Infinite sum over a topological monoid #
THIS FILE IS SYNCHRONIZED WITH MATHLIB4. Any changes to this file require a corresponding PR to mathlib4.
This sum is known as unconditionally convergent, as it sums to the same value under all possible permutations. For Euclidean spaces (finite dimensional Banach spaces) this is equivalent to absolute convergence.
Note: There are summable sequences which are not unconditionally convergent! The other way holds
generally, see has_sum.tendsto_sum_nat
.
References #
- Bourbaki: General Topology (1995), Chapter 3 §5 (Infinite sums in commutative groups)
Infinite sum on a topological monoid
The at_top
filter on finset β
is the limit of all finite sets towards the entire type. So we sum
up bigger and bigger sets. This sum operation is invariant under reordering. In particular,
the function ℕ → ℝ
sending n
to (-1)^n / (n+1)
does not have a
sum for this definition, but a series which is absolutely convergent will have the correct sum.
This is based on Mario Carneiro's
infinite sum df-tsms
in Metamath.
For the definition or many statements, α
does not need to be a topological monoid. We only add
this assumption later, for the lemmas where it is relevant.
Equations
- has_sum f a = filter.tendsto (λ (s : finset β), s.sum (λ (b : β), f b)) filter.at_top (nhds a)
∑' i, f i
is the sum of f
it exists, or 0 otherwise
Constant zero function has sum 0
If a function f
vanishes outside of a finite set s
, then it has_sum
∑ b in s, f b
.
A special case of summable.map_iff_of_left_inverse
for convenience
If f : ℕ → α
has sum a
, then the partial sums ∑_{i=0}^{n-1} f i
converge to a
.
If a series f
on β × γ
has sum a
and for each b
the restriction of f
to {b} × γ
has sum g b
, then the series g
has sum a
.
Version of has_sum.update
for add_comm_monoid
rather than add_comm_group
.
Rather than showing that f.update
has a specific sum in terms of has_sum
,
it gives a relationship between the sums of f
and f.update
given that both exist.
Version of has_sum_ite_sub_has_sum
for add_comm_monoid
rather than add_comm_group
.
Rather than showing that the ite
expression has a specific sum in terms of has_sum
,
it gives a relationship between the sums of f
and ite (n = b) 0 (f n)
given that both exist.
Version of tsum_eq_add_tsum_ite
for add_comm_monoid
rather than add_comm_group
.
Requires a different convergence assumption involving function.update
.
You can compute a sum over an encodably type by summing over the natural numbers and taking a supremum. This is useful for outer measures.
tsum_supr_decode₂
specialized to the complete lattice of sets.
Some properties about measure-like functions.
These could also be functions defined on complete sublattices of sets, with the property
that they are countably sub-additive.
R
will probably be instantiated with (≤)
in all applications.
If a function is countably sub-additive then it is sub-additive on countable types
If a function is countably sub-additive then it is sub-additive on finite sets
If a function is countably sub-additive then it is binary sub-additive
Let f : β → α
be a sequence with summable series and let b ∈ β
be an index.
Lemma tsum_eq_add_tsum_ite
writes Σ f n
as the sum of f b
plus the series of the
remaining terms.
Sums on nat #
We show the formula (∑ i in range k, f i) + (∑' i, f (i + k)) = (∑' i, f i)
, in
sum_add_tsum_nat_add
, as well as several results relating sums on ℕ
and ℤ
.
For f : ℕ → α
, then ∑' k, f (k + i)
tends to zero. This does not require a summability
assumption on f
, as otherwise all sums are zero.
If f₀, f₁, f₂, ...
and g₀, g₁, g₂, ...
are both convergent then so is the ℤ
-indexed
sequence: ..., g₂, g₁, g₀, f₀, f₁, f₂, ...
.
The Cauchy criterion for infinite sums, also known as the Cauchy convergence test
The sum over the complement of a finset tends to 0
when the finset grows to cover the whole
space. This does not need a summability assumption, as otherwise all sums are zero.
Series divergence test: if f
is a convergent series, then f x
tends to zero along
cofinite
.
Infinite sums commute with scalar multiplication. Version for scalars living in a monoid
, but
requiring a summability hypothesis.
Infinite sums commute with scalar multiplication. Version for scalars living in a group
, but
not requiring any summability hypothesis.
Infinite sums commute with scalar multiplication. Version for scalars living in a
division_ring
; no summability hypothesis. This could be made to work for a
[group_with_zero γ]
if there was such a thing as distrib_mul_action_with_zero
.
Product and pi types #
Multiplicative opposite #
Interaction with the star #
Given an additive group α
acting on a type β
, and a function f : β → M
,
we automorphize f
to a function β ⧸ α → M
by summing over α
orbits,
b ↦ ∑' (a : α), f(a • b)
.
Equations
- add_action.automorphize f = quotient.lift (λ (b : β), ∑' (a : α), f (a +ᵥ b)) _
Given a group α
acting on a type β
, and a function f : β → M
, we "automorphize" f
to a
function β ⧸ α → M
by summing over α
orbits, b ↦ ∑' (a : α), f(a • b)
.
Equations
- mul_action.automorphize f = quotient.lift (λ (b : β), ∑' (a : α), f (a • b)) _
Automorphization of a function into an R
-module
distributes, that is, commutes with the
R
-scalar multiplication.
Automorphization of a function into an R
-module
distributes, that is, commutes with the
R
-scalar multiplication.
Given a subgroup Γ
of an additive group G
, and a function f : G → M
, we
automorphize f
to a function G ⧸ Γ → M
by summing over Γ
orbits,
g ↦ ∑' (γ : Γ), f(γ • g)
.
Equations
Given a subgroup Γ
of a group G
, and a function f : G → M
, we "automorphize" f
to a
function G ⧸ Γ → M
by summing over Γ
orbits, g ↦ ∑' (γ : Γ), f(γ • g)
.
Equations
Automorphization of a function into an R
-module
distributes, that is, commutes with the
R
-scalar multiplication.
Automorphization of a function into an R
-module
distributes, that is, commutes with the R
-scalar multiplication.