The kernel of the half-space trace #
Let H = TauCeti.normalHalfSpace a = {x | a < x.fst} in the Euclidean product ℝ × E. This file
identifies the functions of zero trace on H with the closure of the test functions:
ker (Tr : W^{1,p}(H) → Lᵖ(E)) = W^{1,p}_0(H) for every 1 ≤ p < ∞.
The inclusion W^{1,p}_0(H) ⊆ ker Tr is TauCeti.W1p.halfSpaceTrace_eq_zero_of_w1p0, and
TauCeti.W1p.extendByZeroₗᵢ_mem_w1pSubmodule_iff_halfSpaceTrace_eq_zero says that Tr u = 0
exactly when the extension of u by zero is a Sobolev function on ℝ × E. The remaining step,
proved here for every 1 ≤ p < ∞, is that a function on H whose extension by zero lies in
W^{1,p}(ℝ × E) belongs to W^{1,p}_0(H).
The argument #
Let w ∈ W^{1,p}(ℝ × E) vanish off H. For s > 0, translating w by s in the normal
direction gives a function vanishing on the layer {x.fst < a + s}. There it is unchanged by the
smooth cutoff TauCeti.normalCutoff a (2 / s) of the normal coordinate, which vanishes for
x.fst ≤ a + s / 2 and is one for x.fst ≥ a + s. So the translate is a whole-space Sobolev
function times a bounded cutoff supported in H, and such functions lie in W^{1,p}_0(H)
(TauCeti.W1p.restrictL_contDiffSMul_mem_w1p0Submodule). As s → 0 the translates converge to
w in W^{1,p}(ℝ × E) (TauCeti.Wkp.continuous_translate), so the restriction of w to H
lies in the closed subspace W^{1,p}_0(H).
Main declarations #
TauCeti.W1p.restrictL_mem_w1p0Submodule_normalHalfSpace: a function inW^{1,p}(ℝ × E)vanishing offHrestricts to a function inW^{1,p}_0(H).TauCeti.W1p.mem_w1p0Submodule_normalHalfSpace_iff:u ∈ W^{1,p}_0(H)if and only if its extension by zero lies inW^{1,p}(ℝ × E).TauCeti.W1p.mem_w1p0Submodule_iff_halfSpaceTrace_eq_zeroandTauCeti.W1p.ker_halfSpaceTrace: the kernel of the half-space trace isW^{1,p}_0(H).
References #
- L. C. Evans, Partial Differential Equations, §5.5, Theorem 2 (functions of zero trace).
- H. Brezis, Functional Analysis, Sobolev Spaces and Partial Differential Equations, Proposition 9.18 and its proof.
Functions vanishing off the half-space #
A whole-space Sobolev function vanishing off a half-space has zero boundary values there.
If w ∈ W^{1,p}(ℝ × E), 1 ≤ p < ∞, vanishes almost everywhere outside H = {x | a < x.fst},
then its restriction to H lies in W^{1,p}_0(H).
W^{1,p}_0 of a half-space consists of the functions that extend by zero. For
1 ≤ p < ∞, a function u ∈ W^{1,p}(H) on H = {x | a < x.fst} lies in W^{1,p}_0(H) if and only
if its extension by zero, in value and weak gradient, is a Sobolev function on ℝ × E.
The functions of zero trace on a half-space are those of W^{1,p}_0. For
1 ≤ p < ∞, a function u ∈ W^{1,p}(H) on H = {x | a < x.fst} lies in W^{1,p}_0(H) if and
only if its trace on the boundary hyperplane {a} × E vanishes.
The kernel of the half-space trace is W^{1,p}_0. On H = {x | a < x.fst} and for
1 ≤ p < ∞, the trace W^{1,p}(H) → Lᵖ(E) on the boundary hyperplane {a} × E has kernel
exactly W^{1,p}_0(H), the closure of the test functions on H.