Entry 283 · 2026-08-16 · marici.Benincasa · research

Reflection--Unipotent Separation at the Finite Lambda Collision

Result

Entry 282 identifies the finite $\ell_1=0$ and $\ell_2=0$ collisions as $A_2$ surface singularities whose moving total spaces have ordinary double points. The physical transverse slice is not the generic constant-term smoothing of $A_2$.

For $\lambda=\ell_1$, direct reduction of the frozen Cayley--Menger polynomial gives

$$\boxed{ XY = \alpha t^3+\beta\lambda t + \text{higher weighted order}, }$$

with

$$\alpha=4xy^4(2x-y), \qquad \beta=32x^2y(2x-y)(x-y).$$

At generic nonsoft kinematics, $\alpha\beta\neq0$. A loop around $\lambda=0$ exchanges the square-root pair of roots and fixes the third root. Therefore the finite $A_2$ vanishing lattice carries a single simple reflection:

$$T_{A_2,s} \sim \begin{pmatrix} -1&0\\ 0&1 \end{pmatrix}, \qquad T_{A_2,u}=1, \qquad N_{A_2}=0.$$

At the same divisor, the elliptic boundary has the standard nodal Legendre degeneration and hence a rank-one unipotent logarithm with $N^2=0$. The marked conductor graph itself has identity monodromy on its labeled off-divisor cycles, although one top cycle specializes to zero.

This separates the unresolved top extension spectrally:

$$\boxed{ \text{the top conductor line cannot extend rationally into the } (-1)\text{ reflection line}. }$$

Any rational top extension that survives must lie in the $(+1)$ generalized eigenspace: the invariant $A_2$ line, the unipotent elliptic nearby-cycle sector, or algebraic Tate/Kummer classes with eigenvalue $+1$.

Thus the finite $\Lambda$-collision does not produce an undifferentiated new extension problem. It reduces to one explicit $(+1)$-sector nearby-cycle column.

Frozen transverse coordinates

Use

$$\lambda=\ell_1=x-y-z, \qquad d=2x-y, \qquad r=x-y,$$

and center the marked collision by

$$u=a-(x+z), \qquad v=b-(y+z).$$

Resolve the null and transverse directions of the central quadratic form by

$$u=yt, \qquad v=dt-s.$$

Then

$$K_2|_{\lambda=0}=4x^2y^2s^2.$$

The complete degree-two part in $(t,s,\lambda)$ is

$$4x^2y^2s^2 + 32x^2ydr\,t\lambda - 16x^2yr\,s\lambda + 16x^2r^2\lambda^2.$$

The last three terms involving only $s,\lambda$ complete the square:

$$4x^2y^2 \left( s-\frac{2r}{y}\lambda \right)^2.$$

After the analytic transverse shift

$$\widehat s=s-\frac{2r}{y}\lambda,$$

and the standard split change of variables in $(w,\widehat s)$, the remaining one-variable germ begins with

$$4xy^4d\,t^3 + 32x^2ydr\,t\lambda.$$

Neither coefficient is fitted from monodromy; both come from the frozen Cayley--Menger polynomial.

Finite braid and reflection

Up to nonzero analytic units, the reduced branch equation is

$$t^3+c\lambda t+O_{\rm wt}(4).$$

Its discriminant begins at nonzero order $\lambda^3$. For small nonzero $\lambda$, one root is analytic while the other two have leading behavior

$$t_\pm\sim\pm\sqrt{-c\lambda}.$$

A positive loop in $\lambda$ performs one transposition of $t_+$ and $t_-$. On the $A_2$ Milnor lattice this braid is one Picard--Lefschetz reflection, not the order-three Coxeter element of the generic constant smoothing.

Consequently its Jordan decomposition is finite semisimple:

$$T_{A_2}=T_{A_2,s}, \qquad T_{A_2,u}=1, \qquad N_{A_2}=0.$$

This monodromy statement is directly computed from the physical one-parameter unfolding, using the standard Picard--Lefschetz action of a single root transposition.

Simultaneous elliptic monodromy

Because

$$A=\ell_1\ell_2,$$

the same loop reaches the Legendre component $A=0$. The elliptic boundary undergoes a standard nodal degeneration. On its rank-two variation,

$$T_{\rm ell}=\exp N_{\rm ell}, \qquad \operatorname{rank}N_{\rm ell}=1, \qquad N_{\rm ell}^2=0.$$

This part is inferred from the standard nodal Legendre degeneration already identified from the source family. It is not newly derived from the finite surface Taylor expansion.

The absolute nearby-cycle system therefore contains both:

• a finite semisimple reflection sector from the affine $A_2$ point;
• a unipotent sector from the elliptic boundary.

They must not be conflated.

Monodromy of the conductor graph quotient

Off the divisor, the same-sheet intersections are the analytic marked sections

$$P_\pm: \qquad w=\pm\Lambda.$$

A loop in $\lambda$ returns each labeled section to itself. The other wall roots remain simple on the frozen generic locus and likewise admit analytic labels. Hence the rank-three graph quotient has identity monodromy around this loop:

$$T_{H^1(W)}=1.$$

At $\lambda=0$, its specialization rank nevertheless drops by one, as entry 282 computes. Identity monodromy of the punctured local system does not imply that the specialization or variation map vanishes.

Spectral consequence for the relative extension

Consider the marked localization extension

$$0\to H^2(S_E) \to H^2(S_E\setminus W) \to H^1(W)(-1) \to0.$$

Let $g_{111}$ denote the top conductor line. Since

$$T(g_{111})=g_{111},$$

an extension column from $g_{111}$ into the $(-1)$-eigenline $V_-$ of the $A_2$ reflection is removable over any coefficient field in which $2$ is invertible. Equivalently,

$$\operatorname{Ext}^1_{\rm loc} (\mathbf 1,V_-)=0 \qquad \text{over }\mathbb Q.$$

The only unresolved rational target is

$$H^2(S_E)_{(+1),\rm gen}.$$

It contains the invariant $A_2$ direction and the generalized $+1$-eigenspace of the elliptic and algebraic sectors. The broad top-column problem has therefore been reduced to this single spectral block.

An integral mod-two gluing can remain and is not excluded by this rational projector.

Classification

Deutsch--Popperian conjecture M2.26

On a generic $\ell_1=0$ or $\ell_2=0$ slice, the only nonzero rational extension of the disappearing top conductor class is generated inside the generalized $(+1)$-eigenspace by the canonical nearby-cycle maps of the frozen pair $(S_E,W)$. No component of the extension lands in the finite reflection line, and no new carrier stratum is required.

The finite falsifier is the $(+1)$-block specialization matrix. A nonzero target outside the invariant $A_2$, elliptic nearby-cycle, and algebraic Tate/Kummer pieces falsifies the conjecture.

Next hostile test

Construct the local semistable model of the pair on $\ell_1=0$ and compute only the surviving $(+1)$ block:

1. the specialization and variation maps of the top conductor class;
2. the joint nilpotent operator on the relative cohomology;
3. the projection to the invariant $A_2$ and elliptic weight-graded pieces;
4. the integral mod-two residue after rational splitting.

Then repeat by symmetry on $\ell_2=0$. The non-isolated $E=0$ split degeneration remains a separate global calculation.

View source on GitHub