Theorem Relative rank fibers and two-edge detection of rank two
✓ theorem proved

Relative rank fibers and two-edge detection of rank two

Let \(D\) be a Coxeter system with two simple generators, intrinsic length \(\ell\), and Bruhat order \(\le\). Fix \(l\le r\) and write \(J=[l,r]_D\).

Every integer \(n\) with \(\ell(l)\le n\le\ell(r)\) occurs as the length of a vertex of \(J\). At each strict internal rank \[ \ell(l)<n<\ell(r) \] there are exactly two vertices. Each endpoint rank contains just its endpoint.

Two-edge paths detect relative rank two in the following precise forms.

If \(\ell(y)=\ell(x)+2\), there are distinct \(a,b\in D\) such that \[ x\to z\to y\quad\Longleftrightarrow\quad z=a\text{ or }z=b. \]

Conversely, suppose \(x\to a\to y\), and suppose every \(z\) with \(x\to z\to y\) equals \(a\) or a specified vertex \(b\). Then \(\ell(y)=\ell(x)+2\).

Here a two-edge path uses upward reflection edges; their individual length gaps can be odd integers larger than one.

Relation to the OpenAI notes

Exact intermediates and the three-vertex contradiction.

OpenAI notes: The interval-shape lemma counts exactly r two-edge paths when the positive relative length gap r is even. The transport proof uses the equality of ambient intermediate sets to conclude that the target gap is two. dih:shape, dih:relative-paths, dih:transport.

Lean (primary): The native gap-two result lists two intermediates. For the converse, a gap at least four produces two distinct vertices one rank above the bottom and a third three ranks above it. All three are two-edge intermediates, contradicting the assumed two-value bound. Ambient interval membership then permits inverse transport of every target intermediate. ReflectionSchedules:1717, ReflectionSchedules:1673, DyerPaths:302, DyerPaths:341, DyerPaths:634.

Role in the proof: The Blueprint follows the three-intermediate contradiction and proves the ambient and transported intermediate-set equivalences in both directions.

Proof

Step 1. Identify the fibers inside the interval

For a strict internal rank \(n\), one has \(0<n<\ell(r)\). The internal-rank theorem gives exactly two distinct elements of \(D\) of length \(n\).

Every such element \(z\) satisfies \(\ell(l)<\ell(z)<\ell(r)\). The dihedral Bruhat comparison theorem turns each strict length comparison into a strict Bruhat comparison, so \(l<z<r\). Thus both elements lie in \(J\), and no third vertex can have that length.

At the lower endpoint rank, \(l\le z\) together with \(\ell(l)=\ell(z)\) forces \(z=l\). The analogous argument at the upper endpoint gives \(z=r\). These endpoint vertices and the two internal vertices supply every rank, proving the fiber and existence assertions.

The same reasoning works between any two vertices \(x,y\) of a larger interval: if \(\ell(x)<n<\ell(y)\), the two elements of length \(n\) lie above \(x\) and below \(y\), hence remain in that larger interval. This is the form used by the source's two_interval_elements_length.

Step 2. Describe the two intermediates at gap two

Assume \(\ell(y)=\ell(x)+2\). A path \(x\to z\to y\) gives \[ \ell(x)<\ell(z)<\ell(y), \] so the only possible length is \(\ell(z)=\ell(x)+1\).

Conversely, for every element \(z\) of that length, both differences are one. The odd-gap edge criterion therefore gives both edges. Since \(0<\ell(x)+1<\ell(y)\), the internal-rank theorem supplies exactly two such elements \(a,b\). This proves the stated equivalence for all \(z\in D\).

Step 3. Bound the possible total gap in the converse

Suppose \(x\to a\to y\). Each edge has a positive odd length difference, so \[ d=\ell(y)-\ell(x) \] is a positive even integer, with \(d\ge2\). If \(d\ne2\), then \(d\ge4\).

The rank \(\ell(x)+1\) is positive and strictly below \(\ell(y)\), so choose two distinct elements \(p,q\) of that length. The rank \(\ell(x)+3\) is also positive and strictly below \(\ell(y)\), so choose an element \(v\) of that length.

Step 4. Produce three distinct intermediates when the gap is too large

The gaps from \(x\) to \(p,q\) are one, and their gaps to \(y\) are \(d-1\), a positive odd integer. Thus both \(p,q\) are intermediates of two-edge paths.

The gaps through \(v\) are three and \(d-3\), also positive and odd. Thus \(v\) is a third intermediate. It is distinct from \(p,q\) because its length differs from theirs.

The assumed two-value bound would put the three distinct elements \(p,q,v\) in the set \(\{a,b\}\), which has at most two members. This is impossible. Therefore \(d=2\), proving the converse.

References

The local source declarations below give the construction and proof.

Lean formalization

OAI.KLInvariance.Dihedral.two_interval_elements_length: ReflectionSchedules.lean:1717.

OAI.KLInvariance.Dihedral.exists_interval_length: DyerPaths.lean:488.

OAI.KLInvariance.Dihedral.two_step_rank_two: ReflectionSchedules.lean:1673.

OAI.KLInvariance.Dihedral.rank_two_of_two_step_bound: DyerPaths.lean:341.

Used by