Descent Theorem #
We provide a proof of the following result.
Let G be a group and f : G →* G an endomorphism of G that maps every
subgroup of G into itself (e.g., f = fun g ↦ g ^ n when G is commutative).
If there is a finite subset s : Set G and there exists a "height" function h : G → ℝ
and constants a, b, c : ℝ such that
ssurjects onto the quotientG ⧸ f(G),- for all
g ∈ sandx : G,h x ≤ a * h (g * x) + c, - for all
x : G,h (f x) ≥ b * h x - c, - for all
B : ℝ, there are only finitely manyx : Gsuch thath x ≤ B, and 0 ≤ a < b, thenGis finitely generated. SeeGroup.fg_of_descent/AddGroup.fg_of_descent.
We use this to deduce a more specific version when G is commutative and f is the nth power
endomorphism and finally an even more specific version with n = 2, replacing the upper
and lower bound for the height function by the "approximate parallelogram law"
∀ x y, |h (x * y) + h (x / y) - 2 * (h x + h y)| ≤ C.
See CommGroup.fg_of_descent / AddCommGroup.fg_of_descent and
CommGroup.fg_of_descent' / AddCommGroup.fg_of_descent'.
This last version is one of the main ingredients of the standard proof of the
Mordell-Weil Theorem. It allows to reduce the statement to showing that G / 2 • G is finite
(where G is the Mordell-Weil group`).
Implementation note #
Replacing ℝ by an ordered field ({R : Type*} [LinearOrder R] [Field R] [IsOrderedRing R])
works, but makes the type check quite slow (and to_additive needs some help...).
As the application(s) work with ℝ-valued height functions, we think that generalizing
is not really worth the trouble.
If G is a group and f : G →* G is an endomorphism sending subgroups into themselves,
and if there is a "height function" h : G → ℝ with respect to f and a finite subset s
of G, then G is finitely generated.
If G is an additive group and f : G →+ G is an endomorphism sending
subgroups into themselves, and if there is a "height function" h : G → ℝ with respect
to f and a finite subset s of G, then G is finitely generated.
If G is a commutative group and n : ℕ, h : G → ℝ satisfy
G / G ^ nis finite,- for all
g x : G,h x ≤ a * h (g * x) + c g, - for all
x : G,h (x ^ n) ≥ b * h x - c₀, - for all
B : ℝ, there are only finitely manyx : Gsuch thath x ≤ B, where0 ≤ a < bandc₀are real numbers,c : G → ℝ, thenGis finitely generated.
If G is a commutative additive group and n : ℕ, h : G → ℝ satisfy
G / n • Gis finite,- for all
g x : G,h x ≤ a * h (g + x) + c g, - for all
x : G,h (n • x) ≥ b * h x - c₀, - for all
B : ℝ, there are only finitely manyx : Gsuch thath x ≤ B, where0 ≤ a < bandc₀are real numbers,c : G → ℝ, thenGis finitely generated.
If G is a commutative group and n : ℕ, h : G → ℝ satisfy
G / G ^ 2is finite,0 ≤ h xfor allx : G,- there is
C : ℝsuch that for allx y : G,|h (x * y) + h(x / y) - 2 * (h x + h y)| ≤ C, - for all
B : ℝ, there are only finitely manyx : Gsuch thath x ≤ B, thenGis finitely generated.
If G is a commutative additive group and n : ℕ, h : G → ℝ satisfy
G / 2 • Gis finite,0 ≤ h xfor allx : G,- there is
C : ℝsuch that for allx y : G,|h (x + y) + h(x - y) - 2 * (h x + h y)| ≤ C, - for all
B : ℝ, there are only finitely manyx : Gsuch thath x ≤ B, thenGis finitely generated.