Essential bounds from multiplier estimates #
If multiplication by a function a increases Lᵖ norms by at most a factor C, for some
finite exponent p, then a is essentially bounded by C. This file proves two forms of this
multiplier criterion on a sigma-finite measure space:
MeasureTheory.eLpNorm_top_le_of_forall_eLpNorm_indicator_le: it suffices to test the estimate on indicators of sets of finite measure, that is, to know‖a 𝟙_s‖_p ≤ C μ(s) ^ (1 / p)for all such setss;MeasureTheory.eLpNorm_top_le_of_dense_of_eLpNorm_mul_le: it suffices to test the estimate‖a f‖_p ≤ C ‖f‖_pon a dense set of functionsfinLᵖ, for1 ≤ p < ∞.
No integrability of a is assumed beyond almost-everywhere strong measurability. In the second
form, an indicator is approximated in Lᵖ by test functions, a subsequence converges almost
everywhere, and Fatou's lemma for Lᵖ seminorms passes the estimate to the indicator.
A typical application is to trigonometric polynomials on the circle: a function whose
multiplication operator is bounded on trigonometric polynomials in L² is essentially bounded.
This identifies the operators on circle L² commuting with the bilateral shift as the
multiplication operators by essentially bounded symbols.
Multiplier criterion, indicator form. On a sigma-finite measure space, if for some
exponent 0 < p < ∞ every set s of finite measure satisfies ‖a 𝟙_s‖_p ≤ C μ(s) ^ (1 / p),
then a is essentially bounded by C.
Multiplier criterion. On a sigma-finite measure space, if for some exponent
1 ≤ p < ∞ multiplication by a satisfies ‖a f‖_p ≤ C ‖f‖_p for all f in a dense subset
of Lᵖ, then a is essentially bounded by C.