field is discrete and cocompact in its adèles


For brevity, we write Pf for the set of finite places of K, and P∞ for the set of infinite places. We also write ∏′ for a restricted direct product. Then

𝔸K=∏′v∈Pf′⁢Kv×∏v∈P∞Kv
Theorem 1.

K is discrete as a subgroup of AK.

Proof. Since 𝔸K is a topological ring, it suffices to show that there is a neighborhood in 𝔸K meeting K in only 0.

Let

U=∏v∈Pf𝔬v×∏v∈P∞B⁢(0,12)

Since 𝔬v is open in Kv for v finite, this is an open set. (Note that 𝔬v=𝒪Kv, the ring of algebraic integers of Kv).

Now consider an element x∈U∩K⊂𝔸K. If x=(xv), then for v finite, xv∈𝔬v, and for v infiniteMathworldPlanetmath, xv∈B⁢(0,12). Assume x≠0. Then

|xv|v≤{1v⁢ finite12<1v⁢ infinite

but then

∏|x|v=∏|xv|v<1

in contradictionMathworldPlanetmathPlanetmath to the product formula. Thus x=0 and we are done.

The above theorem is very sensitive to the fact that all places are included in 𝔸K. For example, it is clear that the image of 𝔸K in ∏v∈P∞Kv is dense, since Kv is characterized by an embeddingMathworldPlanetmathPlanetmath K↪Kv≅ℝ,ℂ,ℚp. Then by an argument familiar from Minkowski’s theorem, 𝒪K is a full-rank latticeMathworldPlanetmathPlanetmath in the image of K. But K is the ℚ-span of that lattice, so is dense in Kv.

Furthermore, the same is true for the finite places:

Proposition 2.

The image of K in ∏′v∈Pf⁢ov is dense.

Proof. Suppose x=(xv)v∈∏′v∈Pf⁢𝔬v. We show that x can be approximated as closely as desired by an element of K by showing that for any ideal I⊂𝒪K, there is y∈K such that y-xv∈I⁢𝔬v for each v∈Pf.

First multiply through by some z so that everything is in 𝒪K: choose z≠0 such that z⁢xv∈𝔬v for all v∈Pf. This is possible since all but finitely many xv are already in 𝔬v. Thus y-xv∈I⁢𝔬v is equivalentMathworldPlanetmathPlanetmathPlanetmathPlanetmathPlanetmath to z⁢y-z⁢xv∈(z⁢I)⁢𝔬v⊂I⁢𝔬v. So assume wlog that xv∈𝔬v for all v; in the end simply divide by z to recover the general case. But then the existence of y is guaranteed by the Chinese Remainder TheoremMathworldPlanetmathPlanetmathPlanetmath, since if I=∏𝔭iei, then I⁢𝔬v=𝔭iei for some i.

It is true, though somewhat harder to prove, that K is in fact dense in 𝔸K if even one place is missing from the productPlanetmathPlanetmathPlanetmath!

Theorem 3.

𝔸K/K is compactPlanetmathPlanetmath.

Proof. The set

U=∏v∈Pf𝔬v×∏v∈P∞Kv

is open in 𝔸K.

Claim first that K+U=𝔸K. Choose (xv)∈𝔸K. There is a finite setMathworldPlanetmath S of finite places v such that xv∉𝔬v for v∈S. Using an argument identical to the approximation argument above, choose y∈K such that y-xv∈𝔬v,v∈S and y∈𝔬v,v∉S. Then (xv-y)v is in 𝔬v for v∈S, is in 𝔬v for v∉S but finite, and is in Kv for v infinite. Thus (xv-y)v∈U and we are done.

Claim next that K∩U=𝒪K. ⊃ is obvious. To see ⊂, note that an element of K∩U is an element of K that is integralDlmfPlanetmath at every finite place, so it is integral and is in 𝒪K.

Thus we get a natural map U↪𝔸K↠𝔸K/K. This map is surjectivePlanetmathPlanetmath since K+U=𝔸K, and its kernel is K∩U. So it suffices to show that U/(K∩U)=U/𝒪K is compact. There is obviously an exact sequence induced by the projection U→∏v∈P∞Kv,

∏v∈Pf𝔬v→U/𝒪K→∏v∈P∞Kv/𝒪K→0

The left-hand side is compact since each 𝔬v is, and the right-hand side is

ℝr1+2⁢r2/𝒪K

which know is compact since 𝒪K forms a full-rank lattice in ℝn. Thus U/𝒪K is also compact and we are done.

So we have shown that 𝔸K is a locally compact ring, and that K⊂𝔸K is discrete and cocompact. This is analogous to two other situations with which we are familiar:

ℝ⁢ is locally compact, ⁢ℤ⊂ℝ⁢ is discrete and cocompact
∏v∈P∞Kv⁢ is locally compact, ⁢𝒪K⊂∏v∈P∞Kv⁢ is discrete and cocompact

This is a useful concept because in such a situation one can do Fourier analysis. For example, if f:ℝ→ℝ is a C∞ functionMathworldPlanetmath with exponential decay (or at least integrable on all of ℝ), then we can define its Fourier transformDlmfMathworldPlanetmath f^, and the Poisson summation formula

∑n∈ℤf⁢(n)=∑n∈ℤf^⁢(n)

relates the two. The same theory thus exists for appropriately defined functions f:𝔸K→ℝ, and the Poisson formula again holds with the sum over K rather than over ℤ. This can be used to show that the L-functions have analytic continuations, just as the real Poisson formula is used to show this for ζ.

Title field is discrete and cocompact in its adèles
Canonical name FieldIsDiscreteAndCocompactInItsAdeles
Date of creation 2013-03-22 18:00:05
Last modified on 2013-03-22 18:00:05
Owner rm50 (10146)
Last modified by rm50 (10146)
Numerical id 4
Author rm50 (10146)
Entry type Theorem
Classification msc 11R56