Marcinkiewicz interpolation for operators defined on a restricted class #
interpolation_weak11_strong22 (Interpolation.lean) assumes sublinearity, the
weak (1,1) bound and the strong (2,2) bound for every measurable function.
A Calderón--Zygmund operator obtained as the L^2-extension of a singular
integral does not satisfy such hypotheses: outside L^2 it is only defined by a
convention, so the endpoint estimates hold only on the classes on which the
operator is genuinely defined.
This file proves the same conclusion, with the same constant
p · 2^p · (A₁/(p-1) + A₂²/(2-p)) for 1 < p < 2, from hypotheses restricted to
the classes actually visited by the truncation argument. Two variants are
provided:
interpolation_weak11_strong22_of_classes, whose endpoint hypotheses are assumed for integrable functions (weak(1,1)) and forL^2functions (strong(2,2)), with sublinearity only for an (integrable,L^2) pair;interpolation_weak11_strong22_of_l2_classes, whose hypotheses are in addition confined toL^2inputs throughout, which is the form available for an operator constructed as anL^2-extension and only later extended toL^pby density.
Splitting f at the level t/2 produces the two pieces f·1_{|f| > t/2} and
f·1_{|f| ≤ t/2}. For f ∈ L^p with 1 < p < 2 the first is integrable and
the second lies in L^2; these two facts, proved here from the pointwise bounds
of InterpolationTruncBounds.lean, are what lets the restricted hypotheses be
applied. Everything else is the argument of Interpolation.lean.
The two truncations of an L^p function #
The part of an L^p function above a positive level is integrable: on
{|f| > l} the modulus is bounded by l^{1-p} |f|^p.
The part of an L^p function below a positive level lies in L^2: on
{|f| ≤ l} the square of the modulus is bounded by l^{2-p} |f|^p.
The distribution-function bound from the two pieces #
The layer-cake integration #
The layer-cake half of the interpolation theorem: the distribution-function
bound at every level, integrated against the weight p t^{p-1}. This is the proof
of interpolation_weak11_strong22 with the tail estimate taken as a hypothesis.
Interpolation with hypotheses restricted to the endpoint classes #
Marcinkiewicz interpolation for an operator defined on L^1 + L^2. The
weak (1,1) bound is assumed only for integrable functions, the strong (2,2)
bound only for L^2 functions, and sublinearity only for a pair consisting of an
integrable function and an L^2 function. For f ∈ L^p with 1 < p < 2 the
conclusion and the constant are those of interpolation_weak11_strong22.
Because f ∈ L^p is in general neither integrable nor square integrable, the
measurability of T f for the input f itself is not a consequence of hTmeas
and is taken as the explicit hypothesis hTf.
Marcinkiewicz interpolation for an operator defined on L^2. Every
hypothesis is confined to square-integrable inputs: this is the form available for
an operator constructed as the L^2-extension of a singular integral, before it
has been extended to L^p by density. For f ∈ L^p ∩ L^2 with 1 < p < 2 the
conclusion and the constant are those of interpolation_weak11_strong22.
The two truncations of such an f at a level l > 0 lie in L^1 ∩ L^2 and in
L^2 respectively, so all four hypotheses apply to them; and T f is measurable
by hTmeas applied to f itself, so no extra measurability hypothesis is
needed.
The exponents used by the pressure estimates #
interpolation_weak11_strong22_of_classes at the exponent p = 3/2.
interpolation_weak11_strong22_of_classes at the exponent p = 6/5.
interpolation_weak11_strong22_of_l2_classes at the exponent p = 3/2.
interpolation_weak11_strong22_of_l2_classes at the exponent p = 6/5.