Geometry
---
Status: proved universal obstruction
Date: 2026-08-15
Depends on: BINARY-R-DIVISOR-COLLISION.md, FAB-MIRROR-CHARACTER-OBSTRUCTION.md
Claim boundary: this rules out an apparently attractive smooth-denominator strategy. It does not prove Erdős-Straus.
1. Setup
Let p be Mordell-hard. Then
Choose any n>=1 such that
and put
Then r>0, and because p==1 mod 8,
In particular r is odd and r==3 mod 4.
The one-denominator subtraction is
Put
By the binary divisor-collision theorem, the remainder can split into two unit fractions only if
2. Every divisor ratio is dyadic modulo r
From
we have
Every divisor of N=p2^n has the form
Therefore every quotient of two divisors is, modulo r, a power of 2:
Indeed replacing every occurrence of p by 2^(n+2) makes this immediate.
3. Jacobi obstruction
For every odd positive integer
the Jacobi symbols satisfy
and
Hence every power of 2 has Jacobi sign +1 modulo r, while -1 has sign -1.
Thus
Consequently
and the binary remainder does not split.
Theorem
For every Mordell-hard prime p and every positive dyadic first denominator A=2^n with 4A>p, the associated binary numerator
cannot rescue p.
QED.
4. Why this matters
The obstruction is not lack of divisor count. It is a squareclass obstruction.
The maximally smooth choice A=2^n collapses the entire signed divisor box into one Jacobi-positive subgroup, while the binary target -1 is Jacobi-negative.
Therefore a successful smooth-denominator construction must import at least one factor carrying the opposite character.
On the Mordell-hard lane, the frozen small shield satisfies
So the missing ingredient cannot be built solely from 2,3,5,7. It must include a genuine external nonresidue prime or squareclass.
This independently matches:
FAB-HARD-NONRESIDUE-BRIDGE.md;FAB-MIRROR-CHARACTER-OBSTRUCTION.md;FAB-GCD-SURFACE-REFORMULATION.md.
All three routes now point to the same reduced design principle: