← Back to the manuscript

Verification Appendix

This appendix records the computer-assisted certificate used for the scalar inequality in Theorem 2. All reported bounds use directed outward rounding.

Certified statement. For the exact decimal parameters \(\alpha=0.6342\) and \(\lambda=0.4506\),

\[ J(0.6342,0.4506)<-0.00005<0. \]

Validated computation

The verification uses Arb ball arithmetic through python-flint 0.9.0 at 80 decimal digits. It partitions \([0,1/2]\) into \(2^{16}\) rational cells. On each cell, sign-verified interval brackets enclose the unique optimizer in the definition of \(\Gamma_\lambda\), and directed rounding encloses the resulting value. Coordinatewise monotonicity controls the cell adjacent to \(t=0\).

Summing the cellwise upper bounds gives an upper endpoint below \(-0.0000512188\), which is strictly smaller than \(-0.00005\). Ordinary high-precision quadrature is used only as an independent numerical check, not as part of the proof.

Reproduction

From the project root, install the pinned dependency and run:

python -m pip install -r verification/requirements.txt
python verification/certify_j_arb.py --cells 65536 --dps 80

The verification program, dependency specification, and technical notes accompany the manuscript.