Lax485480.CographWelzlTreeLowerBound
Cographs can require logarithmic Welzl trees
concepts/Lax485480/CographWelzlTreeLowerBound.lean · lax-485480
No public endorsements yet.
Loading review…
Sign in with ORCIDConcept map
Evidence
Each proof establishes this claim relative to its assumptions.
Theorem
For every nonnegative integer k, there is a cograph on at least and at most vertices such that every spanning tree has crossing number at least for the open-neighborhood set system.
The witnesses are the same recursively constructed cographs that require k crossings in every Welzl order. Every spanning tree of crossing number c can be linearized with crossing number at most : delete leaves one at a time and, in reverse, insert each leaf beside its tree neighbor. Inserting a leaf changes a row only when its incident tree edge is crossed, and then adds at most two crossings. Hence an order lower bound of k gives a tree lower bound of .
Crespelle and Gambette proved the corresponding asymptotic lower bound in the equivalent language of cograph contiguity, using cographs with complete binary cotrees. The theorem below records the consequence for arbitrary Welzl spanning trees with explicit finite-size constants inherited from the preceding formalized order obstruction.
Lean source view on GitHub
| 1 | import Lax485480.WelzlTrees |
| 2 | import Lax214022.Cographs |
| 3 | import Lax195003.WelzlOrdersNeighborhoodSetSystem |
| 4 | |
| 5 | /-! |
| 6 | --- |
| 7 | title: Cographs can require logarithmic Welzl trees |
| 8 | type: theorem |
| 9 | --- |
| 10 | For every nonnegative integer *k*, there is a cograph on at least `3^k` and |
| 11 | at most `4^k` vertices such that every spanning tree has crossing number at |
| 12 | least `ceil(k/2)` for the open-neighborhood set system. |
| 13 | |
| 14 | The witnesses are the same recursively constructed cographs that require |
| 15 | *k* crossings in every Welzl order. Every spanning tree of crossing number |
| 16 | *c* can be linearized with crossing number at most `2c`: delete leaves one at |
| 17 | a time and, in reverse, insert each leaf beside its tree neighbor. Inserting |
| 18 | a leaf changes a row only when its incident tree edge is crossed, and then |
| 19 | adds at most two crossings. Hence an order lower bound of *k* gives a tree |
| 20 | lower bound of `ceil(k/2)`. |
| 21 | |
| 22 | Crespelle and Gambette proved the corresponding asymptotic lower bound in |
| 23 | the equivalent language of cograph contiguity, using cographs with complete |
| 24 | binary cotrees. The theorem below records the consequence for arbitrary |
| 25 | Welzl spanning trees with explicit finite-size constants inherited from the |
| 26 | preceding formalized order obstruction. |
| 27 | |
| 28 | # Formalization notes |
| 29 | |
| 30 | The ceiling is expressed in natural arithmetic as `(k + 1) / 2`. The two |
| 31 | exponential inequalities retain the rounded-logarithm-free size convention |
| 32 | of the preceding Welzl-order lower bound. |
| 33 | -/ |
| 34 | |
| 35 | namespace Lax485480.CographWelzlTreeLowerBound |
| 36 | |
| 37 | open Lax195003.WelzlOrdersNeighborhoodSetSystem |
| 38 | open Lax214022.Cographs |
| 39 | open Lax485480.WelzlTrees |
| 40 | |
| 41 | /-- There are cographs of order at most `4^k` on which every spanning tree |
| 42 | has crossing number at least `ceil(k/2)`. -/ |
| 43 | axiom exists_cograph_requiring_treeCrossingNumber_at_least (k : ℕ) : |
| 44 | ∃ n : ℕ, 3 ^ k ≤ n ∧ n ≤ 4 ^ k ∧ |
| 45 | ∃ G : SimpleGraph (Fin n), IsCograph G ∧ |
| 46 | ∀ T : SimpleGraph (Fin n), T.IsTree → |
| 47 | (k + 1) / 2 ≤ |
| 48 | treeCrossingNumber (neighborhoodSetSystem G 1) T |
| 49 | |
| 50 | end Lax485480.CographWelzlTreeLowerBound |
| 51 |
Formalization notes
The ceiling is expressed in natural arithmetic as . The two exponential inequalities retain the rounded-logarithm-free size convention of the preceding Welzl-order lower bound.
Used by
none
From Mathlib
none
Community review
Discussion
Ask a question or add context. Endorsements and structured flags are kept in the review panel above; your ORCID profile must share a public name.
0 comments