A Counterexample to the Rockafellar Sum Conjecture on c₀ (2026)

Source excerpts; proofs omitted.

NormalCone

Lean authors: Zichen Wang.

Source header

The normal-cone set-valued operator determined by a dual pairing and a set.

universe u v namespace DualPairing variable {X : Type u} {Xstar : Type v} [NormedAddCommGroup X] [NormedSpace ℝ X] [NormedAddCommGroup Xstar] [NormedSpace ℝ Xstar] def normalCone (P : DualPairing X Xstar) (C : Set X) : SetValuedOperator X Xstar := fun x ↦ {xstar | x ∈ C ∧ ∀ y ∈ C, P.toLinearPairing (y - x) xstar ≤ 0}

ClosedBall

Lean authors: Zichen Wang.

Source header

The normal-cone graph of a positive-radius closed ball is maximally monotone when the pairing map onto the strong dual is surjective.

universe u v namespace DualPairing variable {X : Type u} {Xstar : Type v} [NormedAddCommGroup X] [NormedSpace ℝ X] [NormedAddCommGroup Xstar] [NormedSpace ℝ Xstar] theorem maximalMonotone_normalConeGraph_closedBall (P : DualPairing X Xstar) (h_surjective : Function.Surjective P.toDual) (r : ℝ) (hr : 0 < r) : Maximal P.IsMonotone (P.normalConeGraph (Metric.closedBall (0 : X) r))

Lean authors: Zichen Wang.

Source header

A local nonnegativity law for an operator extends to its sum with the normal cone of a constraint set contained in the local region.

open scoped Pointwise universe u v namespace DualPairing variable {X : Type u} {Xstar : Type v} [NormedAddCommGroup X] [NormedSpace ℝ X] [NormedAddCommGroup Xstar] [NormedSpace ℝ Xstar] theorem quadratic_nonneg_of_mem_graph_add_normalCone (P : DualPairing X Xstar) (M : SetValuedOperator X Xstar) (C U : Set X) (h_zero : 0 ∈ C) (h_subset : C ⊆ U) (h_localNonneg : ∀ x xstar, (x, xstar) ∈ M.graph → x ∈ U → 0 ≤ P.quadratic (x, xstar)) (x : X) (ystar : Xstar) (h_graph : (x, ystar) ∈ (M + P.normalCone C).graph) : 0 ≤ P.quadratic (x, ystar)

NatTriangle

Lean authors: Zichen Wang.

Source header

A summable family on ℕ × ℕ decomposes into its diagonal, strict-upper, and strict-lower triangular parts.

universe u public theorem Summable.tsum_prod_split_triangles {E : Type u} [AddCommGroup E] [UniformSpace E] [IsUniformAddGroup E] [CompleteSpace E] [T0Space E] {f : ℕ × ℕ → E} (hf : Summable f) : ∑' p, f p = (∑' n, f (n, n)) + (∑' n, ∑' m, if n < m then f (n, m) else 0) + ∑' n, ∑' m, if m < n then f (n, m) else 0

HilbertProd2

Lean authors: Zichen Wang.

Source header

Construct a point of HilbertProd2 H from its two coordinates.

open scoped InnerProductSpace universe u namespace HilbertProd2 def mk {H : Type u} (v : H) (r : ℝ) : HilbertProd2 H := WithLp.toLp 2 (v, r)

LorentzCone

Lean authors: Zichen Wang.

Source header

The Lorentz energy P ^ 2 - ‖N‖ ^ 2 of a point (P, N).

universe u namespace Lorentz def energy {H0 : Type u} [Norm H0] (x : ℝ × H0) : ℝ := x.1 ^ 2 - ‖x.2‖ ^ 2

HilbertProd2

Lean authors: Zichen Wang.

Source header

The completed-square expression is the Lorentz energy of the corresponding HilbertProd2 coordinates.

universe u namespace Lorentz theorem completeSquareEnergy {H : Type u} [NormedAddCommGroup H] (v : H) (s t : ℝ) : -(‖v‖ ^ 2) + t * s = energy ((t + s) / 2, HilbertProd2.mk v ((t - s) / 2))

SeedTemplate

Lean authors: Zichen Wang.

Source header

The affine half-line generated by a base point and a direction is Lipschitz whenever the direction has norm at most one.

open scoped InnerProductSpace universe u namespace Lorentz theorem affine_halfLine_norm_sub_le_abs_sub {H : Type u} [SeminormedAddCommGroup H] [NormedSpace ℝ H] {z v : ℝ × H} (hv : ‖v.2‖ ≤ 1) {p q : ℝ} : ‖(z.2 + p • v.2) - (z.2 + q • v.2)‖ ≤ |p - q|

Lean authors: Zichen Wang.

Source header

The real ℓ¹ space of sequences indexed by ℕ.

abbrev L1Seq : Type := lp (fun _ : ℕ ↦ ℝ) 1

L1Synthesis

Lean authors: Zichen Wang.

Source header

The continuous linear synthesis operator for a uniformly bounded vector family.

universe u namespace L1Seq noncomputable def synthesis {H : Type u} [NormedAddCommGroup H] [NormedSpace ℝ H] [CompleteSpace H] (u : ℕ → H) {K : ℝ} (hu : ∀ n, ‖u n‖ ≤ K) : L1Seq →L[ℝ] H := (lp.tsumCLM ℝ ℕ H).comp (lp.mapCLM 1 (fun n ↦ ContinuousLinearMap.toSpanSingleton ℝ (u n)) (synthesis_bound_nonneg u hu) (synthesis_coordinate_norm_le u hu))

SetValuedOperator

Lean authors: Zichen Wang.

Source header

A set-valued operator from X to Y is a function assigning a set in Y to each point of X.

open scoped Pointwise universe u v abbrev SetValuedOperator (X : Type u) (Y : Type v) := X → Set Y

RepeatingSchedule

Lean authors: Zichen Wang.

Source header

A sequence together with a strictly increasing sequence of indices at which each value occurs.

universe u structure RepeatingSchedule (α : Type u) where /-- The underlying sequence. -/ toFun : ℕ → α /-- The sequence of occurrence indices for each value. -/ occurrence : α → ℕ → ℕ /-- The occurrence indices for each value are strictly increasing. -/ strictMono_occurrence : ∀ a, StrictMono (occurrence a) /-- Every designated occurrence index has the requested value. -/ apply_occurrence : ∀ a k, toFun (occurrence a k) = a

DualPairing

Lean authors: Zichen Wang.

Source header

An explicit continuous pairing of X with Xstar, nondegenerate in the Xstar argument.

universe u v structure DualPairing (X : Type u) (Xstar : Type v) [NormedAddCommGroup X] [NormedSpace ℝ X] [NormedAddCommGroup Xstar] [NormedSpace ℝ Xstar] where toDual : Xstar →L[ℝ] StrongDual ℝ X injective_toDual : Function.Injective toDual

MaximalExtension

Lean authors: Zichen Wang.

Source header

Every monotone subset of a dual-pairing product is contained in a maximal monotone subset.

universe u v namespace DualPairing variable {X : Type u} {Xstar : Type v} [NormedAddCommGroup X] [NormedSpace ℝ X] [NormedAddCommGroup Xstar] [NormedSpace ℝ Xstar] theorem exists_maximal_monotone_superset (P : DualPairing X Xstar) {S : Set (X × Xstar)} (hS : P.IsMonotone S) : ∃ G, S ⊆ G ∧ Maximal P.IsMonotone G

Monotone

Lean authors: Zichen Wang.

Source header

A set is monotone with respect to a dual pairing when the pairing's quadratic form is nonnegative on the difference of every two points of the set.

universe u v namespace DualPairing variable {X : Type u} {Xstar : Type v} [NormedAddCommGroup X] [NormedSpace ℝ X] [NormedAddCommGroup Xstar] [NormedSpace ℝ Xstar] def IsMonotone (P : DualPairing X Xstar) (S : Set (X × Xstar)) : Prop := ∀ z ∈ S, ∀ w ∈ S, 0 ≤ P.quadratic (z - w)

Lean authors: Zichen Wang.

Source header

A real sequence tending to zero along Filter.atTop determines an element of C0Seq.

open Topology namespace C0Seq def ofTendsto (x : ℕ → ℝ) (hx : Filter.Tendsto x Filter.atTop (𝓝 0)) : C0Seq

Coordinate

Lean authors: Zichen Wang.

Source header

Evaluation at a fixed coordinate is continuous on C0Seq.

namespace C0Seq theorem continuous_apply (n : ℕ) : Continuous (fun x : C0Seq ↦ x n)

Lean authors: Zichen Wang.

Source header

The isometric linear embedding of L1Seq into the strong dual of C0Seq defined by the coordinatewise pairing.

namespace C0Seq noncomputable def l1ToDual : L1Seq →ₗᵢ[ℝ] StrongDual ℝ C0Seq

DualPairing

Lean authors: Zichen Wang.

Source header

The separating dual pairing between real C0Seq and real L1Seq induced by coordinatewise summation.

namespace C0Seq noncomputable def coordinateDualPairing : DualPairing C0Seq L1Seq := ⟨l1ToDual.toContinuousLinearMap, l1ToDual.injective⟩

Lean authors: Zichen Wang.

Source header

A monotone subset of C0Seq × L1Seq containing a point with negative quadratic pairing does not contain the origin.

namespace C0Seq theorem zero_not_mem_of_quadraticPairing_neg {S : Set (C0Seq × L1Seq)} {z : C0Seq × L1Seq} (hz : z ∈ S) (hz_neg : quadraticPairing z < 0) (hS : coordinateDualPairing.IsMonotone S) : (0, 0) ∉ S

Surjective

Lean authors: Zichen Wang.

Source header

The strong-dual map of the coordinate pairing between real C0Seq and real L1Seq is surjective.

namespace C0Seq theorem coordinateDualPairing_surjective : Function.Surjective coordinateDualPairing.toDual

Lean authors: Zichen Wang.

Source header

Pairing a standard coordinate vector with a summable sequence evaluates that sequence at the chosen coordinate and scales the result.

namespace C0Seq @[simp] theorem pairingL_c0Single_apply (n : ℕ) (r : ℝ) (a : L1Seq) : pairingL (c0Single n r) a = r * a n

Lean authors: Zichen Wang.

Source header

The element of C0Seq whose value is r at coordinate n and zero at every other coordinate.

open Topology def c0Single (n : ℕ) (r : ℝ) : C0Seq

Transpose

Lean authors: Zichen Wang.

Source header

The transpose of a continuous linear map from L1Seq to C0Seq, with its domain reindexed through the canonical identification of the strong dual of C0Seq with L1Seq.

namespace ContinuousLinearMap noncomputable def reindexedTranspose (A : L1Seq →L[ℝ] C0Seq) : L1Seq →L[ℝ] StrongDual ℝ L1Seq := A.paperTranspose.comp C0Seq.l1ToDual.toContinuousLinearMap

Transpose

Lean authors: Zichen Wang.

Source header

The transpose of a continuous linear map from L1Seq, defined by precomposition on continuous linear functionals.

universe u v namespace ContinuousLinearMap noncomputable def paperTranspose {E : Type u} [NormedAddCommGroup E] [NormedSpace ℝ E] (T : L1Seq →L[ℝ] E) : StrongDual ℝ E →L[ℝ] StrongDual ℝ L1Seq := ContinuousLinearMap.precomp ℝ T

RationalTimeOperator

Lean authors: Zichen Wang.

Source header

The sum of a coordinate's absolute value and twice its strict upper absolute tail tends to zero.

open Filter Topology namespace L1Seq theorem positiveOperatorCoordBound_tendsto (a : L1Seq) : Filter.Tendsto (fun n : ℕ ↦ |a n| + 2 * ∑' m : ℕ, if n < m then |a m| else 0) Filter.atTop (nhds 0)

Bilinear

Lean authors: Zichen Wang.

Source header

The continuous bilinear form obtained by pairing the positive operator with a second L1Seq argument.

open scoped InnerProductSpace namespace L1Seq noncomputable def positiveOperatorForm : L1Seq →L[ℝ] L1Seq →L[ℝ] ℝ := C0Seq.pairingL.comp positiveOperator

Lean authors: Zichen Wang.

Source header

The full rational-time Gram series of an L1Seq is the squared UnitL2 norm of its image under intervalCoordinateOperator.

namespace L1Seq open scoped InnerProductSpace public theorem gramTsum_eq_norm_intervalCoordinateOperator_sq (a : L1Seq) : (∑' p : ℕ × ℕ, min (rationalTime p.1) (rationalTime p.2) * a p.1 * a p.2) = ‖intervalCoordinateOperator a‖ ^ 2

Lean authors: Zichen Wang.

Source header

The triangular rational-time operator has pointwise norm bound two.

namespace L1Seq theorem norm_positiveOperator_apply_le_two (a : L1Seq) : ‖positiveOperator a‖ ≤ 2 * ‖a‖

Parametrization

Lean authors: Zichen Wang.

Source header

The continuous linear parametrization (a, t) ↦ (-A a + t • d, a).

namespace Lorentz noncomputable def parametrization (d : C0Seq) : (L1Seq × ℝ) →L[ℝ] (C0Seq × L1Seq) := (((-L1Seq.positiveOperator).comp (ContinuousLinearMap.fst ℝ L1Seq ℝ)) + (ContinuousLinearMap.snd ℝ L1Seq ℝ).smulRight d).prod (ContinuousLinearMap.fst ℝ L1Seq ℝ)

LorentzCoordinates

Lean authors: Zichen Wang.

Source header

The positive Lorentz coordinate P on the parametrized subspace.

namespace Lorentz noncomputable def positiveCoordinate (d : C0Seq) (hd : d ≠ 0) : parametrizedSubspace d →ₗ[ℝ] ℝ

FixedPositive

Lean authors: Zichen Wang.

Source header

Choosing t = 2 * p - C0Seq.pairingL d a makes the positive Lorentz coordinate of parametrizedPoint d a t equal to p.

namespace Lorentz theorem positiveCoordinate_adjustedParameter (d : C0Seq) (hd : d ≠ 0) (a : L1Seq) (p : ℝ) : positiveCoordinate d hd (parametrizedPoint d a (2 * p - C0Seq.pairingL d a)) = p

Lean authors: Zichen Wang.

Source header

Vanishing of the negative coordinate of parametrizedPoint d a t recovers the parameters a = 0 and t = 0.

namespace Lorentz theorem parameters_eq_zero_of_negativeCoordinate_eq_zero (d : C0Seq) (hd : d ≠ 0) (a : L1Seq) (t : ℝ) (hN : negativeCoordinate d hd (parametrizedPoint d a t) = 0) : a = 0 ∧ t = 0

LorentzEnergy

Lean authors: Zichen Wang.

Source header

The quadratic pairing of a parametrized point is its representative-level operator expression.

open scoped InnerProductSpace namespace Lorentz theorem quadraticPairing_parametrization (d : C0Seq) (a : L1Seq) (t : ℝ) : C0Seq.quadraticPairing (parametrization d (a, t)) = -(‖L1Seq.intervalCoordinateOperator a‖ ^ 2) + t * C0Seq.pairingL d a

Lean authors: Zichen Wang.

Source header

The symmetric pairing of (x, u) with the parametrized point at (a, -C0Seq.pairingL d a) expands in terms of the positive operator, interval-coordinate operator, and pairing with d.

open scoped InnerProductSpace namespace Lorentz theorem detectorPairingFormula (d x : C0Seq) (u a : L1Seq) : C0Seq.symmetricForm (x, u) (parametrization d (a, -C0Seq.pairingL d a)) = C0Seq.pairingL (x + L1Seq.positiveOperator u) a - 2 * ⟪L1Seq.intervalCoordinateOperator u, L1Seq.intervalCoordinateOperator a⟫_ℝ - C0Seq.pairingL d u * C0Seq.pairingL d a

Lean authors: Zichen Wang.

Source header

The source point Ψ(a, t), regarded as an element of parametrizedSubspace d.

namespace Lorentz def parametrizedPoint (d : C0Seq) (a : L1Seq) (t : ℝ) : parametrizedSubspace d := ⟨parametrization d (a, t), mem_parametrizedSubspace d _ |>.mpr ⟨a, t, parametrization_apply d a t⟩⟩

SamePositiveCoordinate

Lean authors: Zichen Wang.

Source header

The quadratic pairing of the difference of two parametrized points with the same positive coordinate is the negative squared distance between their negative coordinates.

namespace Lorentz theorem sameP_quadraticPairing (d : C0Seq) (hd : d ≠ 0) (w mP : parametrizedSubspace d) (hP : positiveCoordinate d hd w = positiveCoordinate d hd mP) : C0Seq.quadraticPairing (w - mP) = -(‖negativeCoordinate d hd w - negativeCoordinate d hd mP‖ ^ 2)

ScaledDetector

Lean authors: Zichen Wang.

Source header

The affine detector combination used by the seed construction.

namespace Lorentz noncomputable def scaledDetector {d : C0Seq} (ξ h : parametrizedSubspace d) (r : ℝ) : parametrizedSubspace d := ξ + r • h

SeedAssembly

Lean authors: Zichen Wang.

Source header

The full source seed with the actual scheduled detector points.

open Filter Topology namespace Lorentz def assembledSeed (z v : parametrizedSubspace axisDirection) : Set (C0Seq × L1Seq) := {w | (∃ p : ℝ, p ≤ 1 / 2 ∧ w = (z + p • v : parametrizedSubspace axisDirection)) ∨ (∃ n : ℕ, w = (seedPoint n : C0Seq × L1Seq)) ∨ w = ((2 : ℝ) • v : parametrizedSubspace axisDirection)}

SeedBase

Lean authors: Zichen Wang.

Source header

The carrier point obtained from a source sequence a with prescribed positive Lorentz coordinate s.

namespace Lorentz noncomputable def seedBase {d : C0Seq} (a : L1Seq) (s : ℝ) : parametrizedSubspace d := parametrizedPoint d a (2 * s - C0Seq.pairingL d a)

SeedComparisons

Lean authors: Zichen Wang.

Source header

The half-line's negative coordinate lies in the backward cone from time one.

namespace Lorentz theorem seed_halfLine_cone_bound (z v : parametrizedSubspace axisDirection) (hz : ‖negativeCoordinate axisDirection seed_axis_ne_zero z‖ ≤ 1 / 64) (hv : ‖negativeCoordinate axisDirection seed_axis_ne_zero v‖ ≤ 1 / 64) (p : ℝ) (hp : p ≤ 1 / 2) : ‖negativeCoordinate axisDirection seed_axis_ne_zero (z + p • v)‖ ≤ 1 - p

SeedCounterexample

Lean authors: Zichen Wang.

Source header

The source seed yields a maximal monotone operator whose sum with the radius-twelve ball normal cone is not maximal, despite interior qualification.

open scoped Pointwise namespace Lorentz theorem exists_seedCounterexample : ∃ M : SetValuedOperator C0Seq L1Seq, Maximal C0Seq.coordinateDualPairing.IsMonotone M.graph ∧ Maximal C0Seq.coordinateDualPairing.IsMonotone (C0Seq.coordinateDualPairing.normalConeGraph (Metric.closedBall (0 : C0Seq) 12)) ∧ (M.dom ∩ interior (Metric.closedBall (0 : C0Seq) 12)).Nonempty ∧ ¬ Maximal C0Seq.coordinateDualPairing.IsMonotone (M + C0Seq.coordinateDualPairing.normalCone (Metric.closedBall (0 : C0Seq) 12)).graph

SeedSchedule

Lean authors: Zichen Wang.

Source header

Unit differences of arbitrarily small interval norm can be placed beyond any finite prefix.

open Filter Topology open scoped InnerProductSpace namespace Lorentz theorem exists_remote_unitDifference_norm_lt (q N : ℕ) (ε : ℝ) (hε : 0 < ε) : ∃ j, N < j ∧ ‖L1Seq.intervalCoordinateOperator (unitDifference q j)‖ < ε

SeedWitnesses

Lean authors: Zichen Wang.

Source header

The source direction is an explicit remote unit-difference base with positive time one, small negative coordinate and bounded primal component.

namespace Lorentz theorem exists_seed_direction : ∃ v : parametrizedSubspace axisDirection, positiveCoordinate axisDirection seed_axis_ne_zero v = 1 ∧ ‖negativeCoordinate axisDirection seed_axis_ne_zero v‖ < 1 / 64 ∧ ‖(v : C0Seq × L1Seq).1‖ ≤ 5

UnitDifferenceDetector

Lean authors: Zichen Wang.

Source header

The dual unit difference e_i^* - e_j^*.

open Topology open scoped InnerProductSpace namespace Lorentz noncomputable def unitDifference (i j : ℕ) : L1Seq := lp.single 1 i (1 : ℝ) - lp.single 1 j (1 : ℝ)

Quadratic

Lean authors: Zichen Wang.

Source header

The absolute values of the diagonal terms of the rational-time quadratic series associated with an L1Seq are summable.

namespace L1Seq theorem summable_abs_positiveOperator_diagonal (a : L1Seq) : Summable (fun n : ℕ ↦ |rationalTime n * (a n) ^ 2|)

QuadraticIdentity

Lean authors: Zichen Wang.

Source header

The quadratic pairing of positiveOperator a with a is the rational-time Gram double series.

namespace L1Seq public theorem positiveOperator_quadratic_eq_gramTsum (a : L1Seq) : C0Seq.pairingL (positiveOperator a) a = ∑' p : ℕ × ℕ, min (rationalTime p.1) (rationalTime p.2) * a p.1 * a p.2

SymmetricPart

Lean authors: Zichen Wang.

Source header

The symmetric part of the rational-time positive operator is twice the Gram operator of the interval-coordinate map.

namespace L1Seq public theorem positiveOperator_symmetricPart : C0Seq.pairingL.comp positiveOperator + positiveOperator.reindexedTranspose = 2 • (intervalCoordinateAdjoint.comp intervalCoordinateOperator)

Lean authors: Zichen Wang.

Source header

If E is contained in the open unit interval and has null relative complement, then arbitrarily close points of E can be chosen on both sides of an interior point so that the open interval between them avoids any prescribed finite set not containing that point.

namespace Real public theorem exists_conull_points_around_avoiding_finset {E : Set ℝ} (_hE : E ⊆ Set.Ioo (0 : ℝ) 1) (hconull : MeasureTheory.volume (Set.Ioo (0 : ℝ) 1 \ E) = 0) {t : ℝ} (ht : t ∈ Set.Ioo 0 1) (F : Finset ℝ) (htF : t ∉ F) (ε : ℝ) (hε : 0 < ε) : ∃ sLeft ∈ E ∩ Set.Ioo (t - ε) t, ∃ sRight ∈ E ∩ Set.Ioo t (t + ε), ∀ x ∈ F, x ∉ Set.Ioo sLeft sRight

Lean authors: Zichen Wang.

Source header

The canonical real-linear inclusion of UnitL2 into its AEEqFun representation.

open MeasureTheory namespace UnitL2 def toAEEqFun : UnitL2 →ₗ[ℝ] (ℝ →ₘ[volume.restrict (Set.Ioc (0 : ℝ) 1)] ℝ) := Submodule.subtype (MeasureTheory.Lp.LpSubmodule ℝ ℝ 2 (volume.restrict (Set.Ioc (0 : ℝ) 1)))

IntervalCoordinateOperator

Lean authors: Zichen Wang.

Source header

Applying intervalCoordinateOperator gives the sum of the interval coordinates.

namespace L1Seq theorem intervalCoordinateOperator_apply (a : L1Seq) : intervalCoordinateOperator a = ∑' n : ℕ, a n • UnitL2.rationalIntervalVec n

Lean authors: Zichen Wang.

Source header

Evaluation of the interval-coordinate adjoint at an ℓ¹ vector is the inner product with the corresponding synthesized interval vector.

open scoped InnerProductSpace namespace L1Seq theorem intervalCoordinateAdjoint_apply (y : UnitL2) (a : L1Seq) : intervalCoordinateAdjoint y a = ⟪y, intervalCoordinateOperator a⟫_ℝ

Injective

Lean authors: Zichen Wang.

Source header

The interval-coordinate operator from real summable sequences to UnitL2 is injective.

namespace L1Seq public theorem intervalCoordinateOperator_injective : Function.Injective intervalCoordinateOperator

Lean authors: Zichen Wang.

Source header

The contribution to a pointwise representative jump from coefficients outside a finite set whose rational times lie in a given half-open interval.

namespace L1Seq public noncomputable def jumpRemainder (a : L1Seq) (J : Finset ℕ) (sLeft sRight : ℝ) : ℝ := ∑' n : ℕ, if n ∉ J ∧ rationalTime n ∈ Set.Ioc sLeft sRight then a n else 0

Pointwise

Lean authors: Zichen Wang.

Source header

Evaluation of the pointwise representative is its defining scalar series.

namespace L1Seq public theorem pointwiseRepresentative_apply (a : L1Seq) (s : ℝ) : pointwiseRepresentative a s = ∑' n : ℕ, a n * (Set.Ioo (0 : ℝ) (rationalTime n)).indicator (fun _ ↦ (1 : ℝ)) s

IntervalIndicator

Lean authors: Zichen Wang.

Source header

The UnitL2 vector represented by the indicator of the open interval (0, t).

open MeasureTheory open scoped InnerProductSpace namespace UnitL2 noncomputable def intervalVec (t : unitInterval) : UnitL2 := MeasureTheory.indicatorConstLp 2 measurableSet_Ioo (MeasureTheory.measure_ne_top (volume.restrict (Set.Ioc (0 : ℝ) 1)) (Set.Ioo (0 : ℝ) t)) (1 : ℝ)

RationalIntervalIndicator

Lean authors: Zichen Wang.

Source header

The interval-indicator vector at the nth time in the fixed rational enumeration.

open MeasureTheory namespace UnitL2 noncomputable def rationalIntervalVec (n : ℕ) : UnitL2 := intervalVec ⟨rationalTime n, (rationalTime_mem_Ioo n).1.le, (rationalTime_mem_Ioo n).2.le⟩

RationalTime

Lean authors: Zichen Wang.

Source header

The real-valued sequence obtained from rationalTimeEquiv.

public noncomputable def rationalTime (n : ℕ) : ℝ := ((rationalTimeEquiv n).1 : ℝ)

RationalTime

Lean authors: Zichen Wang.

Source header

Every point of the unit interval is the limit of a strictly increasing subsequence of the canonical enumeration of rational times.

open Topology public theorem exists_strictMono_rationalTime_tendsto (t : ℝ) (ht : t ∈ Set.Icc (0 : ℝ) 1) : ∃ φ : ℕ → ℕ, StrictMono φ ∧ Filter.Tendsto (rationalTime ∘ φ) Filter.atTop (𝓝 t)

Sequences

Lean authors: Zichen Wang.

Source header

If a sequence in a metric space visits every neighborhood strictly after every index, then each point is the limit of a strictly increasing subsequence.

open Filter Topology universe u v theorem exists_strictMono_subseq_tendsto {X : Type u} [MetricSpace X] (x : ℕ → X) (h_tail : ∀ b : X, ∀ U ∈ 𝓝 b, ∀ N : ℕ, ∃ n : ℕ, N < n ∧ x n ∈ U) (a : X) : ∃ φ : ℕ → ℕ, StrictMono φ ∧ Tendsto (x ∘ φ) atTop (𝓝 a)

Formalization contributors

Zichen Wang

Project-level credit · Source roster