The past, germ and tail sigma-fields of a process, and Lipschitz continuity at a point
DefinitionDurrettProbability_BrownianThe -field apparatus of chapter 7 of Durrett, Probability: Theory and Examples, for a real-valued process indexed by on a measurable space , together with the pointwise Lipschitz condition of Theorem 7.1.6.
The past. , formed as the supremum, in the lattice of -algebras on , of the pullbacks of the Borel -algebra along the maps for .
The germ. , the infimum of the past fields over all strictly positive times. An event lies in it exactly when it is decided by the path on every initial interval , however short.
The future and the tail. , and .
Lipschitz at a point. A path is Lipschitz at with constant when there is a scale such that
Formalization Note The index set is , so at the Lipschitz condition is one-sided, as it must be. A path differentiable at is Lipschitz at for some constant, so the negation of this condition at every point and for every constant is a strictly stronger statement than nowhere differentiability. For the condition is satisfiable only by a path constant near , which makes its negation trivially true there; the content of Theorem 7.1.6 is at .
The bounded suprema and infima are over the whole index type, with the out-of-range terms contributing the bottom and top -algebras respectively, which are the neutral elements; the results are the ordinary generated and intersected -fields. No measurability of is assumed in these definitions: they are well-formed for any family of functions, and a theorem that needs to sit inside the ambient -field says so.
import Mathlib
open Filter MeasureTheory ProbabilityTheory
open scoped NNReal Topology
namespace DurrettProbability
variable {Ω : Type*} [MeasurableSpace Ω]
/-- `pastSigma B t = σ(B_s : s ≤ t)`, Durrett's `F⁰_t`: the σ-field generated by the process up
to time `t`. Durrett, *Probability: Theory and Examples*, section 7.2. -/
@[instance_reducible]
def pastSigma (B : ℝ≥0 → Ω → ℝ) (t : ℝ≥0) : MeasurableSpace Ω :=
⨆ s ∈ Set.Iic t, MeasurableSpace.comap (B s) inferInstance
/-- `germSigma B = ⋂_{t > 0} σ(B_s : s ≤ t)`, Durrett's germ σ-field `F⁺₀`: the events decided by
the path on every initial interval, however short. Durrett, section 7.2, Theorem 7.2.3. -/
@[instance_reducible]
def germSigma (B : ℝ≥0 → Ω → ℝ) : MeasurableSpace Ω :=
⨅ t ∈ Set.Ioi (0 : ℝ≥0), pastSigma B t
/-- `futureSigma B t = σ(B_s : s ≥ t)`, Durrett's `F'_t`, the future at time `t`. -/
@[instance_reducible]
def futureSigma (B : ℝ≥0 → Ω → ℝ) (t : ℝ≥0) : MeasurableSpace Ω :=
⨆ s ∈ Set.Ici t, MeasurableSpace.comap (B s) inferInstance
/-- `tailSigma B = ⋂_{t ≥ 0} σ(B_s : s ≥ t)`, Durrett's tail σ-field `T`.
Durrett, *Probability: Theory and Examples*, Theorem 7.2.7. -/
@[instance_reducible]
def tailSigma (B : ℝ≥0 → Ω → ℝ) : MeasurableSpace Ω := ⨅ t : ℝ≥0, futureSigma B t
/-- `LipschitzAtPoint f s C` says the path `f` is Lipschitz at the point `s` with constant `C`:
there is a scale `δ > 0` on which `|f t - f s| ≤ C |t - s|`. A path differentiable at `s` is
Lipschitz at `s`, so failure of this for every `s` and every `C` is failure of differentiability
everywhere. Durrett, *Probability: Theory and Examples*, Theorem 7.1.6. -/
def LipschitzAtPoint (f : ℝ≥0 → ℝ) (s : ℝ≥0) (C : ℝ) : Prop :=
∃ δ : ℝ, 0 < δ ∧ ∀ t : ℝ≥0, |(t : ℝ) - s| ≤ δ → |f t - f s| ≤ C * |(t : ℝ) - s|
end DurrettProbability