Skip to content

C_47: lower bound 1.68550999 via a weighted lattice (Lean, registered on Palomar) - #192

Draft
CoolRmal wants to merge 2 commits into
teorth:mainfrom
CoolRmal:c47-lower-bound-weighted-lattice
Draft

CoolRmal wants to merge 2 commits into
teorth:mainfrom
CoolRmal:c47-lower-bound-weighted-lattice

Conversation

@CoolRmal

Copy link
Copy Markdown

Summary

Records a new lower bound for $C_{47}$, the weak-type $(1,1)$ constant of the centered Hardy–Littlewood maximal operator over axis-parallel squares in the plane:

$$ c_2 \ \ge\ \Phi = \frac{2hV-4ab}{1+w} = 1.6855099933\ldots, $$

improving Aldaz's $\tfrac34-\tfrac{\sqrt2}{4}+\tfrac{\sqrt6}{2}\approx 1.6211915$ (2000). $\Phi$ is an explicit algebraic number of degree 16 (closed form in Challenge.lean of the linked repository).

The proof is a complete Lean 4 / Mathlib formalization, registered on the Palomar registry as PALOMAR-2026-09-19-000002 at commit c6a8cb2: https://github.com/CoolRmal/centered-maximal-constant

The construction

The previous record comes from a unit rectangular lattice. The improvement uses unequal masses: mass $1$ at $(ih,jV)$ for even $i$ and mass $w=(17+4\sqrt{22})/9$ for odd $i$, with $h=(5+\sqrt{22})/6$ and $V=h+1$. On a period cell (area $2hV$, mass $1+w$), six explicit witness squares show the maximal function of the measure is at least $1$ outside four thin open slots of size $a\times b$ (the formulas for $a$, $b$ are in the Comments cell). That gives covered area per unit mass $\Phi=(2hV-4ab)/(1+w)$. Truncating the lattice and smearing each atom over a small square turns the measure into integrable test functions, and letting the truncation grow gives $c_2\ge\Phi$. The Comments cell carries enough data to recompute $\Phi$ without downloading anything.

Verification

  • Lean statements, compared by Comparator against a Mathlib-only Challenge.lean: ofReal_phi_le_weakTypeConstant_two ($\Phi\le c_2$), lt_phi / phi_lt ($1.685<\Phi<1.686$), and weakTypeConstant_le_two_pow ($c_d\le 2^d$). They depend only on propext, Classical.choice and Quot.sound: no sorry, no native_decide, no custom axioms.
  • The Lean definition uses closed cubes and the strict level set ${Mf>\alpha}$; this gives the same constant as the non-strict level set on the constant page, and the note under [Lin2026-construction] says so.
  • Registered on Palomar (PALOMAR-2026-09-19-000002): its mechanical check reran the comparator in a sandbox and replayed the proof through Lean's kernel and the independent NanoDa kernel. The result has not been peer reviewed. I have not added an asterisk to the README cell; please add one if you prefer to reserve unmarked cells for published results.

Changes (Markdown only)

  • constants/47a.md: new row in Known lower bounds, a [Lin2026] reference with a [Lin2026-construction] note, and a contribution note.
  • README.md: the $C_{47}$ lower-bound cell becomes 1.68550999, plus a line in Recent progress.

AI use

The construction was found and certified with AI assistance, and formalized with AI assistance in Claude Code under my direction; formalization.yaml in the repository records the models. This pull request was prepared with assistance from Claude (claude-opus-5).

🤖 Generated with Claude Code

Adds a Known lower bounds row for the two-mass lattice construction, the
[Lin2026] reference to its Lean formalization (registered on Palomar as
PALOMAR-2026-09-19-000002), and updates the README cell and Recent progress.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant