Theorem Slope betweenness is equivalent to a positive root combination
✓ theorem proved

Slope betweenness is equivalent to a positive root combination

Fix a finite-generator real Coxeter system. Let \(h\) be its positive height functional and \(f\) the chosen separating linear functional; its injective geometric slope is \(s(a)=f(a)/h(a)\) on positive roots. All such roots have \(h(a)>0\). Let \(a,b,c\) be positive roots, with \(c\in P(a,b)\) and \(s(a)<s(b)\). Then \[ s(a)<s(c)<s(b) \quad\Longleftrightarrow\quad c=ra+tb\text{ for some }r,t>0. \]

The same cone also controls inversion signs. If \(c=ra+tb\) with \(r,t\ge0\), and \(w^{-1}a,w^{-1}b\) are positive, then \(w^{-1}c\) is positive. If instead both transformed roots are negative, then \(w^{-1}c\) is negative. Equivalently, nonnegative combinations preserve a common noninversion or a common inversion sign.

Proof

Step 1. Normalize the linear combination

Since \(c\in P(a,b)\), write \(c=ra+tb\) for real \(r,t\). Applying height gives \(h(c)=rh(a)+th(b)\). Put \[ u=\frac{rh(a)}{h(c)},\qquad v=\frac{th(b)}{h(c)}. \] Then \(u+v=1\), and applying the numerator functional gives \[ s(c)=u\,s(a)+v\,s(b). \]

Step 2. Derive positive coefficients from strict betweenness

Let \(\delta=s(b)-s(a)>0\). Solving the two equations in Step 1 gives \[ v=\frac{s(c)-s(a)}{\delta},\qquad u=\frac{s(b)-s(c)}{\delta}. \] If \(s(a)<s(c)<s(b)\), both are positive. All heights are positive, so \(r=u h(c)/h(a)>0\) and \(t=v h(c)/h(b)>0\).

Step 3. Derive betweenness from positive coefficients

Conversely, if \(r,t>0\), then \(u,v>0\) and \(u+v=1\). Hence \[ s(c)-s(a)=v\delta>0,\qquad s(b)-s(c)=u\delta>0. \] This gives strict betweenness.

Step 4. Preserve positivity under a common inverse action

For the noninverted assertion, linearity gives \[ w^{-1}c=r\,w^{-1}a+t\,w^{-1}b. \] Each coordinate is a sum of nonnegative terms. The vector is a nonzero root, since the group action is invertible and \(c\ne0\). It is therefore positive, as asserted.

Step 5. Preserve a common inverted sign

If \(w^{-1}a,w^{-1}b\) are negative, their coordinates are nonpositive. For each coordinate \(j\), \[ (w^{-1}c)_j=r(w^{-1}a)_j+t(w^{-1}b)_j\le0 \] because \(r,t\ge0\). Again \(w^{-1}c\) is a nonzero root. It is therefore negative, proving the inverted version.

These two closure statements combine with the endpoint bounds of a plane basis to determine the inversion cuts in that plane.

References

The local source declarations below define the objects and prove the stated properties.

Lean formalization

OAI.KLInvariance.TitsSpace.functionalRatio_between: ReflectionSchedules.lean:138.

OAI.KLInvariance.TitsSpace.positive_combination_of_slope_between: Decomposition.lean:395.

OAI.KLInvariance.TitsSpace.noninverted_of_positive_combination: Decomposition.lean:434.

OAI.KLInvariance.TitsSpace.inverted_of_positive_combination: DyerPaths.lean:844.

Used by