No label defined (Q4950797)

Let \(M\) be a compact \(2n\)-dimensional symplectic manifold, let \(\{\cdot,\cdot\}\) be the induced Poisson brackets. The Koszul differential \(\delta: \Omega^k(M) \to \Omega^{k-1}(M)\) can be defined by \[ \begin{multlined}\delta(f_0 df_1\wedge\cdots\wedge df_k) = \sum_{i=1}^k (-1)^{i+1} \{ f_0,f_i\} df_1 \wedge \cdots \wedge \widehat{df_i} \wedge \cdots \wedge df_k\\ + \sum_{i<j} (-1)^{i+j} f_0 d\{ f_i,f_j\}\wedge df_1 \wedge \cdots \wedge \widehat{df_i} \wedge \cdots \wedge \widehat{df_j} \wedge \cdots df_k .\end{multlined} \] It satisfies \(\delta^2 = 0 \) and defines the canonical homology \(H^{\text{can}}_k(M)\) of \(M\). \textit{J.-L. Brylinski} showed in [J. Differ. Geom. 28, 93-114 (1988; Zbl 0634.58029)] that there is an isomorphism \(H^{\text{can}}_k(M) \cong H^{2n-k}(M)\) between the canonical homology and the de Rham cohomology. In analogy to Hodge theory for Riemannian manifolds he conjectured that each de Rham cohomology class has a representative \(\nu\) satisfying \(d\nu = \delta\nu = 0\), and he proved it for compact Kähler manifolds. In the present paper the authors show that the conjecture does not hold in general. Let \(N\) be the 3-dimensional compact Heisenberg manifold. Then \(M=N\times S^1\) is called the Kodaira-Thurston manifold. The Brylinski conjecture is disproven by exhibiting a suitable 3-dimensional cohomology class on \(M\). This illustrates a theorem of \textit{O. Mathieu} [Comment. Math. Helv. 70, 1-9 (1995; Zbl 0831.58004)] where it is shown that Brylinski's conjecture holds if and only if \(M\) satisfies the hard Lefschetz theorem, i.e. if and only if for any \(k\leq n\) the cup product \([\omega]^k: H^{n-k}(M)\to H^{n+k}(M)\) is an isomorphism, where \(\omega\) denotes the symplectic form. For the entire collection see [Zbl 0934.00021].