The Poincaré inequality fails on the whole space #
A Poincaré inequality bounds the Lᵖ seminorm of a function by the Lᵖ seminorm of its
derivative, ‖u‖_p ≤ C ‖Du‖_p. Mathlib proves such an estimate in
MeasureTheory.eLpNorm_le_eLpNorm_fderiv, for 1 ≤ p < n and functions supported in a fixed
bounded set s, with the explicit s-dependent constant
MeasureTheory.eLpNormLESNormFDerivOfLeConst ℝ μ s p p. The PDE roadmap asks for the companion
fact that pins the role of that hypothesis: on the whole of a finite-dimensional space no single
constant works for all compactly supported functions, so the boundedness of the support is
load-bearing rather than an artefact of the proof.
The obstruction is scaling. Dilating the variable by r > 0 multiplies ‖u‖_p by r ^ (n / p)
and ‖Du‖_p by r ^ (n / p - 1) (TauCeti.eLpNorm_comp_inv_smul and
TauCeti.eLpNorm_fderiv_comp_inv_smul), so applying a putative inequality to the dilates of one
fixed bump function forces ‖u‖_p ≤ C r⁻¹ ‖Du‖_p for every r, and letting r → ∞ makes the
bump's own seminorm vanish. Note that p ≠ 0 is the only hypothesis needed: the failure is not
confined to the subcritical range p < n in which the positive result lives, and includes
p = ∞.
TauCeti.Analysis.Sobolev.Poincare.Slab proves the matching positive statement: bounding the
support in a single direction already suffices, and the constant is then the width of the slab.
Main declarations #
TauCeti.not_exists_eLpNorm_le_const_mul_eLpNorm_fderiv: no constantCsatisfies‖u‖_p ≤ C ‖Du‖_pfor all compactly supportedC¹functions on a finite-dimensional real normed space with an additive Haar measure.TauCeti.not_exists_eLpNorm_le_const_mul_eLpNorm_fderiv_euclideanSpace: the roadmap'sℝⁿform, for Lebesgue measure onEuclideanSpace ℝ (Fin n).
No Poincaré inequality holds on the whole space. There is no constant C with
‖u‖_p ≤ C ‖Du‖_p for every compactly supported C¹ function u, for any p ≠ 0, including
p = ∞.
Compare MeasureTheory.eLpNorm_le_eLpNorm_fderiv, which supplies such a constant once every
u is supported in one fixed bounded set; the proof here shows that hypothesis cannot be
dropped.
The Poincaré inequality fails on ℝⁿ with Lebesgue measure: the roadmap's form of
TauCeti.not_exists_eLpNorm_le_const_mul_eLpNorm_fderiv.