2211.03276v1
2211.03276v1
2211.03276v1
ALGEBRAIC GROUPS
DYLON CHOW
arXiv:2211.03276v1 [math.NT] 7 Nov 2022
1. Introduction
Let F be a number field, and let ΩF denote the set of places of F . If v ∈ ΩF , we denote
by Fv the completion of F for the topology defined by v. Let G be a connected reductive
algebraic group defined over F . Each extension F → Fv defines a map ωv : H 1 (F, G) →
H 1 (Fv , G). We thus get a canonical map
Y
ω : H 1 (F, G) → H 1 (Fv , G).
v∈ΩF
This map is a much-studied map. For instance, it is known that the kernel of ω is finite. If ω
is injective, G is said to satisfy the Hasse principle. Some algebraic groups, including simply
connected groups and groups with trivial center, satisfy the Hasse principle, but some don’t.
The map ω is not always surjective, either. In fact, for tori, the failure of ω to be surjective
is related to the failure of weak approximation for reductive groups. For G = PGLn , the
Galois cohomology set H 1 (F, G) classifies equivalence classes of central simple algebras of
degree n2 over F . Every central simple algebra over F splits almost everywhere, so to get
a reasonable surjectivity result one must restrict to the direct sum of the H 1 (Fv , G). The
map from H 1 (F, G) into the direct sum is not surjective in general, either. However, choose
any non-archimedean place v0 of F and consider the diagonal map
M
ϕ : H 1 (F, G) → H 1 (Fv , G).
v6=v0
This article is concerned with the following question: For which linear algebraic groups G is
ϕ surjective?
There are two papers that address this question. Borel and Harder ([HB78]) proved that
ϕ is surjective whenever G is semisimple. They used this result to prove the existence of
discrete cocompact subgroups in the groups of rational points of reductive groups over non-
archimedean local fields of characteristic zero. Prasad and Rapinchuk ([PR06]) strengthened
Borel and Harder’s results and showed that ϕ is surjective for some reductive groups that
are not semisimple. This has applications to lattice counting problems (cf. [BL19]).
This article gives criteria for ϕ to be surjective. One criterion involves the radical of
G and is essentially given in [PR06]. The other involves the cohomology of maximal tori
in quasisplit groups. To obtain this criterion, we relate ϕ to another local-global map
1
2 DYLON CHOW
involving the hypercohomology groups of a complex of tori, i.e., the "abelian cohomology
groups" studied by Borovoi.
2.1. We use F to denote a field. Let F be an algebraic closure of F and write F s for
the separable closure of F in F . We let Γ = ΓF denote the Galois group of F s over F ; it
is a profinite topological group with the Krull topology. Then H i (F, H) denotes the i-th
cohomology set of the Galois group Γ, with coefficients in H(Fs ) (i = 0, 1) and, if G is
commutative, the i-th cohomology group of Γ in G(Fs ) for all i ∈ N.
2.3. In this section we review the construction of the abelian Galois cohomology groups.
Let F be a field of characteristic 0. For a connected reductive group G over K, let G′ denote
the derived group of G and let Gsc denote the universal covering group of G′ . We consider
the composition
ρ : Gsc → G′ ֒→ G.
The complex [Gsc → G] is a crossed module ([Bor98] Lemma 3.7.1).
2.4. We will use results on crossed modules. For the definition and basic properties of
crossed modules, see [Bor98]. A morphism f : T → U of tori is defined over F is a crossed
module in a natural way and we can consider its cohomology groups H i (F, T → U ). We
refer the reader to [Bor98] for the definition and properties of crossed modules and their
cohomology. At the same time, T → U is also a complex of tori of length 2 and we can
consider the Galois hypercohomology of this complex. In order to do this, we need to specify
in what degrees the terms of this complex are placed. If we place T in degree −1 and U
in degree 0, then the Galois hypercohomology of the resulting complex coincides with the
cohomology of the crossed module T → U . This will be our convention, but we warn the
reader that this convention is different from the one used in [KS99, Appendix A], where
the complexes of tori of length 2 are placed in degrees 0 and 1. Thus H 1 in our notation
coincides with H 2 in the notation of [KS99, Appendix A].
2.5. Let T ⊂ G be a maximal torus defined over F and let Tsc be the inverse image ρ−1 (T )
of T under ρ. We consider the complex of tori T • = [Tsc → T ] where Tsc is in degree −1
and T is in degree 0. We get a complex Tsc (F ) → T (F ) of Gal(F /F )-modules.
2.6. The natural morphism of crossed modules from [1 → G] to [Gsc → G] ([Bor98, section
3.10]) induces a map in crossed module hypercohomology:
2.8. We recall the definition of the abelianization map. For any maximal torus T in the
reductive group G, the morphism of crossed modules
[Tsc (F ) → T (F )] → [Gsc (F ) → G(F )]
is a quasi-isomorphism, and so it defines an isomorphism
3. Surjectivity Results
Now we will give a more flexible criterion in terms of maximal tori. First, we reformulate
our conjecture in terms of the Galois hypercohomology of complexes of tori. Let Gsc be the
simply connected cover of Gder and let ρ be the composition
ρ : Gsc → Gder ֒→ G.
Let T be a maximal torus in G and let Tsc = ρ−1 (T ). The complex of tori Gab := [Tsc → T ]
forms a crossed module, and its first hypercohomology group (Tsc is in degree −1) is called
the first abelian cohomology group of G: Hab 1 (F, G) = H1 (F, [T
sc → T ]). There is an
1 1
"abelianization map" ab1 : H (F, G) → Hab (F, G).
Theorem 3.2. Let G be a reductive group over F . Let v0 be any non-archimedean place of
F . The map ϕ is surjective if and only if the map
M
1 1
Hab (F, G) → Hab (Fv , G)
v6=v0
is surjective.
To state our next criterion, we use the exact sequence [KS99, A.1.1] or [Bor98, Proposition
2.12]
· · · → H 1 (F, T ) → Hab
1
(F, G) → H 2 (F, Tsc ) → H 2 (F, T ) → . . . ,
which gives a commutative diagram with exact rows
H 1 (F, T ) 1 (F, G)
Hab H 2 (F, Tsc ) H 2 (F, T )
L L L L
v6=v0 H 1 (Fv , T ) v6=v0
1 (F , G)
Hab v v6=v0 H 2 (Fv , Tsc ) v6=v0 H 2 (Fv , T ).
Proof. Since G is simply connected, it contains an induced torus T , i.e., T is a finite product
of the form RE/F (GL1 ) with E a finite extension of F . By class field theory, the claim holds
for T = RE/F (GL1 ).
Acknowledgements
The author thanks Mikhail Borovoi for helpful comments and Ramin Takloo-Bighash for
his interest in this work.
References
[BL19] Mikhail Belolipetsky and Alexander Lubotzky. Counting non-uniform lattices. Israel Journal of
Mathematics, 232(1):201–229, 2019.
[Bor98] Mikhail Borovoi. Abelian Galois cohomology of reductive groups, volume 626. American Mathemat-
ical Soc., 1998.
[HB78] G. Harder and A. BOREL. Existence of discrete cocompact subgroups of reductive groups over local
fields. 1978(298):53–64, 1978.
[KS99] Robert E. Kottwitz and Diana Shelstad. Foundations of twisted endoscopy. Number 255 in
Astérisque. Société mathématique de France, 1999.
[PR06] Gopal Prasad and Andrei S. Rapinchuk. On the existence of isotropic forms of semi-simple algebraic
groups over number fields with prescribed local behavior. Advances in Mathematics, 207(2):646–660,
2006.