First moments of the auxiliary clock #
Compact square-mean convergence implies fixed-time convergence in mean. The arbitrary-rate resolvent bound then passes to the clock before recurrence is established.
References #
- J. Bertoin, Lévy Processes, Cambridge University Press, 1996, Chapter IV.
theorem
BGW.StableTrees.ae_supAuxLocalTime_nonneg
{θ : ℝ}
(hθ : 1 < θ ∧ θ ≤ 2)
:
∀ᵐ (f : Skorokhod.DLoc) ∂↑(stableLocalLaw θ), ∀ (t : NNReal), 0 ≤ supAuxLocalTime θ f t
The clock is nonnegative at all times on one event.
theorem
BGW.StableTrees.integrable_supAuxLocalTime
{θ : ℝ}
(hθ : 1 < θ ∧ θ ≤ 2)
(T : NNReal)
:
MeasureTheory.Integrable (fun (f : Skorokhod.DLoc) => supAuxLocalTime θ f T) ↑(stableLocalLaw θ)
The clock has integrable values at every deterministic time.
The compact approximation error used in the square-mean estimate.
Equations
- BGW.StableTrees.supAuxApproxError θ ε T f = ⨆ (t : ↑(Set.Iic T)), |BGW.StableTrees.supOccupation ε f ↑t - BGW.StableTrees.supAuxLocalTime θ f ↑t|
Instances For
theorem
BGW.StableTrees.ae_abs_sub_le_supAuxApproxError
{θ : ℝ}
(hθ : 1 < θ ∧ θ ≤ 2)
{ε : ℝ}
(hε : 0 < ε)
(T : NNReal)
:
∀ᵐ (f : Skorokhod.DLoc) ∂↑(stableLocalLaw θ), |supOccupation ε f T - supAuxLocalTime θ f T| ≤ supAuxApproxError θ ε T f
The fixed-time error is bounded by the compact error on the continuous-path event.
theorem
BGW.StableTrees.integral_abs_sub_le_supAuxApproxError
{θ : ℝ}
(hθ : 1 < θ ∧ θ ≤ 2)
{ε δ : ℝ}
(hε : 0 < ε)
(hδ : 0 < δ)
(T : NNReal)
:
∫ (f : Skorokhod.DLoc), |supOccupation ε f T - supAuxLocalTime θ f T| ∂↑(stableLocalLaw θ) ≤ δ + (∫ (f : Skorokhod.DLoc), supAuxApproxError θ ε T f ^ 2 ∂↑(stableLocalLaw θ)) / δ
An elementary square bound converts compact square error into fixed-time mean error.
theorem
BGW.StableTrees.tendsto_integral_abs_supOccupation_sub
{θ : ℝ}
(hθ : 1 < θ ∧ θ ≤ 2)
(T : NNReal)
:
Filter.Tendsto
(fun (ε : ℝ) => ∫ (f : Skorokhod.DLoc), |supOccupation ε f T - supAuxLocalTime θ f T| ∂↑(stableLocalLaw θ))
(nhdsWithin 0 (Set.Ioi 0)) (nhds 0)
Fixed-time occupations converge in mean along the full positive-cutoff filter.
theorem
BGW.StableTrees.tendsto_integral_supOccupation
{θ : ℝ}
(hθ : 1 < θ ∧ θ ≤ 2)
(T : NNReal)
:
Filter.Tendsto (fun (ε : ℝ) => ∫ (f : Skorokhod.DLoc), supOccupation ε f T ∂↑(stableLocalLaw θ))
(nhdsWithin 0 (Set.Ioi 0)) (nhds (∫ (f : Skorokhod.DLoc), supAuxLocalTime θ f T ∂↑(stableLocalLaw θ)))
The occupation means converge to the clock mean.