A non-split torsor with trivial fixed point obstruction. (Q1812166)

Let \(k\) be an algebraically closed field, and let \(G\) be a linear algebraic group over \(k\). It is known that birational isomorphism classes of generically free \(G\)-varieties such that \(k\left( X\right) ^{G}=K\) are in one-to-one correspondence with \(H^{1}\left( K,G\right) ,\) where \(K\) is the field of \(G\)-invariant rational functions on \(X\). Such an \(X\) is called split if it corresponds to the trivial class in \(H^{1}\left( K,G\right) ,\) equivalently if \(X\) is birationally isomorphic to \(Y\times G\) as a \(G\)-variety for some \(Y\) with trivial \(G\)-action, or the quotient map \( X\rightarrow X/G\) has a rational section. Previously, \textit{Z. Reichstein} and \textit{B. Youssin} proved the following proposition [Pac. J. Math. 200, 207--249 (2001; Zbl 1054.14063)]: Let \(X\) be a generically free \(G\)-variety. If \(X\) has a smooth \(H\)-fixed point for some non-toral diagonalizable subgroup \(H\) of \(G\), then \(X\) is not split. Here, the issue is to discover if the converse to the proposition is true, subject to certain conditions. To be more precise, if \(G\) is connected and \(X \) is a smooth projective generically free \(G\)-variety such that \( X^{H}=\emptyset \) for every diagonalizable non-toral subgroup \(H\) of \(G\), does it follow that \(X\) is split? The answer to the question posed above is ``no'', as the paper shows. The main theorem is that, for \(G=\text{P} \lim\text{GL}_{p}\) for some odd prime \(p\), then there exists a \(G\)-variety \(X\) such that \(X\) is not split, \( X^{H}=\emptyset \) for every diagonalizable non-toral subgroup \(H\) of \(G,\) and \(k\left( X\right) ^{G}/k\) is purely transcendental.