Reduction of structure to parabolic subgroups (Q2169327)

The author works with linear algebraic groups over a field \(k\) of characteristic not \(2\). The paper is motivated by a result of \textit{J. Tits} [J. Algebra 131, No. 2, 648--677 (1990; Zbl 0697.20029)] which shows that, since every semisimple group is an inner form of a unique quasi-split group, classifying semisimple groups over \(k\) is equivalent to classifying all anisotropic forms of a semisimple quasi-split group. Recall that a linear algebraic group is called \textit{isotropic} if it contains a split torus which is non-central, and \textit{anisotropic} otherwise. Since forms of \(G\) are classified by cohomology sets, the author makes the following definitions. Definition: A torsor \([\gamma] \in H^1(K,G)\) is called \textit{isotropic} if there exists a proper parabolic subgroup \(P\subset G\) such that \([\gamma]\) is in the image of \[ H^1(K,P) \to H^1(K,G). \] The group \(G\) is called \textit{strongly isotropic} if all \([\gamma] \in H^1(K,G)\) are isotropic for all field extensions \(k\subset K\). For a quasi-split \(G\), the form \(_{\gamma}G\) is isotropic whenever \([\gamma]\in H^1(K,G)\) is, and so the paper asks, ``what are the strongly isotropic connected reductive linear algebraic groups over \(k\)?'' The main results are the following classifications. Theorem 1.3: The simple strongly isotropic groups over \(k\) are \begin{itemize} \item \(\mathrm{SP}_{2n}\), \item \(\mathrm{Spin}(q)\), where \(q\) is a ten dimensional regular quadratic form with trivial discriminant and split Clifford algebra, and \item \(\mathrm{SL}_n(D)/\mu_d\), where \(D\) is a central division algebra over \(k\), \(n>1\), \(d|n\deg(D)\), and some prime divisor of \(n\) does not divide \(d\). \end{itemize} Theorem 1.4: Let \(G\) be a semisimple group such that for any simple factor of \(G\) of type \(A_{n-1}\), the integer \(n\) is squarefree. Then \(G\) is strongly isotropic if and only if it admits a simple strongly isotropic quotient. Section 2 of the paper shows that Theorem 1.3 for split groups is essentially a result of J.~Tits, as it can be easily assembled from his other results. Section 3 proves various technical lemmas which allow one to transfer the property of being strongly isotropic. For example, some conditions on normal subgroups \(H\subset G\) which cause \(G\) to be strongly isotropic if and only if \(G/H\) is strongly isotropic are given, and it is shown that if \(G\) is reductive then it is strongly isotropic whenever the derived subgroup is strongly isotropic, etc. Section 4 lays groundwork for the remainder of the paper. It uses the fact that when \(G\) is the automorphism group of some structure, say a central simple algebra with involution or perhaps a quadratic form, then \(G\)-torsors correspond to twisted forms of the underlying structure. Notions of anisotropy exist for the underlying structures and the group is anisotropic whenever the structure is. In this section the author constructs, over some field extension of \(k\), anisotropic central simple algebras with symplectic involution or with unitary involution and anisotropic ten dimensional quadratic forms with given non-trivial discriminant and non-split Witt invariant. Section 5 then proves Theorem 1.3 as stated. The result for split groups shows that any strongly isotropic \(G\) must be of type \(A_n\), \(C_n\), or \(D_5\) and this allows the author to proceed by types. For each type they show the listed groups are strongly isotropic, and that any other group of those types corresponds to an underlying anisotropic object constructed in Section \(4\), and therefore is not strongly isotropic. The divisibility conditions on \(n\) and \(d\) for groups of type \(A\) come from considerations involving the dimensions of ideals in twisted flag varieties. Section 6 looks at strongly isotropic semisimple groups and contains the proof of Theorem 1.4. The fact that having a simply strongly isotropic quotient implies strong isotropy is a result of Section \(3\), and so this section focuses on the converse. The author finds a candidate simple quotient \(G\to G_1\), and shows that \(G_1\) is strongly isotropic by using the vanishing of certain Brauer invariants. This is also done by type using the classification of Theorem 1.3. Finally, Section 7 focuses on semisimple groups of type \(A\), in particular groups of the form \(S_C = (\mathrm{SL}_{n_1}\times\ldots\times\mathrm{SL}_{n_r})/C\cap(\mu_{n_1}\times\ldots\times\mu_{n_r})\) where \(C\subset \mathbb{G}_m^r\) is an algebraic subgroup. It is shown in Theorem 7.2 that strong isotropy of \(S_C\) is detected by an intricate divisibility criterion involving the \(n_1,\ldots,n_r\) and characters of \(\mathbb{G}_m^r\) which vanish on \(C\). This criterion is then used in Example 7.3 to show that the squarefree assumptions placed on the roots systems in Theorem 1.4 are necessary.