Measure theory in the product of groups #
In this file we show properties about measure theory in products of measurable groups and properties of iterated integrals in measurable groups.
These lemmas show the uniqueness of left invariant measures on measurable groups, up to scaling. In this file we follow the proof and refer to the book Measure Theory by Paul Halmos.
The idea of the proof is to use the translation invariance of measures to prove μ(t) = c * μ(s)
for two sets s
and t
, where c
is a constant that does not depend on μ
. Let e
and f
be
the characteristic functions of s
and t
.
Assume that μ
and ν
are left-invariant measures. Then the map (x, y) ↦ (y * x, x⁻¹)
preserves the measure μ × ν
, which means that
∫ x, ∫ y, h x y ∂ν ∂μ = ∫ x, ∫ y, h (y * x) x⁻¹ ∂ν ∂μ
If we apply this to h x y := e x * f y⁻¹ / ν ((fun h ↦ h * y⁻¹) ⁻¹' s)
, we can rewrite the RHS to
μ(t)
, and the LHS to c * μ(s)
, where c = c(ν)
does not depend on μ
.
Applying this to μ
and to ν
gives μ (t) / μ (s) = ν (t) / ν (s)
, which is the uniqueness up to
scalar multiplication.
The proof in [Halmos] seems to contain an omission in §60 Th. A, see
MeasureTheory.measure_lintegral_div_measure
.
Note that this theory only applies in measurable groups, i.e., when multiplication and inversion
are measurable. This is not the case in general in locally compact groups, or even in compact
groups, when the topology is not second-countable. For arguments along the same line, but using
continuous functions instead of measurable sets and working in the general locally compact
setting, see the file MeasureTheory.Measure.Haar.Unique.lean
.
The map (x, y) ↦ (x, xy)
as a MeasurableEquiv
.
Equations
- MeasurableEquiv.shearMulRight G = { toEquiv := (Equiv.refl G).prodShear Equiv.mulLeft, measurable_toFun := ⋯, measurable_invFun := ⋯ }
Instances For
The map (x, y) ↦ (x, x + y)
as a MeasurableEquiv
.
Equations
- MeasurableEquiv.shearAddRight G = { toEquiv := (Equiv.refl G).prodShear Equiv.addLeft, measurable_toFun := ⋯, measurable_invFun := ⋯ }
Instances For
The map (x, y) ↦ (x, y / x)
as a MeasurableEquiv
with as inverse (x, y) ↦ (x, yx)
Equations
- MeasurableEquiv.shearDivRight G = { toEquiv := (Equiv.refl G).prodShear Equiv.divRight, measurable_toFun := ⋯, measurable_invFun := ⋯ }
Instances For
The map (x, y) ↦ (x, y - x)
as a MeasurableEquiv
with as inverse (x, y) ↦ (x, y + x)
.
Equations
- MeasurableEquiv.shearSubRight G = { toEquiv := (Equiv.refl G).prodShear Equiv.subRight, measurable_toFun := ⋯, measurable_invFun := ⋯ }
Instances For
The multiplicative shear mapping (x, y) ↦ (x, xy)
preserves the measure μ × ν
.
This condition is part of the definition of a measurable group in [Halmos, §59].
There, the map in this lemma is called S
.
The shear mapping (x, y) ↦ (x, x + y)
preserves the measure μ × ν
.
The map (x, y) ↦ (y, yx)
sends the measure μ × ν
to ν × μ
.
This is the map SR
in [Halmos, §59].
S
is the map (x, y) ↦ (x, xy)
and R
is Prod.swap
.
The map (x, y) ↦ (y, y + x)
sends the measure μ × ν
to ν × μ
.
The map (x, y) ↦ (x, x⁻¹y)
is measure-preserving.
This is the function S⁻¹
in [Halmos, §59],
where S
is the map (x, y) ↦ (x, xy)
.
The map (x, y) ↦ (x, - x + y)
is measure-preserving.
The map (x, y) ↦ (y, y⁻¹x)
sends μ × ν
to ν × μ
.
This is the function S⁻¹R
in [Halmos, §59],
where S
is the map (x, y) ↦ (x, xy)
and R
is Prod.swap
.
The map (x, y) ↦ (y, - y + x)
sends μ × ν
to ν × μ
.
The map (x, y) ↦ (yx, x⁻¹)
is measure-preserving.
This is the function S⁻¹RSR
in [Halmos, §59],
where S
is the map (x, y) ↦ (x, xy)
and R
is Prod.swap
.
The map (x, y) ↦ (y + x, - x)
is measure-preserving.
This is the computation performed in the proof of [Halmos, §60 Th. A].
This is the computation performed in the proof of [Halmos, §60 Th. A].
Any two nonzero left-invariant measures are absolutely continuous w.r.t. each other.
Any two nonzero left-invariant measures are absolutely continuous w.r.t. each other.
A technical lemma relating two different measures. This is basically [Halmos, §60 Th. A].
Note that if f
is the characteristic function of a measurable set t
this states that
μ t = c * μ s
for a constant c
that does not depend on μ
.
Note: There is a gap in the last step of the proof in [Halmos].
In the last line, the equality g(x⁻¹)ν(sx⁻¹) = f(x)
holds if we can prove that
0 < ν(sx⁻¹) < ∞
. The first inequality follows from §59, Th. D, but the second inequality is
not justified. We prove this inequality for almost all x
in
MeasureTheory.ae_measure_preimage_mul_right_lt_top_of_ne_zero
.
A technical lemma relating two different measures. This is basically [Halmos, §60 Th. A]. Note that
if f
is the characteristic function of a measurable set t
this states that μ t = c * μ s
for a
constant c
that does not depend on μ
.
Note: There is a gap in the last step of the proof in [Halmos]. In the last line, the equality
g(-x) + ν(s - x) = f(x)
holds if we can prove that 0 < ν(s - x) < ∞
. The first inequality
follows from §59, Th. D, but the second inequality is not justified. We prove this inequality for
almost all x
in MeasureTheory.ae_measure_preimage_add_right_lt_top_of_ne_zero
.
Left invariant Borel measures on a measurable group are unique (up to a scalar).
Left invariant Borel measures on an additive measurable group are unique (up to a scalar).
The map (x, y) ↦ (y, xy)
sends the measure μ × ν
to ν × μ
.
The map (x, y) ↦ (y, x + y)
sends the measure μ × ν
to ν × μ
.
The map (x, y) ↦ (xy, y)
preserves the measure μ × ν
.
The map (x, y) ↦ (x + y, y)
preserves the measure μ × ν
.
The map (x, y) ↦ (x, y / x)
is measure-preserving.
The map (x, y) ↦ (x, y - x)
is measure-preserving.
The map (x, y) ↦ (y, x / y)
sends μ × ν
to ν × μ
.
The map (x, y) ↦ (y, x - y)
sends μ × ν
to ν × μ
.
The map (x, y) ↦ (x / y, y)
preserves the measure μ × ν
.
The map (x, y) ↦ (x - y, y)
preserves the measure μ × ν
.
The map (x, y) ↦ (xy, x⁻¹)
is measure-preserving.
The map (x, y) ↦ (x + y, - x)
is measure-preserving.
A left-invariant measure is quasi-preserved by right-multiplication.
This should not be confused with (measurePreserving_mul_right μ g).quasiMeasurePreserving
.
A left-invariant measure is quasi-preserved by right-addition.
This should not be confused with (measurePreserving_add_right μ g).quasiMeasurePreserving
.
A right-invariant measure is quasi-preserved by left-multiplication.
This should not be confused with (measurePreserving_mul_left μ g).quasiMeasurePreserving
.
A right-invariant measure is quasi-preserved by left-addition.
This should not be confused with (measurePreserving_add_left μ g).quasiMeasurePreserving
.