March 10, 2005; MSC 2000 classification: primary 57R17, 57R50, 57S05; secondary 22E65, 53D35

<ph f="ptmb7t">Crossed flux homomorphisms and vanishing theorems for flux groups</ph>

J. Kȩ dra, D. Kotschick,

S. Morita

Institute of Mathematics, University of Szczecin, Wielkopolska 15, 70-451 Szczecin, Poland; and Mathematical Sciences, University of Aberdeen, Meston Building, King's College, Aberdeen AB24 3UE, Scotland, UK E-mail address: kedra@maths.abdn.ac.uk Mathematisches Institut, Ludwig-Maximilians-Universität München, Theresienstr. 39, 80333 München, Germany E-mail address: dieter@member.ams.org Department of Mathematical Sciences, University of Tokyo, Komaba, Tokyo 153-8914, Japan E-mail address: morita@ms.u-tokyo.ac.jp

1 Introduction

Let M   be a closed smooth manifold and α   a closed p   -form on M   . We shall denote by Diff α   the group of diffeomorphisms of M   which preserve α   , equipped with the C ∞   topology. Let Diff 0 α   be its identity component. The flux homomorphism associated to α   is defined on the universal covering of Diff 0 α   by the formula
Flux α : Diff ~ 0 α ⟶ H p − 1 ( M ; R ) (1)
φ t ↦ ∫ 0 1 [ i φ ˙ t α ] d t (2)
for any path φ t   in Diff 0 α   with φ 0 = I d M   . It is easy to see that the defining integral depends on the path only up to homotopy with fixed endpoints. Identifying an element of the fundamental group of Diff 0 α   with the corresponding homotopy class of paths based at the identity in the universal covering one obtains a homomorphism Flux α : π 1 ( Diff 0 α ) ⟶ H p − 1 ( M ; R )   whose image is the flux group Γ α   associated with α   . The flux homomorphism descends to a homomorphism defined on Diff 0 α   , also called the flux:
Flux α : Diff 0 α ⟶ H p − 1 ( M ; R ) / Γ α .   There are a number of general questions one can ask about this situation, such as whether the flux group Γ α   is trivial, or at least discrete, and whether the flux homomorphism can be extended from Diff 0 α   to the whole group Diff α   . As far as we know, these questions have only been considered in the literature in the case when α   is a symplectic form, see for example [3, 4, 8, 18, 20, 22, 23, 24, 26, 28] and the papers cited there.
It is the aim of this paper to discuss these questions in some generality. We shall show how certain arguments used in [20] for the case of symplectic forms, mostly on surfaces, can be adapted to the general case, showing that the flux groups are often trivial and the flux homomorphisms extend. In the case of volume forms we shall go further and prove that the non-triviality of the flux group implies that M   is very special indeed, in that it shares many properties with manifolds that admit circle actions with homologically essential orbits. The most striking result we shall prove in this direction is that the non-triviality of the volume flux group implies that the manifold has a finite cover with vanishing minimal volume entropy or asymptotic volume, see Section  4 .
We shall also consider symplectic forms and their powers, for which we obtain generalizations of some results previously proved in [26, 18, 20] . Unlike in the case of volume forms, where the flux group is always discrete for purely topological reasons, the discreteness of the symplectic flux group was an open problem until very recently. This issue, first raised by Banyaga, has just been resolved by Ono's proof [31] using methods of hard symplectic topology. Our results in Section  5 , like those of some of the references mentioned above, show that very often the symplectic flux group actually vanishes.

2 General properties of the flux and crossed flux homomorphisms

For cyclic isotopies representing elements of the fundamental group of Diff 0 α   we can reformulate the definition of the flux as follows.
Lemma 1. If φ t   is a closed loop representing an element of π 1 ( Diff 0 α )   and c   is a cycle representing a homology class in H p − 1 ( M , Z )   , then up to sign we have 〈 Flux α ( φ t ) , [ c ] 〉 = 〈 [ α ] , [ φ t ( c ) ] 〉 ,   where φ t ( c )   denotes the cycle swept out by the loop of diffeomorphisms φ t   applied to c   .
  • Proof. This is immediate from the definition of the flux; compare [26] . □
As every real cohomology class can be detected by mapping closed oriented manifolds Σ   into M   , this lemma shows that the flux group of α   is detected by the evaluation of the pullback of [ α ]   on products S 1 × Σ   . More precisely, define φ : S 1 × Σ → M   by φ ( t , x ) = φ t ( f ( x ) )   , where f : Σ → M   is a representative for (a multiple of) [ c ]   . Then 〈 Flux α ( φ t ) , [ Σ ] 〉 = 〈 φ * [ α ] , [ S 1 × Σ ] 〉 .   One should not be misled by this discussion into thinking that the flux group depends only on the cohomology class of [ α ]   , because which loops in Diff 0   can be deformed to essential loops in Diff 0 α   depends on α   itself, and not just on its cohomology class.
Remark 2. Let [ φ t ] ∈ π 1 ( Diff 0 α )   , and denote by E φ t → S 2   the M   -bundle corresponding to φ t   via the clutching construction. It is known [24, 33] that the flux homomorphism Flux α : π 1 ( Diff 0 α ) → H p − 1 ( M , R )   is equal to the evaluation on [ α ]   of the differential ∂ φ t * : H * ( M ; R ) ⟶ H * − 1 ( M ; R )   in the cohomology spectral sequence associated to E φ t   . More precisely, 〈 Flux α ( φ t ) , c 〉 = 〈 ∂ φ t * [ α ] , c 〉   for all c ∈ H p − 1 ( M )   .
The diffeomorphism group Diff α   acts by conjugation on itself and on the universal covering of its identity component. It also acts on cohomology, and this latter action factors through the mapping class group ℳ α   with respect to α   , defined to be the quotient group Diff α / Diff 0 α   . Our first observation is that the flux is equivariant with respect to these actions:
Lemma 3. The flux homomorphism Flux α : Diff ~ 0 α ⟶ H p − 1 ( M ; R )   is equivariant with respect to the natural actions of Diff α   . In other words, for any two elements ψ ∈ Diff α   and φ t ∈ Diff ~ 0 α   , we have the identity Flux ( ψ φ t ψ − 1 ) = ψ ¯ ( Flux ( φ t ) )   where ψ ¯ ∈ ℳ α   denotes the mapping class of ψ   and ℳ α   acts on H p − 1 ( M ; R )   from the left by the rule ψ ¯ ( w ) = ( ψ ¯ − 1 ) * ( w )   for w ∈ H p − 1 ( M ; R )   .
This follows immediatly from the definition of the flux and the chain rule.
The lemma suggests that one should not expect an extension of the flux to Diff α   to exist as a homomorphism, but rather as a crossed homomorphism with respect to this action of Diff α   on cohomology. Indeed, in certain situations we shall prove the existence of an extension as a crossed homomorphism.
Consider the extension 1 ⟶ Diff 0 α ⟶ Diff α ⟶ ℳ α ⟶ 1   and its associated exact sequence of cohomology groups of discrete groups:
0 ⟶ H 1 ( ℳ α ; H p − 1 ( M ; R ) / Γ α ) ⟶ H 1 ( Diff α ; H p − 1 ( M ; R ) / Γ α )
⟶ H 1 ( Diff 0 α ; H p − 1 ( M ; R ) / Γ α ) ℳ α ⟶ δ H 2 ( ℳ α ; H p − 1 ( M ; R ) / Γ α ) ⟶
Lemma  3 shows that we can think of the flux homomorphism as an element Flux α ∈ H 1 ( Diff 0 α ; H p − 1 ( M ; R ) / Γ α ) ℳ α .   Extending the flux to Diff α   as a crossed homomorphism is equivalent to the vanishing of δ ( Flux α )   in the above exact sequence. We now examine this issue in detail.
For a foliated M   -bundle E → B   whose total holonomy is contained in Diff α   we have a transverse invariant class a ∈ H p ( E ; R )   defined as follows. Pulling back α   from M   to B ~ × M   , we obtain an invariant form α ~   which descends to E = ( B ~ × M ) / π 1 ( B )   as a closed form. We denote its cohomology class by a   .
Lemma 4. Let I = [ 0 , 1 ]   . For any φ ∈ Diff 0 α   let π : M φ → S 1   be the foliated M   -bundle over S 1   with monodromy φ   . It is the quotient of M × I   by the equivalence relation ( p , 0 ) ∼ ( φ ( p ) , 1 )   . By assumption, there is an isotopy φ t ∈ Diff 0 α   such that φ 0 = Id   and φ 1 = φ   . Let f : M φ → M × S 1   be the induced diffeomorphism given by the correspondence M φ ∋ ( p , t ) ↦ ( φ t − 1 ( p ) , t ) ∈ M × S 1 .   Then the transverse invariant class a ∈ H p ( M φ ; R )   is equal to
[ α ] + Flux α ( φ t ) ⊗ ν ∈ H p ( M × S 1 ; R )
∼ = H p ( M ; R ) ⊕ ( H p − 1 ( M ; R ) ⊗ H 1 ( S 1 ; R ) )
under the above isomorphism, where ν ∈ H 1 ( S 1 ; R )   denotes the fundamental cohomology class of S 1   .
  • Proof. The horizontal foliation on M φ   is induced from the trivial foliation on M × I   . Hence the transverse invariant class a   is represented by the form p 1 * α   on M × I   , where p 1 : M × I → M   denotes the projection to the first factor. It is clear that the H p ( M ; R )   -component of a   is equal to [ α ]   , so that we only need to prove that for any ( p − 1 )   -cycle c ⊂ M   , the value of a   on the cycle f − 1 ( c × S 1 ) ⊂ M φ   is equal to Flux ( φ t ) ( [ c ] )   where [ c ] ∈ H p − 1 ( M ; Z )   denotes the homology class of c   . Now on M × I   , the above cycle is expressed as the image of the map c × I ∋ ( q , t ) ↦ ( φ t ( q ) , t ) ∈ M × I   because f − 1 ( q , t ) = ( φ t ( q ) , t ) ( ( q , t ) ∈ M × S 1 )   . Hence the required value is equal to the integral of α   over the image of the map c × I ∋ ( q , t ) ↦ φ t ( q ) ∈ M .   But this is exactly equal to the value of Flux ( φ t )   on the homology class represented by the cycle c ⊂ M   . This completes the proof. □
For the formulation of our result about extensions of the flux homomorphism as a crossed homomorphism we use the following definition:
Definition 5. Let M   be a closed manifold and G ⊂ Diff ( M )   a subgroup. We say that a cohomology class c ∈ H * ( M )   extends G   -universally if there is a class b   on the total space of any M   -bundle with structure group G   restricting to c   on the fibers.
Theorem 6. Let G ⊂ Diff α   be a subgroup. If [ α ]   extends G   -universally, then the flux homomorphism Flux α : G 0 ~ ⟶ H p − 1 ( M ; R )   vanishes on π 1 ( G 0 )   and extends to a crossed homomorphism Flux α ~ : G ⟶ H p − 1 ( M ; R ) .  
  • Proof. By assumption [ α ]   extends G   -universally, so in particular it extends to the M   -bundle over S 2   given by the clutching construction for a loop in G 0   . Therefore, Flux α   vanishes on π 1 ( G 0 )   , cf. Remark  2 .
    Let B G δ   be the classifying space of G   considered as a discrete group, and denote by π : E G δ ⟶ B G δ   the universal foliated M   -bundle over B G δ   with total holonomy group in G   . Let b ∈ H p ( E G δ ; R )   be a universal extension of [ α ]   , and consider the difference u = a − b ∈ H p ( E G δ ; R ) ,   where a   denotes the transverse invariant class represented by the global p   -form α ~   on E G δ   which restricts to α   on each fiber. The restriction of u   to the fiber vanishes, so that, in the spectral sequence { E r p , q }   for the real cohomology, we have the natural projection
    P : Ker ( H p ( E G δ ; R ) → H p ( M ; R ) ) ∋ u
    ⟶ P ( u ) ∈ E ∞ 1 , p − 1 ⊂ E 2 1 , p − 1 = H 1 ( B G δ ; H p − 1 ( M ; R ) ) .
    Now Lemma  4 implies that the restriction of P ( u )   to the identity component of G   coincides with the flux homomorphism: P ( u ) = Flux α : G 0 ⟶ H p − 1 ( M ; R ) .   Thus we see that P ( u )   defines an extension of the flux homomorphism as a cohomology class whose representing cocycles are crossed homomorphisms. □
A particular case where the class [ α ]   does extend universally is when it represents some characteristic class for M   :
Corollary 7. Suppose [ α ] ∈ H p ( M ; R )   is a non-zero multiple of a polynomial in the Euler and Pontryagin classes of M   . Then the flux group Γ α   vanishes and the flux homomorphism extends as a crossed homomorphism Flux α ~ : Diff α ⟶ H p − 1 ( M ; R ) .  
  • Proof. For any M   -bundle E → B   consider the tangent bundle along the fibers. Its characteristic classes extend the characteristic classes of T M   from the fiber to the total space. □
If α   defines a geometric structure on M   , then Diff α   acts by automorphisms of this structure and preserves its characteristic classes. We shall consider the case of a symplectic structure in Section  5 below. Another instance of this is the case of foliations:
Example 8. Suppose α   is of constant rank, and T ℱ ⊂ T M   is its kernel. Then Diff α   preserves T ℱ   , and its characteristic classes extend Diff α   -universally. Therefore, the flux group vanishes and the flux homomorphism extends as a crossed homomorphism if [ α ]   is a non-zero multiple of a polynomial in the Euler and Pontryagin classes of T ℱ   and of T M / T ℱ   .
In general one cannot expect that the extension of the flux homomorphism to a crossed homomorphism is unique (if it exists at all). However, it was proved in [20] that for the case of a symplectic form, equivalently an area form, on a surface of genus ≥ 2   , the extension is unique.
There is another mechanism which can force the vanishing of the flux group Γ α   , stemming from Gromov's notion of bounded cohomology [15] . As usual, we say that a real cohomology class [ α ] ∈ H p ( M ; R )   is a bounded class, if it has a representative which is bounded as a functional on the set of singular simplices. This means that the class in the image of the comparison map from bounded to usual cohomology:
H b p ( M ; R ) ⟶ H p ( M ; R ) .  
Theorem 9. Suppose the closed p   -form α   represents a bounded cohomology class. Then the flux group Γ α   vanishes.
  • Proof. Suppose the flux group Γ α   does not vanish. Then according to Lemma  1 there is a smooth map φ : S 1 × Σ p − 1 → M   for which φ * α   has non-zero integral over S 1 × Σ   . As [ α ]   is assumed to be a bounded class, so is φ * [ α ]   . This means that the cohomology generator in top degree on S 1 × Σ   is bounded, equivalently the simplicial volume of S 1 × Σ   is non-zero. This is clearly false, as S 1 × Σ   maps to itself with arbitrarily large degree. □
In this case we obtain the vanishing of the flux group although we do not have a crossed homomorphism extending the flux homomorphism.

3 Volume-preserving diffeomorphisms

In this section we consider the flux of α   when α = μ   is a volume form on M   . In this case Moser's celebrated result [30] implies that Diff μ   is weakly homotopy equivalent to the full diffeomorphism group Diff ( M )   of M   . Moreover, applying Lemma  1 to a homotopy of volume forms of equal total volume, we deduce that, up to normalization, Γ μ   is independent of μ   . Furthermore, in this situation the mapping class group ℳ μ   does not depend on μ   , as it equals Diff ( M ) / Diff 0 ( M )   .
To see some examples and get a feel for what results to expect, we first consider loops of diffeomorphisms generated by circle actions.

3.1 Circle actions

Suppose we are given an effective circle action on an oriented closed manifold M   . By averaging we can always construct an invariant volume form μ   , so that we have a non-trivial homomorphism S 1 → Diff 0 μ   . We can easily characterize the non-triviality of the flux on the image of this homomorphism:
Proposition 10. A circle action gives rise to a nonzero element in Γ μ   if and only if its orbits are nonzero in real homology.
  • Proof. Let X   be the vector field generating the S 1   -action. Then L X μ = 0   , and i X μ   is a closed S 1   -invariant ( n − 1 )   -form representing the volume flux evaluated on the image of π 1 ( S 1 )   . This form is also a defining form for the (singular) foliation defined by the orbits.
    Choose closed S 1   -invariant 1   -forms α 1 , … , α k   representing a basis for the first de Rham cohomology.
    Then the wedge products α i ∧ i X μ   are S 1   -invariant as well, and their cohomology classes vanish for all i ∈ { 1 , … , k }   if and only if the flux vanishes in cohomology. But α i ∧ i X μ   is constant along each orbit, and vanishes if and only if α i ( X )   vanishes along the orbit. As all the orbits are homologous to each other and fill out the manifold, the flux can only vanish if α i ( X )   vanishes identically for all i   , which is equivalent to the orbits being null-homologous. (Note that it is not possible that α i ∧ i X μ   is exact but not identically zero.) □
If the orbits are non-trivial in homology, then the action has no fixed points. An integral α i   evaluating non-trivially on an orbit has α i ( X ) ≠ 0   everywhere, and therefore defines a smooth fibration over S 1   with fibers transverse to the circle action. The finiteness of the isotropy groups of the circle action implies finiteness of the monodromy of the fibration over S 1   . Thus our manifold has a finite cover which splits off S 1   smoothly and equivariantly, with the standard circle action.
This gives a differential-geometric proof of the Conner–Raymond theorem [11] , originally proved by topological means.
Given any fixed-point-free circle action on M   , Gromov [15] showed that the minimal volume of M   vanishes. A fortiori, the simplicial volume and the real characteristic numbers of M   vanish.
Specializing the general theorems of the previous section to volume forms allows us to extend these vanishing results from circle actions to non-trivial volume flux groups. Further, it is well-known that the orbits of circle actions represent central elements of the fundamental group acting trivially on homotopy groups, see for example Browder–Hsiang [7] or Appendix 2 in [15] . We shall generalize these statements in Theorem  11 .
In the case of 3   -manifolds there are more precise results. Namely, it was proved by Epstein that any fixed-point-free circle action on a closed 3   -manifold occurs by rotating the fibers of a Seifert fibration. Moreover, if the orbits are non-trivial in real homology, it is easy to see that the Euler class of the fibration is trivial, so that M   is finitely covered by a product of a surface with the circle.
We shall show in Theorem  14 below that these circle fibrations account for all non-trivial volume flux elements on 3   -manifolds.

3.2 Topological consequences of non-vanishing volume flux groups

We begin with a characterization of the non-triviality of the volume flux group, together with some homotopical constraints.
Theorem 11. Let M   be any closed n   -manifold with volume form μ   .
  • (1) The volume flux group Γ μ   is trivial if and only if the evaluation map e v : Diff 0 μ → M   induces the trivial map on the first real homology.
  • (2) If Γ μ ≠ 0   , then e v * : π 1 ( Diff 0 μ ) → π 1 ( M )   has an infinite image, which acts trivially on the homotopy groups of M   . In particular, the center of π 1 ( M )   is infinite.
  • Proof. Suppose φ t ∈ π 1 ( Diff 0 μ )   and denote by M → E φ t → S 2   the bundle associated to φ t   by the clutching construction. According to Remark  2 , the non-vanishing of Flux μ ( φ t )   is equivalent to the non-vanishing of ∂ φ t * ( μ )   . Applying Poincaré duality in E φ t   , this in turn is equivalent to the non-triviality of the differential ∂ φ t : H 0 ( M ) → H 1 ( M )   . On the other hand, ∂ φ t [ p t ] = e v * ( φ t )   , where e v * : H 1 ( Diff 0 μ ) → H 1 ( M )   is the map induced by the evaluation at the point p t ∈ M   . This proves the first claim.
    Considering the evaluation on the fundamental group, we conclude from what we proved above that it has infinite image. It is a general property of the image of the evaluation that it acts trivially on all homotopy groups. This was first noticed by Gottlieb [13] , compare also Theorem 2.2 in [33] .
    □
In the case of volume forms, Corollary  7 gives the following:
Corollary 12. Let M   be a closed oriented manifold of dimension 2 n   , and μ   a volume form on M   . If M   has a nonzero real characteristic number, then the flux group Γ μ   is trivial, and the flux homomorphism Flux μ   extends to a crossed homomorphism Flux ~ μ : Diff μ ⟶ H 2 n − 1 ( M ; R ) .  
It was proved in Theorem 2 of [20] that the cohomology class of the extension Flux ~ μ   is uniquely determined in H 1 ( Diff μ ; H 1 ( M ; R ) )   if μ   is a volume form on a surface of genus g ≥ 2   . When M   has dimension at least 4   , and has two different nonzero characteristic numbers, for example the Euler characteristic and the signature, then it may happen that these two choices give rise to different extensions of the flux associated with a volume form. In cohomology, the difference between any two such extensions is in H 1 ( ℳ M ; H 2 n − 1 ( M ; R ) )   , where ℳ M = Diff ( M ) / Diff 0 ( M )   is the mapping class group of M   .
Corollary  12 only applies to even-dimensional manifolds. In all dimensions, we have the following special case of Theorem  9 :
Corollary 13. Let M   be a closed oriented manifold with nonzero simplicial volume. Then the volume flux group Γ μ   is trivial for every volume form μ   .

3.3 The case of 3   -manifolds

In small dimensions we can now prove that a non-trivial volume flux group implies the vanishing of the minimal volume. Furthermore, we show that in small dimensions circle actions account for all non-trivial elements of volume flux groups on closed manifolds.
In dimensions < 3   the only closed oriented manifolds with non-trivial volume flux group are S 1   and T 2   , for which the above statements are trivial. In the first interesting case, for three-manifolds, we have a result modulo the Poincaré conjecture:
Theorem 14. Let M   be a closed oriented 3   -manifold without any fake cells. If M   has nonzero volume flux group Γ μ   , then M   is a Seifert fiber space. In particular, its minimal volume vanishes.
More precisely, a multiple of every loop φ t ∈ π 1 ( Diff μ )   with nonzero volume flux is realized by a fixed-point-free circle action on M   .
  • Proof. As π 1 ( M )   has non-trivial center by Theorem  11 , it is indecomposable as a free product.
    Therefore, in the Kneser–Milnor prime decomposition of M   , all summands but one are simply connected. As M   contains no fake cells by assumption, we conclude that it is prime. Thus, either M   is S 1 × S 2   , which is a Seifert fibration in the obvious way, or it is irreducible.
    If M   is irreducible, then because its first Betti number is positive, it is also sufficiently large1 , meaning that M   contains an incompressible surface, cf. [17] p. 35. Now it is a theorem of Waldhausen [38] that a closed irreducible sufficiently large 3   -manifold M   such that π 1 ( M )   has non-trivial center is Seifert fibered.
    By shrinking an invariant metric in the direction of the circle action, one sees that the minimal volume vanishes, cf. [15] .
    The statement to the effect that circle actions account for all nonzero elements of the volume flux group follows from the fact that on a Seifert manifold every element of the center of the fundamental group is, up to a multiple, represented by the fiber of a Seifert fibering, cf. [17] p. 92/93 or [36] . □
The Seifert fibered 3   -manifolds occuring in the theorem are precisely those with a Thurston geometry of type R 3   , S 2 × R   and H 2 × R   , compare [36] .

3.4 Back to higher dimensions

The proof of Theorem  14 has a partial generalization to higher dimensions:
Theorem 15. Let M   be a closed n   -manifold with Γ μ ≠ 0   . If M   is homotopy equivalent to a connected sum M 1 # M 2   then one of the M i   is a homotopy sphere.
  • Proof. As π 1 ( M )   has non-trivial center by Theorem  11 , it is indecomposable as a free product.
    Therefore, one of the M i   , say M 1   , is simply connected. In particular, H 1 ( M 1 ; Z ) = 0   . If M 1   is not a homotopy sphere, then there is a smallest k ≤ n − 2   for which H k ( M 1 ; Z )   does not vanish. By the Hurewicz theorem π k ( M 1 ) ∼ = H k ( M 1 ; Z ) ≠ 0   .
    Now the universal cover of M   is obtained from the universal cover of M 2   by connected summing with infinitely many copies of M 1   . Every non-trivial element of π 1 ( M ) ∼ = π 1 ( M 2 )   acts non-trivially on H k ( M ~ )   by permuting the different summands coming from the different copies of M 1   . This shows that every element of the fundamental group acts non-trivially on π k ( M )   , contradicting the second part of Theorem  11 . □
The following example shows that the result is sharp.
Example 16. Let M = ( S 1 × S 6 ) # Σ 7   , with Σ   a homotopy 7   -sphere. Now every Σ   is a twisted sphere, i. e. it is of the form D 7 ∪ ψ D 7   for some ψ   in the mapping class group of S 6 = ∂ D 7   .
But then M   is just the mapping torus of ψ   , and as the mapping class group of S 6   is finite (of order 28   ), we conclude that M   fibers over S 1   with finite monodromy ( = ψ   ). There is a fixed-point-free circle action transverse to this fibration generating a non-trivial volume flux group by Proposition  10 .
Note that the generator of π 1 ( M ) ∼ = Z   cannot always be realized by the orbits of a circle action on M   , so that passing to multiples is unavoidable. Indeed, if an S 1   -action on M   surjects π 1 ( S 1 )   onto π 1 ( M )   , then all the orbits have trivial stabilizer, because their homotopy classes are primitive. Then the quotient map M → M / S 1   is a smooth circle bundle over a homotopy 6   -sphere. This bundle is trivial, and as there are no exotic 6   -spheres we conclude that M   is diffeomorphic to S 1 × S 6   . But according to Browder [6] , Corollary 2.8, we can choose Σ   in such a way that M   is not diffeomorphic to S 1 × S 6   .
This example shows that the topological manifold S 1 × S 6   has several distinct smooth structures, all of which admit fixed-point-free circle actions. There are also known pairs of homeomorphic manifolds for which one has a free smooth circle action, and the other one has no smooth fixed-point-free actions at all, see [19] . Unfortunately, the examples of [19] are not really relevant to the discussion of volume flux groups, because their fundamental groups have trivial center.

4 Entropy and volume flux

We have seen that a non-trivial volume flux group forces the vanishing of the simplicial volume, and the vanishing of all real characteristic numbers. In the case when the volume flux comes from a smooth circle action, these vanishing results are consequences of the vanishing of the minimal volume. One might therefore speculate that the non-vanishing of the volume flux may imply the vanishing of the minimal volume. As we are not able to prove this, we shall consider intermediate invariants, which interpolate between the simplicial volume and the minimal volume.
Let M   be a connected closed oriented manifold of dimension n   . The following remarkable chain of inequalities between its topological invariants is known, compare [5, 19, 34] :
n n / 2 n ! | | M | | ≤ 2 n n n / 2 T ( M ) ≤ λ ( M ) n ≤ h ( M ) n ≤ ( n − 1 ) n MinVol ( M ) . (3)
Here T ( M )   is the spherical volume introduced by Besson, Courtois and Gallot [5] , λ ( M )   is the minimal volume entropy or asymptotic volume, h ( M )   is the minimal topological entropy of geodesic flows on M   , and MinVol   is the minimal volume in the sense of Gromov [15] . Babenko proved that the minimal volume entropy λ ( M )   is invariant under homotopy equivalences, and also under certain bordisms over B π 1 ( M )   , see [1, 2] . By contrast, the vanishing of the minimal volume is known to depend subtly on the smooth structure [19] .
Of the above invariants, only the simplicial volume is known to be multiplicative in coverings.
As it will be convenient to allow ourselves passage to finite coverings, we make the following definition in the spirit of [29] , compare also [5] .
Definition 17. Let I   be an invariant of n   -dimensional closed manifolds. Then define I * ( M ) = inf { I ( N ) d | N a d − sheeted covering of M } ,   where the infimum is taken over all finite coverings of M   .
Clearly I * ( M ) ≤ I ( M )   , and if I ( M ) ≤ J ( M )   for all M   , then I * ( M ) ≤ J * ( M )   . Thus,  3 implies
n n / 2 n ! | | M | | ≤ 2 n n n / 2 T * ( M ) ≤ λ * ( M ) n ≤ h * ( M ) n ≤ ( n − 1 ) n MinVol * ( M ) . (4)
We now prove the following vanishing result regarding the quantities in  4 :
Theorem 18. Let M   be a closed oriented manifold with non-vanishing volume flux group Γ μ   . Then M   has a finite covering M ¯   whose volume entropy λ ( M ¯ )   vanishes. In particular, λ * ( M ) = 0   .
  • Proof. If the volume flux group Γ μ   for ( M , μ )   is non-trivial, then by Theorem  11 the evaluation at a point of the corresponding loop φ t   in Diff μ   gives us a loop which is of infinite order in the center of π 1 ( M )   and in H 1 ( M ; Z )   . After replacing M   by a finite cover, we may assume that this element is primitive in H 1 ( M ; Z ) / t o r   , so that the fundamental group of M   splits as a direct product π 1 ( M ) ∼ = Z × π   with the generator of the first factor corresponding to the evaluation of our loop of diffeomorphisms, compare [32] . In this situation, Gottlieb [14] and independently Oprea [32] proved a homotopical analogue of the Conner–Raymond splitting theorem, showing that M   is homotopy equivalent to S 1 × Y   , where Y   is the homotopy fiber of a map f : M → S 1   inducing the projection p 1 : π 1 ( M ) → Z   onto the first factor of the fundamental group. If the homotopy type Y   can be represented by a closed oriented ( n − 1 )   -manifold, then we conclude the proof of Theorem  18 by noting that λ ( S 1 × Y )   vanishes because of the obvious circle action, and the minimal entropy is known to be homotopy-invariant by a result of Babenko [1] .
    Regardless what the homotopy fiber Y   is, we can proceed as follows. Choose a smooth map f : M → S 1   with f * = p 1   and let F ′   be a regular fiber. Then π 1 ( F ′ )   surjects onto π ∼ = Ker f *   , and we can modify F ′   by ambient surgery2 inside M   to obtain an embedded submanifold F ⊂ M   in the same homology class, such that π 1 ( F ) ∼ = π   , and the inclusion induces an isomorphism between π 1 ( F )   and 0 × π ⊂ Z × π ∼ = π 1 ( M )   . Consider then the map Φ   given by the composition
    S 1 × F ⟶ i S 1 × M ⟶ I d × φ S 1 × M ⟶ π 2 M (5)
    ( t , x ) ↦ ( t , x ) ↦ ( t , φ t ( x ) ) ↦ φ t ( x ) , (6)
    where the first map is the inclusion, and the composition of the second and third maps is the evaluation.
    Lemma 19. The map Φ   has the following properties:
    • (1) It induces an isomorphism on fundamental groups.
    • (2) It has degree one.
    • (3) It pulls back the tangent bundle of M   to the tangent bundle of S 1 × F   .
    • Proof. The first claim is clear from the construction of F   and Φ   . The second claim follows from the fact that F ⊂ M   has algebraic intersection number = 1   with the evaluation loop.
      For the third claim, consider the factorization Φ = π 2 ∘ ( I d × φ ) ∘ i   . The diffeomorphism I d × φ   pulls back π 2 * T M   to itself. But this bundle restricts to the image of i   as R ⊕ T F   , which proves the claim. □
    Let c : M → B π 1 ( M )   be the classifying map for the universal cover of M   , and consider the classes [ M , c ]   and [ S 1 × F , c ∘ Φ ]   in the bordism group Ω n ( B π 1 ( M ) )   . If these bordism classes agree, then there is a bordism [ W ′ , α ′ ]   between them. It follows that α * ′ : π 1 ( W ′ ) → π 1 ( M )   is surjective, and we can modify the bordism by surgery in the interior of W ′   so as to obtain a new bordism [ W , α ]   for which α * : π 1 ( W ) → π 1 ( M )   is an isomorphism. This new bordism has the property that the inclusion of each boundary component into W   induces an isomorphism on fundamental groups. In the terminology of Babenko [2] , this is an R   -cobordism, and the result of [2] is that R   -cobordant manifolds have the same minimal volume entropy. As before, the volume entropy of S 1 × F   vanishes because its minimal volume vanishes courtesy of the circle action.
    It remains to prove that the bordism classes [ M , c ]   and [ S 1 × F , c ∘ Φ ]   agree—or to deal with their failure to do so. Consider first the case when the integral homology of B π 1 ( M )   has no odd-order torsion. Then the bordism spectral sequence for B π 1 ( M )   is trivial, and we have an isomorphism
    Ω n ( B π 1 ( M ) ) ∼ = n ⊕ i = 0 H i ( B π 1 ( M ) ; Ω n − i ( ◆ ) ) , (7)
    compare Theorem 15.2 in [10] , or [37] . The elements of the summands on the right-hand side are detected by the collection of all Pontryagin and Stiefel-Whitney numbers twisted by cohomology classes on B π 1 ( M )   . Lemma  19 shows that these twisted characteristic numbers agree for [ M , c ]   and [ S 1 × F , c ∘ Φ ]   , so these two bordism classes in Ω n ( B π 1 ( M ) )   agree.
    In the general case, when the homology of B π 1 ( M )   is allowed to have odd-order torsion, we can still find the required bordism between [ M , c ]   and [ S 1 × F , c ∘ Φ ]   after passing to a suitable finite cover induced from a finite cover of S 1   . For this we only have to prove that on such a cover the map Φ : S 1 × F ⟶ M   is bordant to the identity of M   .
    Proposition 20. In the above situation M   has a finite covering induced from a finite covering of S 1   via the map M ⟶ S 1   , such that the lift of Φ   to the corresponding covering of S 1 × F   by itself is bordant to the identity of the target.
    • Proof. We shall use the language of surgery theory. Our reference for this is [25] .
      First of all, we already know from Lemma  19 that Φ   is a tangential equivalence, so that it is covered by a bundle isomorphism Φ ~ : T ( S 1 × F ) ⟶ T M   . Taking the stable normal bundles instead of the tangent bundles, we obtain a normal map (denoted by the same symbol) Φ ~ : N ( S 1 × F ) ⟶ N ( M )   covering Φ   , where N   denotes the stable normal bundle. Consider the surgery map σ : N M O ( M ) ⟶ [ M , G / O ] ,   where the domain consists of normal bordism classes with target M   . This map is an ismorphism, cf. Theorem 2.23 in [25] . Let α ∈ N M O ( M )   be the element represented by the normal map ( Φ ~ , Φ )   .
      Our task is to show that, after passing to a suitable finite cover, the image σ ( α )   is the homotopy class in [ M , G / O ]   represented by the constant map.
      Passing to finite coverings along S 1   means that we consider Φ k : S 1 × F ⟶ M k   , where the domain is the product of F   with the standard k   -fold cover of the circle, and M k   is the corresponding covering of M   pulled back from the circle. Since Φ k   has degree 1   , we see that Φ k * : H i ( M k ) ⟶ H i ( S 1 × F )   is injective with any coefficients. Moreover, by restriction we have an injection Φ k * : H i ( M k , F ) ⟶ H i ( S 1 × F , p t . × F ) ∼ = ι ⊗ H i − 1 ( F ) ,   where ι   denotes the generator of H 1 ( S 1 )   and F   sits inside M k   in the obvious way for any k   .
      Now Φ   is a tangential equivalence, and therefore a map M ⟶ G / O   representing σ ( α )   can be lifted to g : M ⟶ G   . Moreover, this map is constant on F ⊂ M   because Φ   is the identity on F   . It sufficesto show that g   is homotopic to the constant map, at least after we pass to a suitable finite covering.
      The obstructions to g   being null-homotopic are contained in H i ( M , F ; π i ( G ) ) ⊂ ι ⊗ H i − 1 ( F ; π i ( G ) ) .   But the group π i ( G )   is finite for every i   , and if we pass to the finite covering M k   , then the classifying map of the new surgery problem is just the composition M k ⟶ M ⟶ g G ,   as can be seen by inspecting the construction of the surgery map, cf. pp. 42-43 of [25] . Therefore we can kill the obstructions by taking suitable finite covers along S 1   . In detail, the coverings send ι   to k ι   , and since the obstructions lie in ι ⊗ H i − 1 ( F )   with finite coefficients, we can kill all the obstructions. This means that the normal map S 1 × F ⟶ M k   is normally bordant to the identity of M k   . □
    The proof of Proposition  20 shows that S 1 × F   and M k   are bordant over B π 1 ( M k )   . Together with the preceding discussion, this completes the proof of Theorem  18 . □
Remark 21. An alternative approach to the general case proceeds by observing that  7 always holds after tensoring with Q   , cf. [10, 37] . As [ M , c ]   and [ S 1 × F , c ∘ Φ ]   have the same twisted Pontryagin numbers, their difference is rationally zero-bordant. This means that for some k > 0   there is a bordism between the k   -fold connected sums k M   and k ( S 1 × F )   endowed with the corresponding maps to B π 1 ( M )   . These maps induce the diagonal map D : π 1 ( M ) ◆ … ◆ π 1 ( M ) ⟶ π 1 ( M )   on fundamental groups. Unfortunately it is unclear whether this can be arranged to be an R   -cobordism in the sense of Babenko [2] . If this is possible, then a slightly different proof of Theorem  18 can be given as follows. Babenko's theorem [2] implies that the minimal asymptotic exponential volume growth rates of the covers of k M   and of k ( S 1 × F )   with fundamental groups K e r ( D )   are equal.
(These are not the minimal volume entropies, because these covers are not the universal covers.) Now by a result of Paternain and Petean [34] , Theorem 5.9, the circle action on S 1 × F   gives rise to a T   -structure on the connected sum k ( S 1 × F )   . Another result of the same authors, Theorem A in [34] , shows that the minimal topological entropy h   vanishes for any manifold with a T   -structure.
A fortiori, the minimal volume entropy of k ( S 1 × F )   vanishes, compare  3 . This implies that the intermediate cover of k ( S 1 × F )   with fundamental group K e r ( D )   also has slow volume growth.
By the above discussion we have this conclusion also for the cover of k M   with fundamental group K e r ( D )   . This cover essentially contains a copy of the universal cover of M   , which therefore has small minimal asymptotic exponential volume growth rate. Thus, λ ( M ) = 0   .

5 Powers of a symplectic form

In this section we consider the case when α   is a power ω k   of a symplectic form ω   on M   , with M   of dimension 2 n   . It is clear that Diff α   contains the symplectomorphism group Symp = Diff ω   , but is usually strictly larger when k > 1   . In order to obtain a result parallel to Corollary  12 , we want to use the Chern classes of the tangent bundle along the fibers in the universal foliated M   -bundle. This means that instead of Diff α   we should only consider a smaller group which preserves the homotopy class of an almost complex structure compatible with ω   . We will simply take the symplectomorphism group and consider the k   -flux Flux k : Symp ~ 0 ⟶ H 2 k − 1 ( M ; R ) ,   which is the restriction of the flux with respect to ω k   to the symplectomorphism group. We denote by Γ k   the image of π 1 ( Symp 0 )   under the k   -flux.
The groups Γ k   for different values of k   are related to each other by the equation
Flux k ( φ t ) = k Flux 1 ( φ t ) ∧ ω k − 1 , (8)
which is immediate from the definition of the flux and the identity i X ( ω k ) = k i X ω ∧ ω k − 1   . Thus Γ k   is the image of of the usual symplectic flux group Γ 1 = Γ ω   under multiplication by k ω k − 1   .
Taking k = n   , we can use this to draw consequences about the symplectic flux group from our results about the volume flux. Note that Γ n   is not really the volume flux group, because we are only considering Symp   , and not the usually larger Diff ω n   . Nevertheless, the same arguments apply to prove the following:
Theorem 22. Let ( M , ω )   be a closed symplectic manifold of dimension 2 n   that satisfies one of the following conditions:
  • (1) the evaluation map e v : Symp 0 → M   induces the trivial map on the first real homology, or
  • (2) the fundamental group π 1 ( M )   has finite center, or
  • (3) M   has a nonzero real characteristic number, or has nonzero renormalized minimal volume entropy λ * ( M )   , or
  • (4) M   is homotopy equivalent to a connected sum in which neither summand is a homotopy sphere.
Then the symplectic flux group Γ ω ⊂ H 1 ( M ; R )   is in the kernel of the multiplication map [ ω ] n − 1 : H 1 ( M ; R ) → H 2 n − 1 ( M ; R ) .   In particular, if ( M , ω )   also satisfies the hard Lefschetz condition, then the symplectic flux group vanishes.
Instead of taking the maximal power of the symplectic form, we can consider smaller powers, and we can also use the Chern classes of an almost complex structure compatible with the symplectic form. Theorem  6 has the following immediate consequence:
Corollary 23. Suppose that [ ω k ] ∈ H 2 k ( M ; R )   is proportional to a polynomial in the Chern classes of ( M , ω )   . Then Γ k = 0   , and Flux k   extends to a crossed homomorphism Flux ~ k : Symp ⟶ H 2 k − 1 ( M ; R ) .  
This result has an antecedent in McDuff's paper [26] . The case k = 1   was proved in [20] .
As before, using  8 we obtain the following consequence. The case when k = n   was previously proved in [18] .
Corollary 24. Suppose that [ ω k ] ∈ H 2 k ( M ; R )   is proportional to a polynomial in the Chern classes of ( M , ω )   . Then the symplectic flux group Γ 1 = Γ ω ⊂ H 1 ( M ; R )   is in the kernel of the multiplication map [ ω ] k − 1 : H 1 ( M ; R ) ⟶ H 2 k − 1 ( M ; R ) .   If ( M , ω )   satisfies a weak form of the Lefschetz property, namely if multiplication by [ ω ] k − 1   is injective, then the usual symplectic flux group is trivial.
Example 25. Consider M = F × S 2   , where F   is a surface of genus g ≠ 1   . Then every cohomology class with nonzero square in H 2 ( M ; R )   is realised by a split symplectic form, with the symplectic area of the factors scaled suitably. For all these symplectic forms the Chern classes are the same, namely c 1 = ( 2 − 2 g ) P . D . [ S 2 ] + 2 P . D . [ F ]   and c 2 = ( 4 − 4 g ) P . D . [ M ]   . For those symplectic forms ω   whose cohomology class is a multiple of c 1   , the case k = 1   of Corollary  23 implies the triviality of the flux group Γ ω   . When [ ω ]   is not a multiple of c 1   , we can use the case k = 2   and the fact that c 1 2   and c 2   are nonzero to conclude that Γ 2   is trivial3 . As M   satisfies the hard Lefschetz property for every ω   , we again conclude the vanishing of Γ ω   .
Example 26. Let M = F × S 2   as before, with g ≠ 1   . Then the non-vanishing of the Chern numbers c 1 2   and c 2   gives rise to two potentially different extensions Flux ~ 2 : Symp ⟶ H 3 ( M ; R ) .   The difference between them corresponds to the difference c 1 2 ( ξ ) − 2 c 2 ( ξ ) ∈ H 4 ( E Symp ; R )   , which restricts trivially to the fiber M   . There are symplectic bundles with fiber M   which show that this difference class is non-trivial if g = g ( F ) ≥ 3   . Namely, let X → B   be an F   -bundle with nonzero signature. Then X × S 2   is an M   -bundle over B   for which c 1 2 ( ξ ) − 2 c 2 ( ξ ) ≠ 0 ∈ H 4 ( X × S 2 ; R )   .
However, the two extensions of Flux 2   are essentially the same. As M   satisfies the hard Lefschetz property, these extensions of Flux 2   are given by extensions of the usual flux homomorphism multiplied by the symplectic form. But the extensions of the usual flux homomorphism here come from F   , where we know that the extension is unique as a cohomology class, see [20] .
Example 27. Consider now M = T 2 × S 2   with a split symplectic form. Then clearly the flux group is non-trivial. However, if we pass from M   to its blowup M ^ = M # C P 2 ¯   , then M ^   is reducible (and has nonzero Chern numbers), so that Γ 2   must vanish. As M ^   satisfies the hard Lefschetz property, we conclude that the usual flux group Γ 1 = Γ ω   also vanishes.
Theorem  9 has the following consequence for the symplectomorphism groups.
Corollary 28. Suppose that [ ω ] k   is a bounded cohomology class. Then Γ k = 0   . In particular, if ω   represents a bounded class, then the usual symplectic flux group is trivial.
It is interesting to compare the above vanishing results for the symplectic flux group with the following:
Proposition 29. (cf. [22, 35] ) Assume that ( M , ω )   is symplectically aspherical, i. e. ω | π 2 ( M ) = 0   , and that π 1 ( M )   has finite center. Then Γ ω = 0   .
  • Proof. As the center of π 1 ( M )   is finite, and the flux is multiplicative when we replace a loop by a multiple, we may assume that the evaluation of a loop φ t   in Symp 0   whose flux we want to test bounds a 2   -disk D   in M   . If γ ⊂ M   is any closed loop, the degree 2   homology class of φ t ( γ )   can be represented by a 2   -sphere S   obtained by surgering the torus along a meridian, using two copies of D   . Now ω | π 2 ( M ) = 0   implies that 〈 Flux ω ( φ t ) , [ γ ] 〉 = ∫ φ t ( γ ) ω = ∫ S ω = 0 .   □
Example 30. Consider a surface bundle X   over a surface B   , such that both the base B   and the fiber F   have genus ≥ 2   . Then the second Chern number is nonzero, and the center of π 1 ( X )   is trivial, so that Γ 2 = 0   by Theorem  22 . On the one hand, there are many such X   which cannot satisfy the hard Lefschetz property (for any ω   ), so that we cannot conclude the vanishing of Γ ω   from Theorem  22 . On the other hand, it is always possible to choose ω   in such a way that it represents a bounded cohomology class4 , in which case Corollary  28 implies the vanishing of the flux group.
Again this last argument does not cover all cases, because for suitable surface bundles X   one can also choose ω   so that its cohomology class is not bounded. Nevertheless, Proposition  29 always applies, because X   is aspherical and its fundamental group has trivial center.

6 Final comments and remarks

6.1 Does the volume flux group depend on the smooth structure?

So far we do not know whether the non-triviality of the volume flux group depends on the smooth structure, or not. Example  16 is the closest we have come to seeing a dependence on the smooth structure, but the example is not conclusive. Recall also that there are examples of manifolds with free circle actions which are homeomorphic to smooth manifolds on which every smooth circle action must have fixed points, see [19] . Most of the information we have derived from the non-triviality of the volume flux group, about the fundamental group, homotopical irreducibility, simplicial volume, and about the minimal volume entropy, is homotopy invariant. However, this is not known for the minimal topological entropy, and is definitely false for the minimal volume.
Theorem 2 of [19] shows that in dimension 4   there exist homeomorphic manifolds such that one has vanishing minimal volume and the other one does not. It is also known that the minimal volume depends on the smooth structure in higher dimensions.
A modification of Example  16 that might be interesting is as follows: consider the connected sum of T 7   with a homotopy sphere. In certain cases this gives an exotic smooth structure on the torus. It seems like it should not have a smooth free circle action, but we have not proved this.
What can one say about the volume flux on this manifold?

6.2 Remarks on Gottlieb groups

Recall that the Gottlieb group G ( M )   of a manifold M   is the image of the evaluation homomorphism e v * : π 1 ( M M ) ⟶ π 1 ( M )   , see [13] and [24, 33] . It will be clear to the experts that some of our arguments concerning volume flux groups depend only on the fact that a loop of diffeomorphisms having non-trivial volume flux gives an element of infinite order in the Gottlieb group. Indeed, the non-triviality of the Gottlieb group is enough to conclude that the Euler characteristic of M   vanishes, see [13] , and for the irreducibility conclusion of Theorem  15 . However, it is not clear whether the other consequences of a non-trivial volume flux follow from Gottlieb theory alone. If M   has a Gottlieb element whose image under the Hurewicz map has infinite order in homology, then the simplicial volume of M   vanishes. However our proof of the vanishing of the minimal volume certainly does not apply in this generality.
Once again the situation is better for 3   -manifolds. If a closed 3   -manifold without any fake cells has non-trivial Gottlieb group, then it is Seifert fibered, and up to multiples the elements of the Gottlieb group are represented by circle actions. This follows from the Seifert fiber space conjecture, the final cases of which were settled by Casson–Jungreis [9] and Gabai [12] independently.
Our proof of Theorem  14 did not need these deep results, because the existence of non-trivial volume flux implies that the manifold is Haken.

6.3 Further developments

Extended flux homomorphisms arose first in [20] for the case of monotone symplectic forms.
There, a vanishing theorem for flux groups was proved as a byproduct of the search for extended flux homomorphisms, whereas in the present paper we obtain many more vanishing theorems in the general situation, where an extended flux homomorphism may not necessarily exist.
The results of [20, 21] illustrate how extended flux homomorphisms can help in understanding the homology of the groups Diff α   as discrete groups. For many of the situations where we have proved the existence of extensions of the flux as a crossed homomorphism in this paper, one can try to imitate the constructions of [21] in particular in order to find new non-trivial cohomology classes on diffeomorphism groups made discrete. In a related development, after most of the this work was completed, McDuff [27] took up the constructions of [20, 21] and of the present paper for symplectomorphism groups. Her point of view is somewhat different, with an emphasis on extending the Hamiltonian group to a certain subgroup of the symplectomorphism group that intersects all components of Symp   . Such an extension can be seen as the kernel of an extended flux homomorphism. McDuff [27] provides this extension in some cases where we proved only the vanishing of the flux group, but not the existence of the extended flux homomorphism.

Acknowledgements

The first two authors would like to thank T. Schick for a helpful discussion. This work was begun while the first author held a fellowship at the University of Munich funded by the European Differential Geometry Endeavour (EDGE), Research Training Network HPRN-CT-2000-00101, supported by The European Human Potential Programme. The paper was completed during visits of the second author to the University of Tokyo and to Stanford University. The third author is partially supported by JSPS Grant 16204005. We are grateful to all these institutions for their support.
References

  1. I. K. Babenko, Asymptotic invariants of smooth manifolds, Izv. Ross. Akad. Nauk. Ser. Mat. 56 (1992), 707–751 (Russian); Engl. transl. in Russian Acad. Sci. Izv. Math. 41 (1993), 1–38.
  2. I. K. Babenko, Asymptotic volumes and simply connected surgeries of smooth manifolds, Izv. Ross. Akad. Nauk. Ser. Mat. 58 (1994), 218–221 (Russian); Engl. transl. in Russian Acad. Sci. Izv. Math. 44 (1995), 427–430.
  3. A. Banyaga, Sur la structure du groupe des difféomorphismes qui prèservent une forme symplectique, Comment. Math. Helv. 53 (1978), 174–227.
  4. A. Banyaga, The structure of classical diffeomorphism groups, Mathematics and its Applications, vol. 400, Kluwer Academic Publishers Group, Dordrecht 1997.
  5. G. Besson, G. Courtois et S. Gallot, Volume et entropie minimale des espaces localement symétriques, Invent. math. 103 (1991), 417–445.
  6. W. Browder, On the action of Θ n ( ∂ π )   , in Differential and Combinatorial Topology, ed. S. S. Cairns, Princeton Univ. Press 1965.
  7. W. Browder, W.-C. Hsiang, G   -actions and the fundamental group, Invent. math. 65 (1982), 411–424.
  8. E. Calabi, On the group of automorphisms of a symplectic manifold, in Problems in Analysis, ed. R. Gunning, Princeton University Press, 1970, 1–26.
  9. A. Casson, D. Jungreis, Convergence groups and Seifert fibered 3   -manifolds, Invent. Math. 118 (1994), 441–456.
  10. P. E. Conner, E. E. Floyd, Differentiable periodic maps, Springer Verlag 1964.
  11. P. E. Conner, F. Raymond, Injective operations of the toral groups, Topology 10 (1971), 283–296.
  12. D. Gabai, Convergence groups are Fuchsian groups, Ann. of Math. (2) 136 (1992), 447–510.
  13. D. H. Gottlieb, A certain subgroup of the fundamental group, Amer. J. Math. 87 (1965), 840–856.
  14. D. H. Gottlieb, Splitting off tori and the evaluation subgroup of the fundamental group, Israel J. of Math. 66 (1989), 216–222.
  15. M. Gromov, Volume and bounded cohomology, Publ. Math. I.H.E.S. 56 (1982), 5–99.
  16. M. Hoster, D. Kotschick, On the simplicial volumes of fiber bundles, Proc. Amer. Math. Soc. 129 (2001), 1229–1232.
  17. W. Jaco, Lectures on three-manifold topology, CBMS Regional Conference Series in Mathematics no. 43, American Math. Society 1980.
  18. J. Kȩ dra, Remarks on the flux groups, Math. Res. Lett. 7 (2000), 279–285.
  19. D. Kotschick, Entropies, volumes, and Einstein metrics, Preprint arXiv:math.DG/0410215 .
  20. D. Kotschick, S. Morita, Signatures of foliated surface bundles and the symplectomorphism groups of surfaces, Topology 44 (2005). 131–149.
  21. D. Kotschick, S. Morita, Characteristic classes of foliated surface bundles with area-preserving holonomy, Preprint arXiv:math.SG/0409214 .
  22. F. Lalonde, D. McDuff, L. Polterovich, On the flux conjectures, in Geometry, topology, and dynamics (Montreal, PQ, 1995), 69–85, CRM Proc. Lecture Notes 15, Amer. Math. Soc., Providence, RI, 1998.
  23. F. Lalonde, D. McDuff, L. Polterovich, Topological rigidity of Hamiltonian loops and quantum homology, Invent. Math. 135 (1999), 369–385.
  24. G. Lupton, J. Oprea, Cohomologically symplectic spaces, toral actions and the Gottlieb group, Trans. Amer. Math. Soc. 347 (1995), 261–288.
  25. I. Madsen, R. J. Milgram, The classifying spaces for surgery and cobordism of manifolds, Annals of Math. Studies, vol. 92, Princeton University Press, Princeton 1979.
  26. D. McDuff, Symplectic diffeomorphisms and the flux homomorphism, Invent. Math. 77 (1984), 353–366.
  27. D. McDuff, Enlarging the Hamiltonian group, in preparation.
  28. D. McDuff, D. A. Salamon, Introduction to Symplectic Topology, second edition, Oxford University Press, 1998.
  29. J. W. Milnor, W. P. Thurston, Characteristic numbers of 3   -manifolds, Enseign. Math. 23 (1977), 249–254; reprinted in J. W. Milnor, Collected Papers, Vol. 2, Publish or Perish 1995.
  30. J. Moser, On the volume elements on a manifold, Trans. Amer. Math. Soc. 120 (1965), 286–294.
  31. K. Ono, Floer–Novikov cohomology and the flux conjecture, Preprint 2004.
  32. J. Oprea, A homotopical Conner–Raymond theorem and a question of Gottlieb, Canad. Math. Bull. 33 (1990), 219–229.
  33. J. Oprea, Gottlieb groups, group actions, fixed points and rational homotopy, Lecture Notes Series 29, Seoul National University, Research Institute of Mathematics, Global Analysis Research Center, Seoul 1995.
  34. G. P. Paternain, J. Petean, Minimal entropy and collapsing with curvature bounded from below, Invent. math. 151 (2003), 415–450.
  35. L. Polterovich, Growth of maps, distortion in groups and symplectic geometry, Invent. math. 150 (2002), 655–686.
  36. P. Scott, The geometries of 3   -manifolds, Bull. London Math. Soc. 15 (1983), 401–487.
  37. R. E. Stong, Notes on Cobordism Theory, Princeton University Press 1968.
  38. F. Waldhausen, Gruppen mit Zentrum und 3   -dimensionale Mannigfaltigkeiten, Topology 6 (1967), 505–517.

Institute of Mathematics, University of Szczecin, Wielkopolska 15, 70-451 Szczecin, Poland; and Mathematical Sciences, University of Aberdeen, Meston Building, King's College, Aberdeen AB24 3UE, Scotland, UK E-mail address: kedra@maths.abdn.ac.uk Mathematisches Institut, Ludwig-Maximilians-Universität München, Theresienstr. 39, 80333 München, Germany E-mail address: dieter@member.ams.org Department of Mathematical Sciences, University of Tokyo, Komaba, Tokyo 153-8914, Japan E-mail address: morita@ms.u-tokyo.ac.jp