Positivity and Kleiman transversality in equivariant \(K\)-theory of homogeneous spaces (Q610656)

The authors prove the conjectures of \textit{W. Graham} and \textit{S. Kumar} [Int. Math. Res. Not. 2008, Article ID rnn 093, 43 p. (2008; Zbl 1185.14043)] and \textit{S. Griffeth} and \textit{A. Ram} [Eur. J. Comb. 25, No. 8, 1263--1283 (2004; Zbl 1076.14068)] about the equivariant \(K\)-theory of generalized flag varieties \(G/P\). Several special cases of these conjectures have been studied by \textit{W. Fulton} and \textit{A. Lascoux} [Duke Math. J. 76, No. 3, 711--729 (1994; Zbl 0840.14007)], \textit{H. Pittie} and \textit{A. Ram} [Electron. Res. Announc. Am. Math. Soc. 5, No. 14, 102--107 (1999; Zbl 0947.14025)] and \textit{O. Mathieu} [J. Pure Appl. Algebra 152, No. 1-3, 231--243 (2000; Zbl 0978.22016)]. The cohomology ring of the homogeneous space \(G/P\) has the following positivity property (here \(G\) is a complex linear algebraic group): it has an additive basis formed by classes of algebraic subvarieties, the Schubert varieties. Using the transitive group action, as pioneered by Kleiman, these varieties can be translated generically; subsequently intersecting them yields cycles whose multiplicities, i.e., the structure constants, are positive by virtue of being algebraic. This positivity extends beyond ordinary cohomology. Graham generalized it to torus equivariant cohomology of \(G/P\), confirming conjectures of Billey and Peterson. In this context the coefficients are polynomials; positivity means that, expressed as polynomials in the simple roots, the coefficients are nonnegative integers. Moreover, \textit{M. Brion} [J. Algebra 258, No. 1, 137--159 (2002; Zbl 1052.14054)] proved the positivity for ordinary \(K\)-theory, after it had been conjectured by \textit{A. S. Buch} [Acta Math. 189, No. 1, 37--78 (2002; Zbl 1090.14015)]. In this context, positivity means sign alternation: if the dimension of a subvariety differs from the expected dimension by \(i\) in a given intersection product, then the sign of the coefficient on its class is \((-1)^i\). In equivariant \(K\)-theory, the previous conjectures makes precise the notion of positivity for polynomials in terms of which the alternation is phrased. The author proves the previous conjectures using an appropriate generalization of Kleiman transversality. Ordinary Kleiman transversality concerns the movement of subvarieties of \(G/P\) into general position using the transitive group action. This has consequences for non-equivariant cohomology theories because translation preserves rational equivalence. Equivariantly, on the other hand, translation alters the classes of cycles. Thus, the authors' transversality principle takes place on (finite-dimensional approximations) of Borel mixing spaces of \(G/P\) and of the \(S\)-subvarieties of \(G/P\) (here \(S\) is a torus of \(G\) with some positivity conditions). The non-equivariant homological invariants of the approximation \(\mathcal{X}\) of the Borel mixing space of \(G/P\) are the equivariant invariants of \(G/P\). This space \(\mathcal{X}\) is a bundle over a product of sufficiently large projective spaces with fiber \(G/P\). There is a group scheme, called the mixing group, which acts on \(\mathcal{X}\) with finitely many orbits. This action is derived from the structure of \(\mathcal{X}\) as a bundle such that the fibers and the basis have transitive automorphism groups. What makes things simpler in cohomology, as opposed to \(K\)-theory, is that cohomology only requires knowledge on a Zariski open subset. In Anderson's proof [Positivity in the cohomology of flag bundles (after Graham). Preprint, \url{arXiv:math.AG/0711.0983}] of Graham equivariant cohomology positivity each of the relevant cohomology computations is carried out by intersection that occurs in one cell of a paving of the mixed space. As an appropriate group action is transitive, ordinary Kleiman transversality suffices. One must then push down to the base of the mixing space, but this operation transfers the positivity to the resulting class. In \(K\)-theory, one cannot restrict to an open cell, but must work with closed subvarieties where the group actions is not transitive. Moreover, pushing forward to the base can have higher direct images, a priori causing some negative coefficients. The obstacle of non-transitivity is dealt with by using a result of \textit{S. J. Sierra} [Algebra Number Theory 3, No. 5, 597--609 (2009; Zbl 1180.14048)]. Getting around the second obstacle is accomplished by stipulating rational singularities, taking the cue from Brion's phrasing of the results in ordinary \(K\)-theory [\textit{M. Brion}, J. Algebra 258, No. 1, 137--159 (2002; Zbl 1052.14054)]). Then, the authors modify Brion's geometric argument that brings Kawamata-Viehweg vanishing to bear. The computation of coefficients is reduced to an Euler characteristic using the homological transversality, then it is shown that this class is a sum of at most one term with well-defined sign by using this vanishing.