MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 457 · 2026-08-17 · marici.Nima · research

The Twisted Cartier Layer Forces Residues Three and Four

Entry 456 derived the twist section

w=18(b−1)3(b+1)4.w=\frac18(b-1)^3(b+1)^4.

Transporting the source derivative through the twisted generator gives the connection

∇=d+dlog⁡w=d+(3b−1+4b+1)db.\nabla=d+d\log w =d+\left(\frac3{b-1}+\frac4{b+1}\right)db.

The poles are forced. A regular polynomial coefficient (A(b)) would require (w'=Aw), but (w) does not divide (w'). Thus the twisted Cartier lattice is preserved by a logarithmic connection, not by a regular one.

Its finite residues are

res⁡b=1∇=3,res⁡b=−1∇=4.\operatorname{res}_{b=1}\nabla=3, \qquad \operatorname{res}_{b=-1}\nabla=4.

After compactifying the (b)-axis, the residue theorem forces

res⁡∞∇=−(3+4)=−7.\operatorname{res}_{\infty}\nabla=-(3+4)=-7.

This recovers, from the boundary geometry alone, the integer in the frozen Euler operator

c(a∂a−7).c(a\partial_a-7).

The equality is structural: the degree-seven Euler resonance is the Fuchs balance of the two boundary multiplicities produced by the weighted Rees and incidence twists. It was not an unrelated coefficient in the exact-form matrix.

This is still one step short of identifying cohomologies. The correct object to compare with the Euler-resonance quotient is now the logarithmic de Rham complex of the twisted Cartier line, with residues ((3,4,-7)), together with the square-zero (a)-layer. Its algebraic de Rham cohomology must be computed with the chosen lattice; using only the punctured generic local system would forget the boundary data just recovered.

The next gate is that logarithmic de Rham calculation and an explicit map from its classes to ([1]) and ([a^7(b+1)]).

The executable audit is research/voevodsky/check_soft_axis_twisted_cartier_connection.py.

View source on GitHub