]) = æ2oe \ (~æ + ~oe) \ [X] = (~æ3oe + ~æ2oe2) * *\ [X] = ~æ2oe2 \ [X] = ~0b, ~06= 0. Thus, u3 and u2v are linearly independent over R. Now we prove that every Massey product in H*(V ) is trivial, i.* *e., it contains zero provided that it is defined. Notice that all the products u2, uv* * and v2 are non-zero because j* : H4(X) ! H4(V ) is a monomorphism. Case 1. H5(V ), i.e., ff| = |fi| = |fl| = 2. Then the Masse* *y product is not defined, since fffi 6= 0. Case 2. H6(V ). Clearly, if ff 6= 0 6= fl and * *is defined then |fi| = 3. Furthermore, if ff = u or fl = u then H6(V ) = (ff, fl), and so 0 2 <* *ff, fi, fl>. So, it remains to consider the case . But every Massey triple produc* *t of the form , |x| = 2, |y| = 3 contains zero. You can see it directly or * *use the equlity = - from [K, Theorem 8]. Case 3. H7(V ) = 0. Trivial. Case 4. H8(V ) = R. Then there exist z 2 H*(X) with ffz 6= 0,* * and so (ff, fl) = H8(V ). Thus, is trivial. Finally, all the higher (i.e, quadruple, etc.) Massey products are trivial f* *or the dimensional reasons. HOMOTOPY PROPERTIES OF SYMPLECTIC MANIFOLDS 7 3. Relation between homotopy properties of closed symplectic manifolds 3.1. Theorem. The relations between the Hard Lefschetz property, evenness of odd-degree Betti numbers and vanishing of the Massey products for closed symple* *ctic manifolds are given by the following tables: TABLE 1: symplectically aspherical case; TABLE 2: simply-connected case. The word Impossible in the table means that there is no closed symplectic manifold (aspherical or simply connected) that realizes the combination in the corresponding line. The sign ? means that we (the authors) do not know whether a manifold with corresponding properties exists. 8 RA'UL IB'A~NEZ, YULI RUDYAK, ALEKSY TRALLE AND LUIS UGARTE Table 1: Symplectically Aspherical Symplectic Manifolds ______________________________________________________________________________|* *|||| | Triviality of |Hard Lefschetz | Evenness of | | |_Massey_Products_____|___Property______|____b2i+1_____|______________________|* *||||| 2n ||_______yes__________||_____yes________||____yes______||___Kähler_(T__)______|| | | | | | ||_______yes__________||_____yes________||____no_______||____Impossible_______|| | | | | | ||_______yes__________||_____no_________||____yes______||_________?___________|| | | | | | ||_______yes__________||_____no_________||____no_______||_________?___________|| | | | | | ||_______no___________||_____yes________||____yes______||_________?___________|| | | | | | ||_______no___________||_____yes________||____no_______||____Impossible_______|| | | | | | ||_______no___________||_____no_________||____yes______||______K_x_K__________|| | | | | | ||_______no___________||_____no_________||____no_______||________K____________|| Table 2: Simply-Connected Symplectic Manifolds ______________________________________________________________________________|* *|||| | Triviality of |Hard Lefschetz |Evenness of | | |_Massey_Products_____|___Property_______|___b2i+1_____|______________________|* *||||| n ||_______yes__________||_____yes_________||___yes______||___Kähler_(CP_)______|| | | | | | ||_______yes__________||_____yes_________||___no_______||____Impossible_______|| | | | | | ||_______yes__________||_____no__________||___yes______||____M(6,{e},0)_______|| | | | | | ||_______yes__________||_____no__________||___no_______||________V____________|| | | | | | ||_______no___________||_____yes_________||___yes______||________?____________|| | | | | | ||_______no___________||_____yes_________||___no_______||____Impossible_______|| || || || || || ||_______no___________||_____no__________||___yes______||___CgP5_x_gCP_5______|| || || || || || ||_______no___________||_____no__________||___no_______||_______gCP5__________|| HOMOTOPY PROPERTIES OF SYMPLECTIC MANIFOLDS 9 Proof. We prove the theorem via line-by-line analysis of Tables 1 and 2. Line 1 in Tables 1 and 2. For closed Kähler manifolds, the Hard Lefschetz property is proved in [GH], the evenness of b2i+1 follows from the Hodge theory [W], the triviality of Massey products follows from the formality of any closed Kähler manifold [DGMS]. One can ask if there are non-Kähler manifolds having the properties from line 1. In the symplectically aspherical case the answer is affirmative. Let G = R* * xOE R2 be the semidirect product determined by the one-parameter subgroup OE(t) = diag (ekt, e-kt), t 2 R, ek + e-k 6= 2. One can check that G contains a lattice* *, say . Then the compact solvmanifold M = G= x S1 is symplectic and has the same minimal model as the Kähler manifold S2 x T 2. Hence such manifold fits into line 1. It cannot be Kähler, since it admits no c* *om- plex structure. The latter follows from the Kodaira-Yau classification of compa* *ct complex surfaces (see [TO] for details). Line 2 in Tables 1 and 2. Any manifold satisfying the Hard Lefschetz prop- erty must have even b2i+1. Indeed, consider the usual non-singular pairing p : H2k+1 (M) H2m-2k-1 (M) ! R of the form Z p ([ff], [fi])= ff ^ fi. M Define a skew-symmetric bilinear form <-, -> : H2k+1 (M) H2k+1 (M) ! R via the formula <[ff], [fl]> = p [ff], Lm-2k-1 [fl], for [ff], [fl] 2 H2k+1 (M). Since this form is non-degenerate and skew-symmetri* *c, its domain H2k+1 (M) must be even-dimensional, i.e. b2k+1 is even. Line 3 in Table 1. We do not know any non-simply-connected (and, in particular, symplectically aspherical) examples to fill in this line. Line 3 in Table 2. We use Theorem 2.2 with n = 6 and G = {e}. Then, for every b, the corresponding manifold M(6, {e}, b) has even odd-degree Betti numbers and does not have the Hard Lefschetz property. Furthermore, all the Massey products in M vanish by 2.1. Line 4 in Table 2. Consider the manifold V from Theorem 2.6. Since b3(V ) = 1, V does not have the Hard Lefschetz property. Line 5 in Tables 1 and 2 and Line 4 in Table 2. We do not know any examples to fill in these lines. Line 6 in Tables 1 and 2. This is impossible, see the argument concerning line 2. Lines 7 and 8 in Table 1. Consider the Kodaira-Thurston manifold K [Th]. Recall that this manifold is defined as a nilmanifold K = N3= x S1, 10 RA'UL IB'A~NEZ, YULI RUDYAK, ALEKSY TRALLE AND LUIS UGARTE where N3 denotes the 3-dimensional nilpotent Lie group of triangular unipotent matrices and denotes the lattice of such matrices with integer entries. One c* *an check that the Chevalley-Eilenberg complex of the Lie algebra n3 is of the form ( (e1, e2, e3), d), de1 = de2 = 0, de3 = e1e2. with |ei| = 1. We have already mentioned that the minimal model of any nilman- ifold N= is isomorphic to the Chevalley-Eilenberg complex of the Lie algebra n. In particular, one can get the minimal model of the Kodaira-Thurston manifold in the form ( (x, e1, e2, e3), d), dx = de1 = de2 = 0, de3 = e1e2 with degrees of all generators equal 1. One can check that the vector space H1(* *K) has the basis {[x], [e1], [e2]}. Hence, b1(K) = 3, which also shows that K does not have the Hard Lefschetz property. Furthermore, K possesses a symplectic form ! with [!] = [e1e3 + e2x], and one can prove that the Massey triple product <[e1], [e1], [!]> is non-trivial. Thus, K realizes Line 8 of Table 1. Finally, K x K realizes Line 7 of Table 1. Lines 7 and 8 in Table 2. We use Theorem 2.5. Consider a symplectic embedding i : K ! CP m, m 5, and perform the blow-up along i. Then, by 2.5(i), C]P m is simply-connected. Furthermore, it realizes Line 8 of Table 2 by 2.5(ii) and 2.5* *(iii). Finally, ]CP m x ]CP m realizes Line 7 of Table 2. 3.2. Remark. The result of Lupton [L] shows that the problem of constructing of a non-formal manifold with the Hard Lefschetz property turns our to be very delicate. In [L] there is an example of a DGA, whose cohomology has the Hard Lefschetz property, but which is not intrinsically formal. This means that the* *re is also a non-formal minimal algebra with the same cohomology ring. Sometimes, using Browder-Novikov theory, one can construct a smooth closed manifold M with such non-formal Sullivan minimal model. However, there is no way in sight to get a symplectic structure on M. 4. Flexible symplectic manifolds Let (M2m , !) be a symplectic manifold. It is known that there exists a uniq* *ue non-degenerate Poisson structure associated with the symplectic structure (se* *e, for example [LM, TO]). Recall that is a skew symmetric tensor field of order 2 such that [ , ] = 0, where [-, -] is the Schouten-Nijenhuis bracket. The Koszul differential ffi : k(M) ! k-1 (M) is defined for Poisson, in pa* *rtic- ular symplectic, manifolds as ffi = [i( ), d]. Brylinski has proved in [B] that the Koszul differential is a symplectic codiff* *erential of the exterior differential with respect to the symplectic star operator. We c* *hoose the volume form associated to the symplectic form, say vM = !m =m!. Then we define the symplectic star operator * : k(M) ! 2m-k (M) HOMOTOPY PROPERTIES OF SYMPLECTIC MANIFOLDS 11 by the condition fi ^ (*ff) = k( )(fi, ff)vM , for all ff, fi 2 k(M). It turn* *s out to be that ffi = (-1)k+1 (* O d O *). 4.1. Definition. A k-form ff on the symplectic manifold M is called symplectica* *lly harmonic, if dff = 0 = ffiff. We denote by khr(M) the space of symplectically harmonic k-forms on M. We set Hkhr(M, !) = khr(M)=(Im d \ khr(M)), hk(M) = hk(M, !) = dim Hkhr(M, !). We say that a de Rham cohomology class is symplectically harmonic if it contains a symplectically harmonic representative, i.e. if it belongs to the subgroup H** *hr(M) of H*(M). 4.2. Definition. We say that a closed smooth manifold M is flexible, if M pos- sesses a continuous family of symplectic forms !t, t 2 [a, b], such that hk(M, * *!a) 6= hk(M, !b) for some k. So, the McDuff-Khesin Question (see the introduction) asks about existence of flexible manifolds. In order to prove our result on the existence of flexible 6-dimensional nilm* *ani- folds, we need some preliminaries. The following lemma is proved in [IRTU1] and generalizes an observation of Yan [Y]. 4.3. Lemma. For any symplectic manifold (M2m , !) and k = 0, 1, 2 we have H2m-khr(M) = Im {Lm-k : Hk (M) ! H2m-k (M)} H2m-k (M). In other words, h2m-k (M, !) = dim Im {Lm-k : Hk (M) ! H2m-k (M)}. The following fact can be deduced from 4.3 using standard arguments from lin* *ear algebra, see [IRTU1]. 4.4. Lemma. Let !1 and !2 be two symplectic forms on a closed manifold M2m . Suppose that, for k = 1 or k = 2, we have h2m-k (M, !1) 6= h2m-k (M, !2). Then M is flexible. 4.5. Proposition. Let G be a simply connected 6-dimensional nilpotent Lie group such that its Lie algebra g has the basis {Xi}6i=1and the following structure r* *ela- tions: [X1, X2] = -X4, [X1, X4] = -X5, [X1, X5] = [X2, X3] = [X2, X4] = -X6 (all the other brackets [Xi, Xj] are assumed to be zero). Then G admits a latti* *ce , and the corresponding compact nilmanifold N := G= admits two symplectic forms !1 and !2 such that dim Im L2[!1]= 0, dim Im L2[!2]= 2. 12 RA'UL IB'A~NEZ, YULI RUDYAK, ALEKSY TRALLE AND LUIS UGARTE Proof. First, G has a lattice by 2.4(ii). Furthermore, by 2.4(iii), in the Chev* *alley- Eilenberg complex ( *g*, d) we have dff1= dff2 = dff3 = 0, dff4= ff1ff2, dff5= ff1ff4, dff6= ff1ff5 + ff2ff3 + ff2ff4, where we write ffiffj instead of ffi^ ffj. One can check that the following ele* *ments represent closed homogeneous 2-forms on N: !1 = ff1ff6 + ff2ff5 - ff3ff4, !2 = ff1ff3 + ff2ff6 - ff4ff5. Since [!31] 6= 0 6= [!32], these homogeneous forms are symplectic. Indeed, by 2* *.4(iii) the cohomology classes [!0] and [!1] have homogeneous representatives whose thi* *rd powers are non-zero. Then the same is valid for their pull-backs to invariant 2* *-forms on the Lie group G. But for invariant 2-forms this condition implies non-degene* *racy. Since G ! N is a covering, the homogeneous forms !1 and !2 on N are also non- degenerate. Obviously, the R-vector space H1(N) has the basis {[ff1], [ff2], [ff3]}. On* *e can check by direct calculation that [!1]2[ffi] = 0, i = 1, 2, 3 and that [!2]2[ff1] = -2[ff1ff2ff4ff5ff6], [!2]2[ff2] = 0, [!2]2[ff3] = 2[ff2ff3ff* *4ff5ff6]. Finally, it is straightforward that the above cohomology classes span 2-dimensi* *onal subspace in H5(N). 4.6. Theorem. There exists a flexible 6-dimensional nilmanifold. Proof. Consider the nilmanifold N as in 4.5. Because of 4.3 and 4.5, we conclude that h5(N, !1) = 0 6= 2 = h5(N, !2), and the result follows from 4.4. Acknowledgment. The first and the fourth authors were partially supported by the project UPV 127.310-EA147/98. This work was partially done in Oberwol- fach and financed by Volkswagen-Stiftung. The second and third authors were also partially supported by Max-Planck Institut für Mathematik, Bonn. REFERENCES [B] J.-L. Brylinski, A differential complex for Poisson manifolds, J. Diff.* * Geom. 28 (1988), 93-114. [DGMS] P. Deligne, P. Griffiths, J. Morgan and D. Sullivan, Real homotopy theo* *ry of Kähler manifolds, Invent. Math. 29 (1975), 245-274. HOMOTOPY PROPERTIES OF SYMPLECTIC MANIFOLDS 13 [D] S. Donaldson, Symplectic submanifolds and almost-complex geometry, J. D* *iff. Geom. 44 (1996), 666-705. [FG] M. Fern'andez and A. Gray, Compact symplectic solvmanifold not admittin* *g complex structures, Geom. Dedic. 34 (1990), 295-299. [F] A. Floer, Symplectic fixed points and holomorphic spheres, Commun. Math* *. Phys 120 (1989), 575-611. [G1] R. Gompf, A new construction of symplectic manifolds, Ann. Math. 142 (1* *995), 527-597. [G2] R. Gompf, On symplectically aspherical manifolds with nontrivial i2, Ma* *th. Res. Letters 5 (1999), 599-603. [GH] P. Griffiths and J. Harris, Principles of Algebraic Geometry, Wiley, Ne* *w York, 1978. [Gr] M. Gromov, Partial Differential Relations, Springer, Berlin, 1986. [H] H. Hofer, Lusternik-Schnirelmann theory for Lagrangian intersections, A* *nnales de l'inst. Henri Poincar'e- analyse nonlineare 5 (1988), 465-499. [IRTU1] R. Ib'a~nez, Yu. Rudyak, A. Tralle and L. Ugarte, On symplectically har* *monic forms on six-dimensional nilmanifolds, Comment. Math. Helv. 76, no. 1 (2001), 89* *-109; Erratum 76, no. 3 (2001), 576. [IRTU2] R. Ib'a~nez, Yu. Rudyak, A. Tralle and L. Ugarte, Symplectically Harmon* *ic Cohomology of Nilmanifolds, Available from xxx.lanl.gov archive as SG/0111074. [K] D. Kraines, Massey higher products, Trans. Amer. Math. Soc. 124 (1966),* * 431-449. [LM] P. Libermann and C. Marle, Symplectic Geometry and Analytical Mechanics* *, Kluwer, Dordrecht, 1987. [L] G. Lupton, Intrinsic formality and certain types of algebras, Trans. Am* *er. Math. Soc. 319 (1990), 257-283. [M] O. Mathieu, Harmonic cohomology classes of symplectic manifolds, Commen* *t. Math. Helvetici 70 (1995), 1-9. [Ma] J. P. May, Matric Massey products, J. Algebra 12 (1969), 533-568. [McD] D. McDuff, Examples of symplectic simply connected manifolds with no Kä* *hler structure, J. Diff. Geom. 20 (1984), 267-277. [MS] D. McDuff and D. Salamon, Introduction to Symplectic Topology, Clarendo* *n Press, Ox- ford, 1998. [Mi] T. J. Miller, On the formality of (k - 1)-connected compact manifolds o* *f dimension less than or equal to 4k - 2, Illinois J. Math. 23 (1979), 253-258. [RO] Yu. Rudyak and J. Oprea, On the Lusternik-Schnirelmann category of symp* *lectic man- ifolds and the Arnold conjecture, Math. Z. 230 (1999), 673-678. [RT1] Yu. Rudyak and A. Tralle, On Thom spaces, Massey products and non-forma* *l symplectic manifolds, available as DG9907035, MPI Preprint Series 71 (1999). [RT2] Yu. Rudyak and A. Tralle, On symplectic manifolds with aspherical sympl* *ectic form, available as DG9908001, MPI Preprint Series 88 (1999). [S] D. Sullivan, Infinitesimal computations in topology, Publ. Math. IHES 4* *7 (1978), 269- 331. [Th] W. P. Thurston, Some simple examples of compact symplectic manifolds, P* *roc. Amer. Math. Soc. 55 (1976), 467-468. [Ti] D. Tischler, Closed 2-forms and an embedding theorem for symplectic man* *ifolds, J. Diff. Geom. 12 (1977), 229-235. 14 RA'UL IB'A~NEZ, YULI RUDYAK, ALEKSY TRALLE AND LUIS UGARTE [TO] A. Tralle and J. Oprea, Symplectic manifolds with no Kähler structure, * *Springer, LNM 1661, 1997. [W] R.O. Wells, Differential Analysis on Complex Manifolds, Prentice Hall, * *1978. [Y] D. Yan, Hodge structure on symplectic manifolds, Adv. Math. 120 (1996),* * 143-154. Departamento de Matem'aticas, Facultad de Ciencias, Universidad del Pais Vas* *co, Apdo. 644, 48080 Bilbao, Spain, E-mail address: : mtpibtor@lg.ehu.es Yu. B. Rudyak, Department of Mathematics, University of Florida, 358 Little Hall, PO Box 118105 Gainesville, FL 32611-8105, USA, E-mail address: rudyak@math.ufl.edu, july@mathi.uni-heidelberg.de University of Warmia and Masuria, 10561 Olsztyn, Poland E-mail address: tralle@tufi.wsp.olsztyn.pl Departamento de Matem'aticas, Facultad de Ciencias, Universidad de Zaragoza, 50009 Zaragoza, Spain E-mail address: : ugarte@posta.unizar.es