Let G be a reductive algebraic group over a field k. Let S be a maximal split torus, Z its centraliser and N its normaliser. The Weyl group W is then defined to be the quotient N(k)/Z(k). Now we cannot hope for W to be realisable as a subgroup of G, but I would like to know how close we can get.

There is a classical result of Tits in the case where G is split which says that for each simple reflection s in the Weyl group, we can find a lift w_{s} in G with the property that these lifts satisfy the braid relations. They do not however square to the identity, instead square to an order 2 element of S, and we get an extension of W by an elementary abelian 2-group embedding in G.

So my question ends up becoming, what generalisation of the above theorem of Tits exists when G is no longer assumed to be split? Ideally I'd get an answer for general reductive G, and if there happens to be a simpler formulation in the quasi-split case, I'd be interested in hearing that too.

Normalisateurs de toresby Tits (part I, with the promised second part never published) appears in J. Algebra 4 (1966), 96-116; the comment about the nonsplit case pointed out by Guntram occurs already in the first paragraph. Also, the Borel-Tits framework allows for reductive rather than just semisimple groups; but the essential arguments here involve the latter case. – Jim Humphreys Oct 14 '10 at 11:54