On the push-forwards for motivic cohomology theories with invertible stable Hopf element (Q276043)

The author proves a general structural result for \(\eta\)-inverted extraordinary cohomology theories of algebraic varieties, by showing that these admit push-forward maps for certain classes of morphisms. To set the stage, recall that a (bigraded) cohomology theory on algebraic varieties is a contravariant functor \(A^{*,*}: Sm(k) \to\mathrm{Ab}\) (where \(Sm(k)\) is the category of smooth varieties and Ab is the category of abelian groups) satisfying certain conditions (typically homotopy invariance, localisation sequences of some sort, Mayer-Vietoris, etc). When choosing the right set of conditions (in particular homotopy invariance) then cohomology theories are represented in the stable motivic homotopy category \(SH(k)\), and in fact every object of \(SH(k)\) represents a cohomology theory, just as in classical topology. For the purpose of of this review, ``cohomology theory'' shall mean ``functor represented on \(SH(k)\)''. Of particular interest are the \textit{oriented} cohomology theories. There are various equivalent ways of specifying this condition: such a theory admits Chern classes, satisfies the projective bundle formula, has Thom isomorphisms, splits the infinite projective space, admits pushfowards for proper morphisms, etc. There is a different way of looking at this. Namely, recall the algebraic Hopf map \(\eta: \mathbb{A}^2 \setminus 0 \to \mathbb{P}^1\). Since in \(SH(k)\) there is a weak equivalence \(\mathbb{A}^2 \setminus 0 \simeq \mathbb{P}^1 \wedge \mathbb{G}_m\), this defines a map \(\eta: \mathbb{G}_m \to S \in SH(k)\). A necessary (but not sufficient) condition for orientability is that \(A(\eta) = 0\). Examples of orientable theories abound: motivic cohomology, algebraic \(K\)-theory, algebraic cobordism, and so on. What about non-orientable theories? Those certainly exist. Examples include Hermitian \(K\)-theory, derived Witt theory, oriented Chow groups, and also (tautologously) motivic stable cohomotopy theory. Handling these theories is much harder -- there is no projective bundle formula, no proper pushforward, no Thom isomorphism, and so on. One might argue that those theories farthest from orientable are the ones which have the property that instead of \(A(\eta)=0\), \(A(\eta)\) is actually an isomorphism. These are the so-called \(\eta\)-inverted theories under investigation in the article under review. Generalising a geometric method for derived Witt theory, the author computes the cohomology of projective spaces twisted by an arbitrary vector bundle in terms of the cohomology of the base, for any \(\eta\)-inverted theory. Using this he defines twisted pushforwards along projective morphisms, again for all \(\eta\)-inverted theories, and establishes their basic properties (functoriality, base change, and the projection formula).