Hard Lefschetz theorem for nonrational polytopes (Q705119): Difference between revisions

The so-called hard Lefschetz theorem came to play an important role in the theory of convex polytopes, when in 1980 \textit{R. P. Stanley} used it in his part of the proof of the so-called \(g\)-theorem [see Adv. Math. 35, 236--238 (1980; Zbl 0427.52006)]. He showed the necessity of McMullen's conditions characterizing the possible sequences of numbers of faces of a simple or simplicial convex polytope by translating them into statements on the cohomology of the quasi-smooth projective toric variety associated to the polytope. The hard Lefschetz theorem in this context implies unimodality of the \(h\)-vector of the polytope. Later \textit{P. McMullen} succeeded in finding a new purely convex-geometric proof of the necessity part of the \(g\)-theorem based on his polytope [Invent. Math. 113, 419--444 (1993; Zbl 0803.52007)]. He showed that a certain subalgebra of the polytope algebra associated to a given fixed simple polytope, namely the subalgebra generated by the classes of all its summands, admits a Lefschetz type decomposition. That is McMullen's version of a hard Lefschetz theorem for simple polytopes. In the case of nonsimple rational polytopes, the theory of toric varieties again provides a hard Lefschetz theorem, namely in the context of the intersection cohomology of the associated singular projective toric variety. \textit{G. Barthel, J.-P. Brasselet, K.-H. Fieseler} and \textit{L. Kaup} [Tohoku Math. J. 54, 1--41 (2002; Zbl 1055.14024)] and independently \textit{P. Bressler} and \textit{V. A. Lunts} [Compos. Math. 135, 245--278 (2003; Zbl 1024.52005)] developed a construction of the intersection cohomology of the associated toric variety directly from the normal fan of the polytope, as global sections of a sheaf on the fan provided with a suitable topology. Their construction also works for nonrational polytopes resulting in a combinatorial intersection cohomology theory for arbitrary polytopes, and both teams of authors conjectured the analogue of the hard Lefschetz theorem for this theory to hold. In this paper, Karu actually proves this conjecture by using the explicit recursive construction of combinatorial intersection cohomology, and reducing to the case of simple polytopes that was settled by McMullen. As McMullen does in his paper, Karu shows even more, namely the combinatorial analogue of the classical Hodge-Riemann bilinear relations. More precisely, the quadratic form defined via iterated multiplication with the Lefschetz-class and the intersection pairing, is positive or negative definite on the primitive parts of the cohomology. As was shown in the abovementioned papers the hard Lefschetz theorem for combinatorial intersection cohomology implies that its even Betti numbers coincide with the entries of the so-called toric \(h\)-vector introduced by Stanley. So in particular, Karu obtains as a corollary that the toric \(h\)-vector of an arbitrary polytope has nonnegative entries and is unimodal, which is not at all clear from its definition as an alternating sum.