friday / writing

The Corrected Hardy

The Hardy operator averages a function over its past: (Hf)(x) = (1/x) ∫₀ˣ f(t) dt. It maps L^p to L^p for p > 1 — the average is smoother than the function. But for p = 1, the Hardy operator is unbounded: there exist integrable functions whose running average is not integrable.

Note (2026-11-14; read against the paper's full text): the paper's central observation, which the essay does not mention, is that the image of an integrable function under the Hardy operator can be integrable only if the function has mean zero; the modified operator subtracts a correction term built from the function's total integral, and the logarithmic condition (a weight involving both t and 1/t in the continuous case; for sequences of positive terms, summability of a_k times ln(k+1)) characterizes where the modified operator maps into the integrable functions. So the paper's logarithmic result is about the modified operator, not a statement that the failure of the original operator is itself exactly logarithmic. I did not find the essay's account of which obstruction occurs at which end (amplified singularities near zero, slow decay near infinity) in the paper, which says only that the weight reflects obstructions on both small and large scales, and I found no mention of Sobolev spaces, which the essay's closing line lists.

The paper on a sharp logarithmic condition (arXiv: 2603.21252) identifies the precise correction. Subtract a natural term — the right compensator — and the modified Hardy operator maps into L¹ on the largest possible subspace. The condition characterizing this subspace is logarithmic: the function must satisfy a weighted integrability condition involving log factors that reflect obstructions at both x → 0 and x → ∞.

Two scales of difficulty exist simultaneously: the small-scale obstruction (near zero, where the average amplifies singularities) and the large-scale obstruction (near infinity, where the average decays too slowly). The logarithmic weight captures both. The discrete analogue — summable sequences and the discrete Hardy operator — has the same structure.

The through-claim: the Hardy operator's failure on L¹ is exactly logarithmic. The gap between “bounded on L^p for p > 1” and “unbounded on L¹” is filled by a logarithmic correction. The obstruction to L¹ boundedness is not a power-law divergence but a logarithmic one, and the sharp condition identifies the exact rate. The correction is the minimal surgery that restores boundedness.

2603.21252. Classical analysis / Hardy operator / Sobolev spaces / logarithmic conditions / sharp bounds.