Chapter 01 — Background

Proofs omitted. Hover over highlighted terms to see their Lean types.

Text 1.0.1

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Pointwise addition of two subsets of an additive commutative group is the set of all pairwise sums.

open Set open scoped Pointwise universe u variable {X : Type u} [AddCommGroup X] theorem pointwise_add_eq_setOf_add (C D : Set X) : C + D = {x | ∃ c ∈ C, ∃ d ∈ D, c + d = x}

Text 1.0.2

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.2 (1): A subset of a real vector space is a cone when it equals its set of positive real multiples.

open Set open scoped Pointwise universe u variable {X : Type u} [AddCommGroup X] [Module ℝ X] def IsCone (C : Set X) : Prop := C = (Ioi (0 : ℝ) : Set ℝ) • C

Text 1.0.3

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.3: a subset of a real vector space is the nonempty carrier of a real affine subspace exactly when it is nonempty and equal to its affine span.

open scoped Pointwise universe u namespace Set variable {X : Type u} [AddCommGroup X] [Module ℝ X] theorem exists_nonempty_affineSubspace_iff_nonempty_eq_affineSpan (C : Set X) : (∃ S : AffineSubspace ℝ X, (S : Set X) = C ∧ (S : Set X).Nonempty) ↔ C.Nonempty ∧ (affineSpan ℝ C : Set X) = C

Text 1.0.4

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.4 (1): the span of C is the intersection of all linear subspaces of X containing C.

open Set universe u variable {X : Type u} [AddCommGroup X] [Module ℝ X] theorem span_eq_sInf_submodule (C : Set X) : Submodule.span ℝ C = sInf {S : Submodule ℝ X | C ⊆ S}

Text 1.0.5

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.5: if C is a nonempty affine subspace of a real vector space, then the textbook difference set C - C is the carrier of the canonical direction subspace C.direction, and C is the translate of this direction through any point of C.

universe u open Set open AffineSubspace variable {X : Type u} [AddCommGroup X] [Module ℝ X] theorem text_1_0_5 (C : AffineSubspace ℝ X) (hC : (C : Set X).Nonempty) : (C.direction : Set X) = (C : Set X) -ᵥ (C : Set X) ∧ ∀ ⦃x : X⦄, x ∈ C → (C : Set X) = (fun v ↦ v +ᵥ x) '' (C.direction : Set X)

Text 1.0.6

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.6 (1): The closed line segment [x,y] in a real vector space is mathlib's segment ℝ x y.

open Set universe u variable {X : Type u} [AddCommGroup X] [Module ℝ X] recall segment

Text 1.0.7

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.7: the half-open segment ]x,y] coincides with the half-open segment [y,x[.

universe u open Set variable {X : Type u} [AddCommGroup X] [Module ℝ X] theorem text_1_0_7 (x y : X) : openClosedSegment x y = closedOpenSegment y x

Text 1.0.8

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.8: a set-valued operator from X to Y is canonically just a map X → Set Y.

universe u v variable (X : Type u) (Y : Type v) #check X → Set Y

Text 1.0.9

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.9: The image A(C) of a set C under a set-valued operator A is the union of the value sets A x for x ∈ C.

universe u v w namespace SetValuedOperator variable {X : Type u} {Y : Type v} {Z : Type w} def image (A : SetValuedOperator X Y) (C : Set X) : Set Y := ⋃ x ∈ C, A x

Text 1.0.10

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.10: The domain of a set-valued operator A is the set of points x for which A x is nonempty.

universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} def dom (A : SetValuedOperator X Y) : Set X := { x | (A x).Nonempty }

Text 1.0.11

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.11: The inverse of a set-valued operator is the operator obtained by reversing its graph, equivalently by sending u : Y to the set of x : X such that u ∈ A x.

universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} def inverse (A : SetValuedOperator X Y) : SetValuedOperator Y X := fun u ↦ { x | u ∈ A x }

Text 1.0.12

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.12 (1): the domain of the inverse of a set-valued operator is the range of the original operator.

universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} @[simp] theorem dom_inverse (A : SetValuedOperator X Y) : A.inverse.dom = A.range

Text 1.0.13

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.13: a set-valued operator is at most single-valued when each value set has at most one element, equivalently when every value over its domain is a singleton.

universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} def IsAtMostSingleValued (A : SetValuedOperator X Y) : Prop := ∀ x, (A x).Subsingleton

Text 1.0.14

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.14: a selection of A is an operator T : A.dom → Y such that T x ∈ A x for every x ∈ A.dom.

universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} variable {A : SetValuedOperator X Y} abbrev Selection (A : SetValuedOperator X Y) : Type (max u v) := ∀ x : A.dom, A x

Text 1.0.15

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.15: the image T(C) of a subset C ⊆ X under a single-valued map T : X → Y is the canonical set-theoretic construction Set.image T C.

universe u v variable {X : Type u} {Y : Type v} theorem image_eq_set_image (T : X → Y) (C : Set X) : T '' C = Set.image T C

Text 1.0.16

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.16: for set-valued operators A B : X → Set Y and a real scalar c (λ in the text), the linear combination A + λ B is the pointwise operator satisfying (A + c • B) x = A x + c • B x.

open scoped Pointwise universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} [AddCommGroup Y] [Module ℝ Y] @[simp] theorem add_smul_apply (A B : SetValuedOperator X Y) (c : ℝ) (x : X) : (A + c • B) x = A x + c • B x

Text 1.0.17

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.17 (1): the graph of the pointwise linear combination A + λ B consists of the pairs (x, u + λ • v) coming from matching graph points of A and B.

open scoped Pointwise universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} [AddCommGroup Y] [Module ℝ Y] theorem graph_add_smul (A B : SetValuedOperator X Y) (r : ℝ) : (A + r • B).graph = { xw | ∃ u v, (xw.1, u) ∈ A.graph ∧ (xw.1, v) ∈ B.graph ∧ xw.2 = u + r • v }

Text 1.0.18

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.18: an operator between real scalar-action spaces is positively homogeneous when it commutes with scalar multiplication by every positive real scalar.

universe u v def PositivelyHomogeneous {X : Type u} {Y : Type v} [SMul ℝ X] [SMul ℝ Y] (T : X → Y) : Prop := ∀ ⦃a : ℝ⦄, 0 < a → ∀ x : X, T (a • x) = a • T x

Text 1.0.19

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.19: affine mappings between real vector spaces are the canonical affine maps X →ᵃ[ℝ] Y.

universe u v theorem affineMap_notation_eq (X : Type u) (Y : Type v) [AddCommGroup X] [Module ℝ X] [AddCommGroup Y] [Module ℝ Y] : (X →ᵃ[ℝ] Y) = AffineMap ℝ X Y

Text 1.0.20

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.20: a map between real vector spaces is affine, in the canonical sense of arising from an element of X →ᵃ[ℝ] Y, exactly when its translate through the origin, x ↦ T x - T 0, is linear.

universe u v variable {X : Type u} {Y : Type v} [AddCommGroup X] [Module ℝ X] [AddCommGroup Y] [Module ℝ Y] theorem affine_iff_isLinearMap_sub_apply_zero (T : X → Y) : (∃ A : X →ᵃ[ℝ] Y, (A : X → Y) = T) ↔ IsLinearMap ℝ (fun x : X ↦ T x - T 0)

Text 1.0.21

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.21: The translation of a set-valued operator A by y sends x to the value of A at x - y.

universe u v namespace SetValuedOperator variable {X : Type u} {Y : Type v} [AddGroup X] def translate (A : SetValuedOperator X Y) (y : X) : SetValuedOperator X Y := fun x ↦ A (x - y)

Text 1.0.22

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.22: the textbook notion of a directed set is exactly a nonempty type equipped with a reflexive, transitive, and directed relation, equivalently a nonempty type with a directed preorder relation.

universe u theorem isDirectedSet_iff {α : Type u} {r : α → α → Prop} : (Nonempty α ∧ Std.Refl r ∧ IsTrans α r ∧ IsDirected α r) ↔ Nonempty α ∧ IsPreorder α r ∧ IsDirected α r

Text 1.0.23

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.23: once the directed-set structure on the index type is fixed in the ambient context, a net in X indexed by A is formalized by the canonical function type A → X. The textbook's nonemptiness assumption on X does not affect this underlying data.

universe u v variable (A : Type u) (X : Type v) #check A → X

Text 1.0.24

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

A net is eventually in Y exactly when all sufficiently large indices map into Y.

universe u v namespace Net variable {A : Type u} {X : Type v} [Preorder A] [IsDirectedOrder A] [Nonempty A] theorem eventuallyIn_iff_exists_forall_ge_mem (x : A → X) (Y : Set X) : (∀ᶠ a in Filter.atTop, x a ∈ Y) ↔ ∃ a : A, ∀ b ≥ a, x b ∈ Y

Text 1.0.25

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.25: a net y is a subnet of a net x via k exactly when the textbook reindexing-and-cofinality condition holds, equivalently when y = x ∘ k and k tends to atTop.

open Filter universe u v w namespace Net variable {X : Type u} {A : Type v} {B : Type w} theorem isSubnetOfVia_iff [Preorder A] [Preorder B] [IsDirectedOrder A] [IsDirectedOrder B] [Nonempty B] {x : A → X} {y : B → X} {k : B → A} : (y = x ∘ k ∧ ∀ a : A, ∃ d : B, ∀ b : B, d ≤ b → a ≤ k b) ↔ y = x ∘ k ∧ Tendsto k atTop atTop

Text 1.0.26

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.26: the extended real line is formalized by the canonical type EReal of real numbers with adjoined bottom and top elements ⊥ and ⊤, equipped with the extended order; the textbook's conventions about indeterminate arithmetic expressions are handled separately when needed.

recall EReal : Type

Text 1.0.27

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.27: the extended interval ]ξ, +∞] is the usual real interval ]ξ, +∞[, embedded into EReal, with the point +∞ adjoined.

open Set namespace EReal theorem openClosedUpperInfiniteInterval_eq_image_Ioi_union_top (ξ : ℝ) : Ioi (ξ : EReal) = Real.toEReal '' Ioi ξ ∪ ({(⊤ : EReal)} : Set EReal)

Text 1.0.28

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.28: for a set S of extended reals, the canonical complete-lattice infimum sInf S is its infimum, i.e. the greatest lower bound of S; this is exactly the EReal specialization of the standard theorem isGLB_sInf.

recall isGLB_sInf {α : Type*} [CompleteSemilatticeInf α] (s : Set α) : IsGLB s (sInf s)

Text 1.0.29

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.29: for an extended-real-valued net indexed by a nonempty directed set, the limit inferior is the supremum of the infima of the tails and the limit superior is the infimum of the suprema of the tails.

open Filter universe u namespace Net variable {A : Type u} [Preorder A] [IsDirectedOrder A] theorem tail_liminf_limsup_formulas [Nonempty A] (ξ : A → EReal) : Filter.liminf ξ atTop = sSup (Set.range fun a : A ↦ sInf (ξ '' Set.Ici a)) ∧ Filter.limsup ξ atTop = sInf (Set.range fun a : A ↦ sSup (ξ '' Set.Ici a))

Text 1.0.30

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.30: for an EReal-valued net on a nonempty directed index set, the lower limit is bounded above by the upper limit.

open Filter universe u namespace Net variable {A : Type u} [Nonempty A] [Preorder A] [IsDirectedOrder A] theorem liminf_le_limsup (ξ : A → EReal) : liminf ξ atTop ≤ limsup ξ atTop

Text 1.0.31

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.31: the effective domain of an extended-real-valued function is the set of points at which the function takes a finite value. Equivalently, it is the preimage of the open interval Ioo ⊥ ⊤, i.e. the points where the value is neither +∞ nor -∞.

universe u namespace ERealFunction variable {X : Type u} def effectiveDom (f : X → EReal) : Set X := f ⁻¹' Set.Ioo (⊥ : EReal) ⊤

Text 1.0.32

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.32: addition on EReal with the textbook convention that the indeterminate sums +∞ + -∞ and -∞ + +∞ are assigned the value +∞.

universe u namespace EReal variable {X : Type u} def sumWithTopBotAsTop : EReal → EReal → EReal

Text 1.0.33

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.33: the effective domain of the textbook pointwise sum of two extended-real-valued functions is the intersection of their effective domains.

universe u open EReal namespace ERealFunction variable {X : Type u} theorem effectiveDom_pointwiseSum_eq_inter (f g : X → EReal) : effectiveDom (pointwiseSum f g) = effectiveDom f ∩ effectiveDom g

Text 1.0.34

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.34: a topology on a type X is the canonical structure TopologicalSpace X, encoding the empty set, the whole space, arbitrary unions, and finite intersections of open sets; a topological space is therefore a type equipped with an instance of TopologicalSpace X, and the associated notions of open and closed sets are IsOpen and IsClosed.

universe u recall TopologicalSpace (X : Type u) : Type u

Text 1.0.35

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.35: for a topological space X and a point x : X, the family 𝓥(x) of neighborhoods of x is formalized by the neighborhood filter nhds x.

universe u recall nhds {X : Type u} [TopologicalSpace X] (x : X) : Filter X

Text 1.0.36

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.36: a base of a topology on X is formalized by the canonical predicate TopologicalSpace.IsTopologicalBasis B, expressing that B is a family of open sets whose members locally refine every neighborhood.

universe u recall TopologicalSpace.IsTopologicalBasis {X : Type u} [TopologicalSpace X] (B : Set (Set X)) : Prop

Text 1.0.37

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.37: if B is a basis for the topology on X, then every open set is the union of the basis sets contained in it. This is the canonical theorem TopologicalSpace.IsTopologicalBasis.open_eq_sUnion'.

universe u namespace TopologicalSpace variable {X : Type u} [TopologicalSpace X] recall IsTopologicalBasis.open_eq_sUnion'

Text 1.0.38

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.38: equivalently, a point lies in interior C exactly when C contains a neighborhood of that point.

open Filter Topology universe u theorem mem_interior_iff_exists_mem_nhds_subset {X : Type u} [TopologicalSpace X] {x : X} {C : Set X} : x ∈ interior C ↔ ∃ V : Set X, V ∈ 𝓝 x ∧ V ⊆ C

Text 1.0.39

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.39: for a subset C of a topological space, the closure of C is the canonical set closure C, namely the smallest closed set containing C.

universe u recall closure {X : Type u} [TopologicalSpace X] (C : Set X) : Set X

Text 1.0.40

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.40: the boundary bdry C of a subset C of a topological space is formalized by the canonical set frontier C.

universe u recall frontier {X : Type u} [TopologicalSpace X] (C : Set X) : Set X

Text 1.0.41

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.41: if B₁ and B₂ are bases for the given topologies on X₁ and X₂, then the family of rectangles U ×ˢ V with U ∈ B₁ and V ∈ B₂ is a basis for the product topology on X₁ × X₂. This is the textbook set-builder form of the canonical theorem TopologicalSpace.IsTopologicalBasis.prod.

universe u v namespace TopologicalSpace variable {X₁ : Type u} [TopologicalSpace X₁] variable {X₂ : Type v} [TopologicalSpace X₂] theorem IsTopologicalBasis.rectangles {B₁ : Set (Set X₁)} {B₂ : Set (Set X₂)} (h₁ : IsTopologicalBasis B₁) (h₂ : IsTopologicalBasis B₂) : IsTopologicalBasis {s : Set (X₁ × X₂) | ∃ U ∈ B₁, ∃ V ∈ B₂, s = U ×ˢ V}

Text 1.0.42

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.42: a Hausdorff topological space is formalized by the canonical predicate T2Space X, expressing that distinct points admit disjoint neighborhoods.

universe u recall T2Space {X : Type u} [TopologicalSpace X] : Prop

Text 1.0.43

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.43: in the Hausdorff setting of the text, compactness of a subset C is formalized by the canonical predicate IsCompact C.

universe u recall IsCompact {X : Type u} [TopologicalSpace X] (C : Set X) : Prop

Text 1.0.44

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.44: a net ξ converges to x exactly when it is eventually contained in every neighborhood of x; in mathlib this convergence is expressed by Tendsto ξ atTop (nhds x).

universe u v open Filter variable {A : Type u} [Preorder A] [IsDirectedOrder A] [Nonempty A] variable {X : Type v} [TopologicalSpace X] theorem tendsto_atTop_nhds_iff_forall_exists_forall_ge_mem (ξ : A → X) (x : X) : Tendsto ξ atTop (nhds x) ↔ ∀ V ∈ nhds x, ∃ b, ∀ a ≥ b, ξ a ∈ V

Text 1.0.45

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.45: for a net ξ : A → X, the canonical predicate MapClusterPt x Filter.atTop ξ means that every neighborhood of x is visited arbitrarily far out in the directed index set.

universe u v variable {A : Type u} [Nonempty A] [Preorder A] [IsDirectedOrder A] variable {X : Type v} [TopologicalSpace X] theorem mapClusterPt_atTop_iff_forall_forall_exists_ge_mem_nhds (ξ : A → X) (x : X) : MapClusterPt x Filter.atTop ξ ↔ ∀ V : Set X, V ∈ nhds x → ∀ b : A, ∃ a : A, b ≤ a ∧ ξ a ∈ V

Text 1.0.46

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.46: A point x is a sequential cluster point of a sequence u if some subsequence of u converges to x.

open Filter open scoped Topology universe u variable {X : Type u} [TopologicalSpace X] def IsSequentialClusterPt (u : ℕ → X) (x : X) : Prop := ∃ φ : ℕ → ℕ, StrictMono φ ∧ Tendsto (u ∘ φ) atTop (𝓝 x)

Text 1.0.47

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.47: a point x is a cluster point of a net u iff there is a subnet, given by a monotone cofinal reindexing map, whose reindexed net converges to x.

universe u v w open Filter theorem mapClusterPt_atTop_iff_exists_subnet_tendsto {X : Type u} [TopologicalSpace X] {A : Type v} [Nonempty A] [Preorder A] [IsDirectedOrder A] {u : A → X} {x : X} : MapClusterPt x atTop u ↔ ∃ (B : Type (max u v)) (_ : Nonempty B) (_ : Preorder B) (_ : IsDirectedOrder B) (φ : B → A), Monotone φ ∧ Tendsto φ atTop atTop ∧ Tendsto (u ∘ φ) atTop (nhds x)

Text 1.0.48

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.48: the usual topology on ℝ has as a basis the family of open intervals (a, b) with a < b.

open Set TopologicalSpace theorem real_isTopologicalBasis_Ioo : TopologicalSpace.IsTopologicalBasis { s : Set ℝ | ∃ a b, a < b ∧ s = Ioo a b }

Text 1.0.49

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.49: the canonical topology on the extended real line EReal has as a basis the family of all embedded real open intervals (a, b) together with the lower rays [-∞, ξ) and the upper rays (ξ, +∞], represented in Lean by Ioo (a : EReal) (b : EReal), Iio (ξ : EReal), and Ioi (ξ : EReal).

open Set open scoped Topology namespace EReal theorem realIntervalRayBasis_isTopologicalBasis : TopologicalSpace.IsTopologicalBasis realIntervalRayBasis

Text 1.0.50

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.50 (1): the real line with its usual metric topology is Hausdorff.

theorem real_t2Space : T2Space ℝ

Text 1.0.51

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.51: the extended real line EReal = [-∞, +∞], equipped with its canonical order topology, is compact. The textbook description by real intervals together with rays at ⊥ and ⊤ describes this same topology.

theorem extendedReal_compactSpace : CompactSpace EReal

Text 1.0.53.1.33

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.53 (1.33): the source’s notion of lower semicontinuity at a point is the canonical predicate LowerSemicontinuousAt; the textbook’s extended-real-valued case is a specialization of this mathlib definition.

open Filter open scoped Topology universe u v recall LowerSemicontinuousAt {X : Type u} {Y : Type v} [TopologicalSpace X] [Preorder Y] (f : X → Y) (x : X) : Prop

Text 1.0.54

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.54 (1): an extended-real-valued function is lower semicontinuous on the whole space exactly when it is lower semicontinuous at every point.

universe u v open Filter open scoped Topology theorem lowerSemicontinuous_iff_forall_lowerSemicontinuousAt {X : Type u} [TopologicalSpace X] {f : X → EReal} : LowerSemicontinuous f ↔ ∀ x, LowerSemicontinuousAt f x

Text 1.0.55

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.55: the points where an EReal-valued function takes a real value and is continuous are exactly the points in its effective domain at which it is continuous.

universe u namespace ERealFunction variable {X : Type u} [TopologicalSpace X] theorem mem_effectiveDom_inter_continuousAt_iff_exists_real (f : X → EReal) (x : X) : x ∈ effectiveDom f ∩ {y | ContinuousAt f y} ↔ (∃ r : ℝ, f x = (r : EReal)) ∧ ContinuousAt f x

Text 1.0.56.1.36

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.56 (1.36): the limit inferior of an extended-real-valued function f at x is the filter liminf of f along the neighborhood filter nhds x.

open Filter universe u namespace ERealFunction variable {X : Type u} [TopologicalSpace X] noncomputable abbrev liminfAt (f : X → EReal) (x : X) : EReal := liminf f (nhds x)

Text 1.0.57

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.57: for a topological space X and a subset C ⊆ X, the extended-real indicator ν_C, equal to 0 on C and +∞ on Cᶜ, is lower semicontinuous if and only if C is closed.

universe u open Set theorem lowerSemicontinuous_indicator_compl_top_iff_isClosed {X : Type u} [TopologicalSpace X] (C : Set X) : LowerSemicontinuous (indicator Cᶜ fun _ : X ↦ (⊤ : EReal)) ↔ IsClosed C

Text 1.0.58

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.58: for a subset C of a Hausdorff topological space, being sequentially closed is formalized by the canonical predicate IsSeqClosed C, meaning that every convergent sequence in C has its limit in C. The separation assumption is not needed for this definition.

universe u open Filter open scoped Topology recall IsSeqClosed

Text 1.0.59

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.59: if C is a closed subset of a topological space X, then C is sequentially closed; equivalently, every sequence in C converging to x has limit x ∈ C.

recall IsClosed.isSeqClosed

Text 1.0.60

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.60: a topological space is sequential exactly when every sequentially closed subset is closed, so closedness and sequential closedness coincide. This is the canonical class SequentialSpace.

recall SequentialSpace

Text 1.0.61

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.61: for an extended-real-valued function on a sequential topological space, lower semicontinuity is equivalent to sequential lower semicontinuity, meaning that every convergent sequence satisfies the liminf inequality at its limit.

universe u open Filter variable {X : Type u} [TopologicalSpace X] [SequentialSpace X] theorem lowerSemicontinuous_iff_forall_seq_tendsto_le_liminf (f : X → EReal) : LowerSemicontinuous f ↔ ∀ ⦃x : X⦄ ⦃u : ℕ → X⦄, Tendsto u atTop (nhds x) → f x ≤ liminf (f ∘ u) atTop

Text 1.0.62

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.62: a topological space is sequential if and only if every map from it to an arbitrary topological space is continuous exactly when it is sequentially continuous.

universe u open Filter variable {X : Type u} [TopologicalSpace X] theorem sequentialSpace_iff_forall_continuous_iff_seqContinuous : SequentialSpace X ↔ ∀ ⦃Y : Type*⦄ [TopologicalSpace Y] (f : X → Y), Continuous f ↔ SeqContinuous f

Text 1.0.63

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.63: for a subset C of a metric space X, the diameter is the canonical extended diameter Metric.ediam C, and the distance function to C, with value ∞ when C = ∅, is the canonical extended infimum-edistance map fun x ↦ Metric.infEDist x C.

universe u recall Metric.ediam {α : Type u} [PseudoEMetricSpace α] (s : Set α) : ENNReal

Text 1.0.64

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.64: in a metric space, the open ball of center x and positive radius ρ is the canonical set Metric.ball x ρ; the closed ball and the open-ball description of the induced metric topology are recalled by the companion canonical API below.

universe u recall Metric.ball {α : Type u} [PseudoMetricSpace α] (x : α) (ε : ℝ) : Set α

Text 1.0.65

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.65: a topological space is metrizable when there exists a metric on the space whose induced topology coincides with the given topology, formalized by the canonical predicate TopologicalSpace.MetrizableSpace.

universe u recall TopologicalSpace.MetrizableSpace (X : Type u) [TopologicalSpace X] : Prop

Text 1.0.66

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.66: in a metric space, convergence of a sequence to a point is expressed by the canonical filter-theoretic notion Tendsto u atTop (nhds x).

universe u v open Filter variable {X : Type u} [MetricSpace X] recall Tendsto

Text 1.0.67

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.67: a sequence in a metric space is Cauchy when it is formalized by the canonical predicate CauchySeq; the usual metric-space distance criterion and the textbook sequential formulation of completeness are recorded below as companion statements.

universe u v open Filter variable {X : Type u} [MetricSpace X] recall CauchySeq {α : Type u} {β : Type v} [UniformSpace α] [Preorder β] (u : β → α) : Prop

Text 1.0.68

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.68: a subset of a topological space, and hence in particular of a Hausdorff space, is called a G_δ set when it is a countable intersection of open sets, formalized by the canonical predicate IsGδ.

universe u recall IsGδ {X : Type u} [TopologicalSpace X] (s : Set X) : Prop

Text 1.0.69

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Text 1.0.69: for an operator T : X → X, the textbook fixed point set Fix T is the canonical mathlib set Function.fixedPoints T.

universe u recall Function.fixedPoints {α : Type u} (T : α → α) : Set α

Fact 1.1

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.1 (Zorn's lemma): if every chain in a preorder has an upper bound, then there exists a maximal element.

universe u theorem fact_1_1 {α : Type u} [Preorder α] (h : ∀ c : Set α, IsChain (fun a b : α ↦ a ≤ b) c → BddAbove c) : ∃ m : α, IsMax m

Remark 1.3

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Remark 1.3 (1): the textbook ray [0,+∞[ is the closed ray Set.Ici 0.

open Set theorem positiveReals_eq_Ici : {ξ : ℝ | 0 ≤ ξ} = Ici 0

Definition 1.4

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.4: an extended-real-valued function is proper when it never attains -∞ and its domain is nonempty.

universe u namespace ERealFunction variable {X : Type u} def IsProper (f : X → EReal) : Prop := (∀ x, f x ≠ ⊥) ∧ (dom f).Nonempty

Lemma 1.6

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.6 (1): the epigraph of the pointwise supremum of a family of extended-real-valued functions is the intersection of the epigraphs of the family members.

open Set universe u v namespace ERealFunction -- Proof sketch: extensionality on pairs `(x, ξ)` reduces the claim to the order fact -- `iSup_le_iff`, which says `⨆ i, f i x ≤ ξ` iff `∀ i, f i x ≤ ξ`. theorem epigraph_iSup {X : Type u} {I : Type v} (f : I → X → EReal) : epigraph (⨆ i, f i) = ⋂ i, epigraph (f i)

Definition 1.7

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.7 (1): the infimum of an extended-real-valued function over a set is the infimum of its image on that set, namely sInf (f '' C).

open Set universe u namespace ERealFunction variable {X : Type u} theorem isGLB_sInf_image (f : X → EReal) (C : Set X) : IsGLB (f '' C) (sInf (f '' C))

Definition 1.8

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.8: a sequence in the domain of an extended-real-valued function is minimizing when its function values converge to the infimum of the image of the function.

open Filter universe u namespace ERealFunction variable {X : Type u} def IsMinimizingSequence (f : X → EReal) (x : ℕ → X) : Prop := (∀ n : ℕ, x n ∈ dom f) ∧ Tendsto (f ∘ x) atTop (nhds (sInf (Set.range f)))

Fact 1.9

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.9 in textbook subnet form: a monotone cofinal reindexing preserves the limit of a convergent net.

open Filter universe u v w variable {A : Type u} {B : Type v} {X : Type w} variable [Preorder A] [Preorder B] [TopologicalSpace X] variable {x : A → X} {k : B → A} {x₀ : X} theorem subnet_tendsto_same_limit_of_monotone_cofinal (hx : Tendsto x atTop (nhds x₀)) (hk_mono : Monotone k) (hk_cofinal : ∀ a₀ : A, ∃ b₀ : B, ∀ b : B, b₀ ≤ b → a₀ ≤ k b) : Tendsto (x ∘ k) atTop (nhds x₀)

Lemma 1.10

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.10: a point belongs to the closure of a subset C exactly when there exists a nonempty directed net with values in C that converges to that point.

universe u open Filter theorem mem_closure_iff_exists_net_tendsto {X : Type u} [TopologicalSpace X] {C : Set X} {x : X} : x ∈ closure C ↔ ∃ (A : Type u) (_ : Preorder A) (_ : IsDirectedOrder A) (_ : Nonempty A) (ξ : A → C), Tendsto (fun a ↦ (ξ a : X)) atTop (nhds x)

Fact 1.11

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.11: for a subset C of a Hausdorff topological space, compactness, the closed-set finite-intersection property relative to C, the existence of cluster points for all nets in C, and the existence of convergent subnets for all nets in C are equivalent.

open Filter open scoped Topology universe u v w variable {X : Type u} [TopologicalSpace X] theorem isCompact_tfae [T2Space X] (C : Set X) : List.TFAE [ IsCompact C, ClosedSetFiniteIntersectionPropertyIn C, EveryNetHasClusterPointIn C, EveryNetHasConvergentSubnetIn C ]

Lemma 1.12

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.12: a compact subset of a Hausdorff topological space is closed.

universe u theorem isClosed_of_isCompact {X : Type u} [TopologicalSpace X] [T2Space X] {C : Set X} : IsCompact C → IsClosed C

Lemma 1.14

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.14: if a net in a compact subset of a Hausdorff space has x as its unique cluster point, then the net converges to x.

universe u v open Filter theorem tendsto_of_unique_cluster_point_on_compact {X : Type u} [TopologicalSpace X] [T2Space X] {A : Type v} [Preorder A] [IsDirectedOrder A] {C : Set X} (hC : IsCompact C) {ξ : A → X} (hξC : ∀ a, ξ a ∈ C) {x : X} (_hx : MapClusterPt x atTop ξ) (hunique : ∀ y : X, MapClusterPt y atTop ξ → y = x) : Tendsto ξ atTop (nhds x)

Fact 1.15

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.15 (1): the net of tail infima of an extended-real-valued net converges to its limit inferior.

open Filter universe u namespace Net variable {A : Type u} [Nonempty A] [Preorder A] [IsDirectedOrder A] theorem tendsto_tailInf_liminf (ξ : A → EReal) : Tendsto (fun a ↦ sInf (ξ '' Set.Ici a)) atTop (nhds (Filter.liminf ξ atTop))

Lemma 1.16

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.16: if two extended-real-valued nets have lower limits strictly above -∞, then the sum of their lower limits is bounded above by the lower limit of their pointwise sum.

universe u open Filter namespace Net variable {A : Type u} [Preorder A] theorem le_liminf_add [IsDirectedOrder A] (ξ η : A → EReal) (_hξ : (⊥ : EReal) < liminf ξ atTop) (_hη : (⊥ : EReal) < liminf η atTop) : liminf ξ atTop + liminf η atTop ≤ liminf (ξ + η) atTop

Definition 1.17

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.17: continuity of a map f : X → Y at a point x : X is formalized by the canonical predicate ContinuousAt f x; global continuity is formalized by Continuous f.

recall ContinuousAt

Fact 1.18

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.18: the basis criterion for continuity is already the canonical mathlib theorem TopologicalSpace.IsTopologicalBasis.continuous_iff. It states that a map between topological spaces is continuous exactly when the preimage of every basis element of the codomain is open.

universe u v recall TopologicalSpace.IsTopologicalBasis.continuous_iff

Fact 1.19

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.19: a map T : X → Y is continuous at x exactly when it sends every net converging to x to a net converging to T x.

universe u v w open Filter theorem continuousAt_iff_tendsto_nets {X : Type u} {Y : Type v} [TopologicalSpace X] [TopologicalSpace Y] {T : X → Y} {x : X} : ContinuousAt T x ↔ ∀ {A : Type (max u w)} [Preorder A] [IsDirectedOrder A] (ξ : A → X), Tendsto ξ atTop (nhds x) → Tendsto (T ∘ ξ) atTop (nhds (T x))

Lemma 1.20

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.20: if X and Y are Hausdorff spaces, T : X → Y is continuous, and C ⊆ X is compact, then T(C) is compact. This is the textbook Hausdorff-space specialization of the canonical theorem IsCompact.image.

universe u v theorem lemma_1_20 {X : Type u} {Y : Type v} [TopologicalSpace X] [TopologicalSpace Y] [T2Space X] [T2Space Y] {C : Set X} {T : X → Y} (hC : IsCompact C) (hT : Continuous T) : IsCompact (T '' C)

Definition 1.21

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.21: for an extended-real-valued function f : X → EReal, lower semicontinuity at a point x is the canonical predicate LowerSemicontinuousAt f x, and lower semicontinuity on X is the canonical predicate LowerSemicontinuous f.

recall LowerSemicontinuousAt

Lemma 1.23

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.23: the limit inferior of f at x is the minimum of the limit inferiors of f along all nets converging to x.

open Filter ERealFunction universe u v variable {X : Type u} [TopologicalSpace X] {x : X} theorem liminfAt_isLeast_netLiminfValues {X : Type u} [TopologicalSpace X] (f : X → EReal) (x : X) : IsLeast (netLiminfValuesAt.{u, v} f x) (liminfAt f x)

Lemma 1.24

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.24: an extended-real-valued function on a topological space is lower semicontinuous if and only if each of its real lower level sets is closed.

universe u namespace ERealFunction variable {X : Type u} variable [TopologicalSpace X] theorem lowerSemicontinuous_iff_isClosed_lowerLevelSet (f : X → EReal) : LowerSemicontinuous f ↔ ∀ ξ : ℝ, IsClosed (lowerLevelSet f ξ)

Lemma 1.26

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.26 (1): the extended-real statement in the text is the EReal specialization of the canonical mathlib theorem lowerSemicontinuous_iSup.

universe u v variable {X : Type u} [TopologicalSpace X] recall lowerSemicontinuous_iSup

Lemma 1.27

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.27: a finite positive weighted sum of lower semicontinuous extended-real-valued functions is lower semicontinuous. This formulation is valid on any topological space, hence in particular on a Hausdorff space.

universe u v open scoped BigOperators variable {X : Type u} [TopologicalSpace X] theorem lowerSemicontinuous_weighted_sum {ι : Type v} [Finite ι] {f : ι → X → EReal} {α : ι → ℝ} (hα : ∀ i, 0 < α i) (hf : ∀ i, LowerSemicontinuous (f i)) : let _ : Fintype ι := Fintype.ofFinite ι LowerSemicontinuous (fun x ↦ ∑ i, ((α i : EReal) * f i x))

Lemma 1.28

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.28: composing a lower semicontinuous extended-real-valued function with a continuous map preserves lower semicontinuity.

universe u v namespace ERealFunction variable {X : Type u} {Y : Type v} [TopologicalSpace X] [TopologicalSpace Y] theorem lowerSemicontinuous_comp {f : Y → EReal} {g : X → Y} (hf : LowerSemicontinuous f) (hg : Continuous g) : LowerSemicontinuous (f ∘ g)

Theorem 1.29

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Theorem 1.29: a lower semicontinuous extended-real-valued function on a Hausdorff space attains a minimum on a compact set as soon as that compact set meets the domain of the function.

universe u namespace ERealFunction theorem lowerSemicontinuous_exists_isMinOn_of_isCompact {X : Type u} [TopologicalSpace X] [T2Space X] {f : X → EReal} {C : Set X} (hf : LowerSemicontinuous f) (hC : IsCompact C) (hdom : (C ∩ { x | f x < ⊤ }).Nonempty) : ∃ x ∈ C, IsMinOn f C x

Lemma 1.30

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.30: over a nonempty compact second factor, the infimum of a lower semicontinuous extended-real-valued function over the second variable is lower semicontinuous, and each fiber section attains its minimum.

universe u v theorem lowerSemicontinuous_compactRightInfimum_and_exists_isMinOn {X : Type u} {Y : Type v} [TopologicalSpace X] [TopologicalSpace Y] [CompactSpace Y] [Nonempty Y] {φ : X × Y → EReal} (hφ : LowerSemicontinuous φ) : LowerSemicontinuous (fun x ↦ sInf (Set.range fun y : Y ↦ φ (x, y))) ∧ ∀ x : X, ∃ y : Y, IsMinOn (fun y' : Y ↦ φ (x, y')) Set.univ y

Definition 1.31

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.31: the lower semicontinuous envelope of f is the supremum, in the function lattice, of all lower semicontinuous extended-real-valued minorants of f.

universe u open Set namespace ERealFunction variable {X : Type u} [TopologicalSpace X] noncomputable def lowerSemicontinuousEnvelope (f : X → EReal) : X → EReal := sSup (lowerSemicontinuousMinorants f)

Lemma 1.32

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.32: the lower semicontinuous envelope is the greatest lower semicontinuous minorant of f.

open Set universe u namespace ERealFunction variable {X : Type u} variable [TopologicalSpace X] theorem lowerSemicontinuousHull_isGreatest (f : X → EReal) : IsGreatest {g : X → EReal | LowerSemicontinuous g ∧ g ≤ f} (lowerSemicontinuousEnvelope f)

Definition 1.33

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.33: sequential compactness of a subset C of a topological space is formalized by the canonical predicate IsSeqCompact C. In the Hausdorff setting of the textbook, this is exactly the statement that every sequence in C admits a subsequence, indexed by a strictly increasing map, that converges to a point of C.

universe u recall IsSeqCompact {X : Type u} [TopologicalSpace X] (C : Set X) : Prop

Lemma 1.34

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.34 (1): a sequentially compact subset of a Hausdorff space is sequentially closed.

universe u variable {X : Type u} [TopologicalSpace X] [T2Space X] variable {C : Set X} theorem IsSeqCompact.isSeqClosed (hC : IsSeqCompact C) : IsSeqClosed C

Lemma 1.35

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.35: if a sequence in a sequentially compact subset has a unique sequential cluster point, expressed by the canonical predicate IsSequentialClusterPt, then the whole sequence converges to that point.

open Filter open scoped Topology universe u variable {X : Type u} [TopologicalSpace X] theorem tendsto_of_unique_sequential_cluster_point {C : Set X} (hC : IsSeqCompact C) {u : ℕ → X} (huC : ∀ n, u n ∈ C) {x : X} (hunique : ∀ y : X, IsSequentialClusterPt u y → y = x) : Tendsto u atTop (nhds x)

Lemma 1.36

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.36: for an extended-real-valued function on a topological space, sequential lower semicontinuity, sequential closedness of the real-height epigraph, and sequential closedness of every real lower level set are equivalent.

universe u open Filter ERealFunction variable {X : Type u} [TopologicalSpace X] theorem sequentialLowerSemicontinuous_real_epigraph_seqClosed_lowerLevelSet_seqClosed_tfae (f : X → EReal) : List.TFAE [ ∀ ⦃u : ℕ → X⦄ ⦃x : X⦄, Tendsto u atTop (nhds x) → f x ≤ liminf (fun n ↦ f (u n)) atTop, IsSeqClosed (epigraph f), ∀ ξ : ℝ, IsSeqClosed (lowerLevelSet f ξ) ]

Fact 1.38

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.38: for a map from a metric space to a topological space, continuity is equivalent to sequential continuity; no Hausdorff assumption on the codomain is needed.

universe u v variable {X : Type u} [MetricSpace X] variable {Y : Type v} [TopologicalSpace Y] theorem continuous_iff_seqContinuous_of_metricSpace {f : X → Y} : Continuous f ↔ SeqContinuous f

Fact 1.39

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Fact 1.39: a subset of a metric space is compact if and only if it is sequentially compact.

universe u variable {X : Type u} [MetricSpace X] theorem isCompact_iff_isSeqCompact_of_metricSpace (C : Set X) : IsCompact C ↔ IsSeqCompact C

Lemma 1.40

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.40: if every intersection C ∩ B[z; n] with a closed ball of integer radius n ≥ 1 centered at z is closed, then C itself is closed.

open Metric universe u variable {α : Type u} [MetricSpace α] theorem isClosed_of_isClosed_inter_closedBall_nat (C : Set α) (z : α) (hC : ∀ n : ℕ, 1 ≤ n → IsClosed (C ∩ closedBall z n)) : IsClosed C

Lemma 1.41

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.41: a compact subset of a metric space is closed and bounded.

universe u variable {X : Type u} [MetricSpace X] {C : Set X} theorem isClosed_and_isBounded_of_isCompact (hC : IsCompact C) : IsClosed C ∧ Bornology.IsBounded C

Lemma 1.42

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.42: in a metric space, the neighborhood-filter limit inferior of f at x is the minimum of the sequential limit inferiors along sequences converging to x.

open Filter universe u variable {X : Type u} theorem liminfAt_isLeast_sequentialLiminfValues [MetricSpace X] (f : X → EReal) (x : X) : IsLeast (sequentialLiminfValuesAt f x) (liminf f (nhds x))

Lemma 1.43

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.43 (Cantor): in a complete metric space, a bounded antitone sequence of nonempty closed sets whose diameters converge to 0 has intersection equal to a singleton.

universe u open Set Filter variable {X : Type u} [MetricSpace X] variable [CompleteSpace X] theorem exists_eq_singleton_iInter_of_nonempty_isClosed_antitone_diam_tendsto_zero (C : ℕ → Set X) (hnonempty : ∀ n, (C n).Nonempty) (hclosed : ∀ n, IsClosed (C n)) (hanti : Antitone C) (hbdd : Bornology.IsBounded (C 0)) (hdiam : Tendsto (Metric.diam ∘ C) atTop (nhds 0)) : ∃ x : X, (⋂ n, C n) = {x}

Lemma 1.44

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Lemma 1.44 [Ursescu]: specialized to complete metric spaces through their canonical BaireSpace instance, the corrected Baire-category identities for countable unions of closed sets and countable intersections of open sets hold with the necessary closure operators.

universe u open Set variable {X : Type u} [MetricSpace X] [CompleteSpace X] theorem lemma_1_44 : (∀ C : ℕ → Set X, (∀ n, IsClosed (C n)) → closure (⋃ n, interior (C n)) = closure (interior (⋃ n, C n))) ∧ ∀ U : ℕ → Set X, (∀ n, IsOpen (U n)) → interior (closure (⋂ n, U n)) = interior (⋂ n, closure (U n))

Corollary 1.45

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Corollary 1.45: in a complete metric space, the intersection of a sequence of dense open subsets is again a dense Gδ subset.

universe u open Set variable {X : Type u} [MetricSpace X] [CompleteSpace X] theorem dense_isGδ_iInter_of_dense_open (C : ℕ → Set X) (h_open : ∀ n, IsOpen (C n)) (h_dense : ∀ n, Dense (C n)) : Dense (⋂ n, C n) ∧ IsGδ (⋂ n, C n)

Theorem 1.46

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Theorem 1.46 (Ekeland): on a complete metric space, a proper lower semicontinuous extended-real-valued function that is bounded below satisfies Ekeland's variational principle. If y ∈ ERealFunction.dom f and f y ≤ α + inf f(X) with α ≥ 0 and β > 0, then there exists z such that f z + (α / β) d(y, z) ≤ f y, d(y, z) ≤ β, and f z < f x + (α / β) d(x, z) for every x ≠ z.

universe u open Set theorem exists_ekeland_variational_point {X : Type u} [MetricSpace X] [CompleteSpace X] (f : X → EReal) (hf_proper : ERealFunction.IsProper f) (hf_lsc : LowerSemicontinuous f) (hf_bddBelow : BddBelow (Set.range f)) {α β : ℝ} (hα : 0 ≤ α) (hβ : 0 < β) {y : X} (hy : y ∈ ERealFunction.dom f) (hy_approx : f y ≤ (α : EReal) + sInf (Set.range f)) : ∃ z : X, f z + ((α / β) * dist y z : ℝ) ≤ f y ∧ dist y z ≤ β ∧ ∀ ⦃x : X⦄, x ≠ z → f z < f x + ((α / β) * dist x z : ℝ)

Definition 1.47

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Definition 1.47(1): a map between metric spaces is Lipschitz continuous with constant β when it is formalized by the canonical predicate LipschitzWith β T.

open scoped Topology universe u v recall LipschitzWith

Example 1.48

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Example 1.48: the textbook estimate (1.67) is the pointwise form of the canonical 1-Lipschitz estimate for x ↦ Metric.infDist x C; the book's nonemptiness hypothesis is unnecessary for this inequality.

universe u theorem abs_infDist_sub_infDist_le_dist {α : Type u} [PseudoMetricSpace α] {C : Set α} (x y : α) :

Theorem 1.49

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Theorem 1.49: (Ostrowski) in a metric space, if C is compact, a sequence u takes values in C, and dist (u n) (u (n + 1)) → 0, then the cluster-point set of u is connected. Under the hypothesis u n ∈ C, every cluster point already lies in C, so this is equivalent to the textbook formulation with C ∩ {x | MapClusterPt x atTop u}.

open Filter open scoped Topology universe u variable {X : Type u} [MetricSpace X] theorem isConnected_clusterPoints_of_tendsto_dist_succ_eq_zero {C : Set X} {u : ℕ → X} (hC : IsCompact C) (huC : ∀ n, u n ∈ C) (hstep : Tendsto (fun n ↦ dist (u n) (u (n + 1))) atTop (𝓝 0)) : IsConnected {x | MapClusterPt x atTop u}

Theorem 1.50

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Theorem 1.50 (Banach-Picard): a contraction on a nonempty complete metric space has a unique fixed point, expressed canonically as fixedPoints T = {x}, and this point satisfies the textbook one-step, geometric, a priori, a posteriori, and initial two-sided bounds for every starting point x₀.

universe u open Function variable {X : Type u} [MetricSpace X] [Nonempty X] [CompleteSpace X] variable {β : NNReal} {T : X → X} theorem exists_fixedPoint_with_banach_picard_estimates (hT : ContractingWith β T) : ∃ x : X, fixedPoints T = {x} ∧ (∀ x₀ : X, ∀ n : ℕ, dist (T^[n + 1] x₀) x ≤ β * dist (T^[n] x₀) x) ∧ (∀ x₀ : X, ∀ n : ℕ, dist (T^[n] x₀) x ≤ β ^ n * dist x₀ x) ∧ (∀ x₀ : X, ∀ n : ℕ, dist (T^[n] x₀) x ≤ dist x₀ (T x₀) * (β : ℝ) ^ n / (1 - β)) ∧ (∀ x₀ : X, ∀ n : ℕ, dist (T^[n] x₀) x ≤ dist (T^[n] x₀) (T^[n + 1] x₀) / (1 - β)) ∧ ∀ x₀ : X, dist x₀ (T x₀) / (1 + β) ≤ dist x₀ x ∧ dist x₀ x ≤ dist x₀ (T x₀) / (1 - β)

Theorem 1.51

Lean contributors (project-level attribution): Yifan Bai, Zichen Wang, Zaiwen Wen.

Source header · Attribution source

Theorem 1.51: if a self-map of a complete metric space has iterates whose pairwise distances are bounded by a summable nonnegative sequence, then every orbit converges to the unique fixed point, with the quantitative tail estimate given by the coefficient series.

open Filter Function universe u theorem exists_fixed_point_of_summable_iterate_bound {X : Type u} [MetricSpace X] [CompleteSpace X] {T : X → X} {β : ℕ → NNReal} (hβ : Summable β) (hiterate : ∀ n : ℕ, ∀ x y : X, dist (T^[n] x) (T^[n] y) ≤ (β n : ℝ) * dist x y) (x₀ : X) : ∃ x : X, IsFixedPt T x ∧ (∀ ⦃y : X⦄, IsFixedPt T y → y = x) ∧ Tendsto (fun n ↦ T^[n] x₀) atTop (nhds x) ∧ ∀ n : ℕ, dist (T^[n] x₀) x ≤ (summable_iterate_bound_tail β n : ℝ) * dist x₀ (T x₀)