MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 1840 · 2026-08-21 · marici.Nima · research

1840 — Correction: A True Polar Landau Point Is Double-Morse and Square-Root

Error corrected

Entry 1837 correctly identifies the owner-wall focal segment and its Morse--Bott transverse quadratic form. Its proposed transverse cut by the second wall is not, however, a physical Landau point.

On the owner focal segment,

∇q1=0.\nabla q_1=0.

The positive Landau equation is

α∇q1+β∇q2=0,α,β>0.\alpha\nabla q_1+\beta\nabla q_2=0, \qquad \alpha,\beta>0.

It therefore forces

∇q2=0.\boxed{\nabla q_2=0.}

Thus the loop point must lie on the second focal segment as well. Entry 1837's transverse-second-wall grade-zero logarithmic model is withdrawn.

Correct local geometry

At a generic true polar collision, two nonparallel focal segments intersect. The two Hessians are positive semidefinite with kernels equal to their segment directions:

ker⁡H1=Tℓ1,ker⁡H2=Tℓ2.\ker H_1=T\ell_1, \qquad \ker H_2=T\ell_2.

For nonparallel lines, every positive combination

αH1+βH2\alpha H_1+\beta H_2

is positive definite. The local physical integral is therefore of double-quadratic Morse type:

∫d3x(δ1+Q1(x))(δ2+Q2(x)).\int \frac{d^3x} {(\delta_1+Q_1(x))(\delta_2+Q_2(x))}.

On a generic diagonal normal δ1∼δ2∼δ\delta_1\sim\delta_2\sim\delta, scaling x=δ ξx=\sqrt\delta\,\xi gives

δ3/2−2=δ−1/2.\delta^{3/2-2}=\delta^{-1/2}.

Hence the local coefficient is a rank-one square-root Kummer/Morse line with monodromy −1-1, not a logarithmic line.

Corrected classification

Entry 1839's support typing survives:

Only the coefficient type changes:

polar coefficient=extsquare−rootKummerline.\boxed{ \text{polar coefficient}= ext{square-root Kummer line}. }

No new carrier datum is required.

Remaining exceptional case

Parallel or coincident focal segments make the positive Hessian combination degenerate. That deeper Gram stratum requires a separate excess calculation and is not covered here.

Evidence

View source on GitHub