Sublevel sets of the plumbing-lattice weight function #
Némethi builds lattice homology out of the sublevel sets S_c = {x | χ_k(x) ≤ c} of the
characteristic weight function of a negative-definite plumbing: the lattice complex is
filtered by them, and the invariants read off from that filtration — including the d-invariant
analogues — only make sense because each S_c is a finite set of lattice points and because
χ_k attains a minimum. This file proves both facts.
The mechanism is completing the square. Write A for the intersection matrix, d = det A, and
u = adjugate A *ᵥ k, so that A *ᵥ u = d • k and hence A(x, u) = d ⟪k, x⟫. Expanding the
self-pairing of 2d • x + u therefore gives
A(2d • x + u, 2d • x + u) = 4 d² (⟪k, x⟫ + A(x, x)) + A(u, u),
whose middle factor ⟪k, x⟫ + A(x, x) is exactly the weight numerator
characteristicWeightNumerator k x, which is -2 χ_k(x). So a sublevel set of χ_k is carried,
by the injective affine map x ↦ 2d • x + u, into a superlevel set of the intersection-form
self-pairing, and those are finite on a negative-definite plumbing by
IsNegativeDefinite.finite_setOfPred_le_intersectionForm_self. Negative-definiteness enters twice:
it makes d ≠ 0, so the affine map is injective, and it makes the target set finite.
Finiteness immediately yields a minimizer: the lattice points with χ_k(x) ≤ χ_k(0) form a finite
nonempty set, and a minimum over it is a global minimum. That minimum is the numerical input to the
d-invariant of the plumbed three-manifold in the spin^c structure recorded by k, and it
transforms under the two moves that change the representative k of a spin^c structure: passing
to k + 2 · A m shifts it by -χ_k(m), and spin^c conjugation leaves it fixed.
Main results #
TauCeti.PlumbingGraph.finite_setOfPred_le_characteristicWeightNumerator: superlevel sets of the weight numerator are finite.TauCeti.PlumbingGraph.finite_setOfPred_characteristicWeight_le: sublevel sets of the characteristic weight are finite.TauCeti.PlumbingGraph.exists_forall_characteristicWeight_le: the characteristic weight attains a global minimum, andTauCeti.PlumbingGraph.bddBelow_range_characteristicWeight: its range is bounded below.TauCeti.PlumbingGraph.sInfCharacteristicWeight: the infimum of the weight as an integer, withTauCeti.PlumbingGraph.sInfCharacteristicWeight_defunfolding it,TauCeti.PlumbingGraph.le_sInfCharacteristicWeightthe greatest-lower-bound rule, andTauCeti.PlumbingGraph.sInfCharacteristicWeight_leandTauCeti.PlumbingGraph.exists_characteristicWeight_eq_sInfCharacteristicWeightsaying that on a negative-definite plumbing it is a lower bound and that it is attained.TauCeti.PlumbingGraph.sInfCharacteristicWeight_add_two_mulVecandTauCeti.PlumbingGraph.sInfCharacteristicWeight_conjugate: how that value transforms under the two moves on the representative of a spin^c structure.TauCeti.PlumbingGraph.IsNegativeDefinite.not_bddAbove_range_characteristicWeight: on a nonempty vertex set the weight is unbounded above, so the sublevel sets are proper.
References #
This advances TauCetiRoadmap/CombinatorialHeegaardFloer/README.md, Lane L ("lattice homology"),
whose target is "Némethi's lattice (co)homology ℍ⁻/ℍ⁰ as a ℤ[U]-module from lattice points
and weight functions … and d-invariant analogues": the sublevel sets treated here are the ones
that filtration is taken over. See Némethi,
arXiv:0709.0841, after Ozsváth--Szabó,
arXiv:math/0203265.
On a negative-definite plumbing only finitely many lattice points have characteristic-weight numerator above a fixed bound.
The affine map x ↦ 2 (det A) • x + adjugate A *ᵥ k is injective because negative-definiteness
forces det A ≠ 0, and by the completed square it carries the superlevel set into a superlevel
set of the intersection-form self-pairing, which is finite.
On a negative-definite plumbing every sublevel set of the characteristic weight function is
finite: only finitely many lattice points satisfy χ_k(x) ≤ N.
This is the finiteness that makes Némethi's filtration of the lattice complex by weight sublevel sets a filtration by finitely generated pieces.
On a negative-definite plumbing the characteristic weight function attains a global minimum.
The lattice points with χ_k(x) ≤ χ_k(0) form a finite nonempty set, and a point minimizing χ_k
over it minimizes χ_k everywhere. The value at such a point is the numerical input to the
d-invariant of the plumbed three-manifold in the spin^c structure recorded by k.
On a negative-definite plumbing the characteristic weight function is bounded below.
The infimum of the characteristic weight function, as an integer.
On a negative-definite plumbing this is a genuine minimum — sInfCharacteristicWeight_le bounds
the weight below by it and exists_characteristicWeight_eq_sInfCharacteristicWeight attains it —
and it is the numerical input to the d-invariant of the plumbed three-manifold in the spin^c
structure recorded by k. Without negative-definiteness the range need not be bounded below and
the value is the junk value 0 of Int.csInf_of_not_bddBelow, which is why the name records the
infimum rather than a minimum.
Equations
- P.sInfCharacteristicWeight k = sInf (Set.range (P.characteristicWeight k))
Instances For
The infimum of the characteristic weight, as the infimum of its range.
The greatest-lower-bound rule for the infimum of the characteristic weight: any lower bound for the weight is a lower bound for the infimum. No hypothesis on the plumbing is needed, since the range is nonempty.
On a negative-definite plumbing the infimum of the characteristic weight is a lower bound for the characteristic weight.
On a negative-definite plumbing the infimum of the characteristic weight is attained: some lattice point has that exact weight, so it is a minimum.
Passing to the other representative k + 2 · A m of the same spin^c structure shifts the
minimal characteristic weight by the constant -χ_k(m).
The two weight functions differ by the translation χ_{k + 2·A m}(x) = χ_k(x + m) - χ_k(m) of
characteristicWeight_add_two_mulVec_eq_sub, and translating the lattice argument permutes the
range; so the minimum picks up exactly the constant. This is what lets a d-invariant computed
from a chosen representative be transported to any other.
The infimum of the characteristic weight is invariant under spin^c conjugation: negating the covector and the lattice point together preserves the weight, so it permutes the range.
Conjugation is an involution on characteristic covectors, and
characteristicWeight_conjugate_neg says χ_{-k}(-x) = χ_k(x); hence the two weight functions
have the same range, and in particular the same infimum.
On a negative-definite plumbing with at least one vertex the characteristic weight function is unbounded above: a bound would make every lattice point lie in one sublevel set, which is finite while the lattice is infinite.
Together with finite_setOfPred_characteristicWeight_le this says that the weight sublevel sets
form a genuine exhaustion of the lattice by finite subsets each of which is proper: no single
sublevel set is the whole lattice.