Pairing with a density #
The pairing f ↦ ∑ e, f e * η e of a density η : Density E against real-valued functions
f : E → ℝ, bundled as a continuous linear functional (an element of StrongDual ℝ (E → ℝ)).
Assembling it from already-bundled pieces (ContinuousLinearMap.proj, scalar multiples, sums)
makes linearity and continuity automatic.
For strictly positive η the pairing is coercive on nonnegative functions: a bound
η.pairing f ≤ c forces the coordinatewise bound f e ≤ c / η e. Consequently the sublevel
sets of the pairing over Density.toReal '' Γ.Adm are compact — the hypothesis needed to apply
the minimum-at-an-extreme-point machinery of LeanModulus.Common.ExtremePoints in the duality
argument.
The pairing f ↦ ∑ e, f e * η e of the density η against real-valued functions,
as a continuous linear functional on E → ℝ.
Equations
- η.pairing = ∑ e : E, ↑(η e) • ContinuousLinearMap.proj e
Instances For
Pairing η against the coercion of a density ρ computes the ℝ≥0-valued length
η.length ρ, coerced to ℝ. This bridges the real-analysis side (pairing) back to the
ℝ≥0 admissibility language of LeanModulus.Common.FamilyOfObjects.
If a nonnegative function f pairs against a strictly positive density η to at most c,
then each coordinate satisfies f e ≤ c / η e. This is the coercivity estimate that traps
sublevel sets of the pairing in a compact box.
Sublevel sets of the pairing with a strictly positive density η, taken within the image
of the admissible set, are compact: they are closed and trapped in the box
∏ e, [0, c / η e] by the coercivity estimate Density.apply_le_div_of_pairing_le.