Planar Emulators Conjecture

Is Nearly True for Cubic Graphs

Petr Hliněný
Faculty of Informatics,
Informatics Masaryk University
Botanická 68a, 602 00 Brno, Czech Rep.

e-mail: [email protected]

* Joint work with Martin Derka,

Derka Brno / Waterloo.
Petr Hliněný, Eurocomb, Pisa, 2013 1 / 16 Planar Emulators Conjecture for Cubic . . .
Small math miracle for start. . .

Petr Hliněný, Eurocomb, Pisa, 2013 2 / 16 Planar Emulators Conjecture for Cubic . . .
Small math miracle for start. . .

Petr Hliněný, Eurocomb, Pisa, 2013 2 / 16 Planar Emulators Conjecture for Cubic . . .
Small math miracle for start. . .
– turning nonplanar into planar

Petr Hliněný, Eurocomb, Pisa, 2013 2 / 16 Planar Emulators Conjecture for Cubic . . .
Small math miracle for start. . .
– turning nonplanar into planar

Petr Hliněný, Eurocomb, Pisa, 2013 2 / 16 Planar Emulators Conjecture for Cubic . . .
1 Definitions
Motivation: Exploring the two graphs locally, we cannot see any difference. . .

Petr Hliněný, Eurocomb, Pisa, 2013 3 / 16 Planar Emulators Conjecture for Cubic . . .
1 Definitions
Motivation: Exploring the two graphs locally, we cannot see any difference. . .
A graph H is a cover of a graph G if there exists a pair of onto mappings
(a projection) ϕ : V(H) → V(G), ψ : E(H) → E(G)
such that ψ maps the edges incident with each vertex v in H
bijectively onto the edges incident with ϕ(v) in G.

e3 ψ(e3 )
v s e2 ϕ(v) s ψ(e2 )
H e1 → ψ(e1 ) G

Petr Hliněný, Eurocomb, Pisa, 2013 3 / 16 Planar Emulators Conjecture for Cubic . . .
1 Definitions
Motivation: Exploring the two graphs locally, we cannot see any difference. . .
A graph H is a cover of a graph G if there exists a pair of onto mappings
(a projection) ϕ : V(H) → V(G), ψ : E(H) → E(G)
such that ψ maps the edges incident with each vertex v in H
bijectively onto the edges incident with ϕ(v) in G.

e3 ψ(e3 )
v s e2 ϕ(v) s ψ(e2 )
H e1 → ψ(e1 ) G

Remark. The edge ψ(uv) has always ends ϕ(u), ϕ(v), and hence only
ϕ : V(H) → V(G), the vertex projection,
is enough to be specified for simple graphs.

Petr Hliněný, Eurocomb, Pisa, 2013 3 / 16 Planar Emulators Conjecture for Cubic . . .
Planar covers
• We speak about a planar cover if H is a finite planar graph.
s s ϕ(v1 ) = ϕ(v2 ) = v

s s s
v2
s s
s
→ v
s s s s
s
v1
H s s G = K5

Petr Hliněný, Eurocomb, Pisa, 2013 4 / 16 Planar Emulators Conjecture for Cubic . . .
Planar covers
• We speak about a planar cover if H is a finite planar graph.
s s ϕ(v1 ) = ϕ(v2 ) = v

s s s
v2
s s
s
→ v
s s s s
s
v1
H s s G = K5

• Graph embedded in the projective plane has a double planar cover,

via the universal covering map from the sphere onto the proj. plane.

Petr Hliněný, Eurocomb, Pisa, 2013 4 / 16 Planar Emulators Conjecture for Cubic . . .
Planar emulators
• ϕ : V(H) → V(G), an emulator vs. a cover:
. . . map the edges inc. with v in H surjectively
onto the edges inc. with ϕ(v) in G.

Petr Hliněný, Eurocomb, Pisa, 2013 5 / 16 Planar Emulators Conjecture for Cubic . . .
Planar emulators
• ϕ : V(H) → V(G), an emulator vs. a cover:
. . . map the edges inc. with v in H surjectively
onto the edges inc. with ϕ(v) in G.
c2 s s
b2

a3 s s a2 a s s
b

b3 s s →
s
c1

s s c
H c3 s b1 G = K3
a1

Petr Hliněný, Eurocomb, Pisa, 2013 5 / 16 Planar Emulators Conjecture for Cubic . . .
Planar emulators
• ϕ : V(H) → V(G), an emulator vs. a cover:
. . . map the edges inc. with v in H surjectively
onto the edges inc. with ϕ(v) in G.
c2 s s
b2

a3 s s a2 a s s
b

b3 s s →
s
c1

s s c
H c3 s b1 G = K3
a1

• Could a planar emulator be “more than” a planar cover?

Petr Hliněný, Eurocomb, Pisa, 2013 5 / 16 Planar Emulators Conjecture for Cubic . . .
Planar emulators
• ϕ : V(H) → V(G), an emulator vs. a cover:
. . . map the edges inc. with v in H surjectively
onto the edges inc. with ϕ(v) in G.
c2 s s
b2

a3 s s a2 a s s
b

b3 s s →
s
c1

s s c
H c3 s b1 G = K3
a1

• Could a planar emulator be “more than” a planar cover?

• Not really, at least until 2008. . .
Petr Hliněný, Eurocomb, Pisa, 2013 5 / 16 Planar Emulators Conjecture for Cubic . . .
2 Fellows’ planar emulator conjecture
Conjecture 1 (Negami, 1988)
A connected graph has a finite planar cover
⇐⇒ it embeds in the projective plane.

Petr Hliněný, Eurocomb, Pisa, 2013 6 / 16 Planar Emulators Conjecture for Cubic . . .
2 Fellows’ planar emulator conjecture
Conjecture 1 (Negami, 1988)
A connected graph has a finite planar cover
⇐⇒ it embeds in the projective plane.

Fact. A planar cover is also a planar emulator.

Why a planar emulator should be “more than” a planar cover?

Petr Hliněný, Eurocomb, Pisa, 2013 6 / 16 Planar Emulators Conjecture for Cubic . . .
2 Fellows’ planar emulator conjecture
Conjecture 1 (Negami, 1988)
A connected graph has a finite planar cover
⇐⇒ it embeds in the projective plane.

Fact. A planar cover is also a planar emulator.

Why a planar emulator should be “more than” a planar cover?

• We only use “more edges” – takes us farther away from planarity!

Petr Hliněný, Eurocomb, Pisa, 2013 6 / 16 Planar Emulators Conjecture for Cubic . . .
2 Fellows’ planar emulator conjecture
Conjecture 1 (Negami, 1988)
A connected graph has a finite planar cover
⇐⇒ it embeds in the projective plane.

Fact. A planar cover is also a planar emulator.

Why a planar emulator should be “more than” a planar cover?

• We only use “more edges” – takes us farther away from planarity!

• Until the end of 2008, most people perhaps considered planar emu-
lators just as a curious redefinition of planar covers. . .

Petr Hliněný, Eurocomb, Pisa, 2013 6 / 16 Planar Emulators Conjecture for Cubic . . .
2 Fellows’ planar emulator conjecture
Conjecture 1 (Negami, 1988)
A connected graph has a finite planar cover
⇐⇒ it embeds in the projective plane.

Fact. A planar cover is also a planar emulator.

Why a planar emulator should be “more than” a planar cover?

• We only use “more edges” – takes us farther away from planarity!

• Until the end of 2008, most people perhaps considered planar emu-
lators just as a curious redefinition of planar covers. . .

Conjecture 2 (Fellows, 1989)

A connected graph has a finite planar emulator
⇐⇒ it has a finite planar cover.

Petr Hliněný, Eurocomb, Pisa, 2013 6 / 16 Planar Emulators Conjecture for Cubic . . .
Slow progress, and a sudden surprise
Long-term development around Negami’s conjecture led to. . .

Theorem 3 (A+N+F+H, since 1998)

If K1,2,2,2 had no finite planar cover, then Negami’s conj. would be proved.

Petr Hliněný, Eurocomb, Pisa, 2013 7 / 16 Planar Emulators Conjecture for Cubic . . .
Slow progress, and a sudden surprise
Long-term development around Negami’s conjecture led to. . .

Theorem 3 (A+N+F+H, since 1998)

If K1,2,2,2 had no finite planar cover, then Negami’s conj. would be proved.

. . . and then. . .

Petr Hliněný, Eurocomb, Pisa, 2013 7 / 16 Planar Emulators Conjecture for Cubic . . .
Slow progress, and a sudden surprise
Long-term development around Negami’s conjecture led to. . .

Theorem 3 (A+N+F+H, since 1998)

If K1,2,2,2 had no finite planar cover, then Negami’s conj. would be proved.

. . . and then. . . Suddenly, Fellows’ conjecture falls down. . .

Fact. The graph K4,5 −4K2 has no finite planar cover.

Theorem 4 (Rieck and Yamashita 2008)

The graphs K1,2,2,2 and K4,5 −4K2 do have finite planar emulators!!!

Petr Hliněný, Eurocomb, Pisa, 2013 7 / 16 Planar Emulators Conjecture for Cubic . . .
Slow progress, and a sudden surprise
Long-term development around Negami’s conjecture led to. . .

Theorem 3 (A+N+F+H, since 1998)

If K1,2,2,2 had no finite planar cover, then Negami’s conj. would be proved.

. . . and then. . . Suddenly, Fellows’ conjecture falls down. . .

Fact. The graph K4,5 −4K2 has no finite planar cover.

Theorem 4 (Rieck and Yamashita 2008)

The graphs K1,2,2,2 and K4,5 −4K2 do have finite planar emulators!!!

• Now we know that the class of graphs having finite planar emulators
– is different from the class of graphs having finite planar covers,
– and different from the class of projective planar graphs, too.
• So, let us study this class. . .
Petr Hliněný, Eurocomb, Pisa, 2013 7 / 16 Planar Emulators Conjecture for Cubic . . .
K4,5 −4K2

(A picture by Yamashita.)

Petr Hliněný, Eurocomb, Pisa, 2013 8 / 16 Planar Emulators Conjecture for Cubic . . .
3 What graphs do have planar emulators?
Recall. . .

K3,3 ·K3,3 K5 · K3,3 K5 · K5 B3 C2 C7

D1 D4 D9 D12 D17 E6 E 11

E 19 E 20 E 27 F4 F6 G1

K3,5 K4,5−4K2 K4,4 −e K7−C4 D3 E5 F1

K1,2,2,2 B7 C3 C4 D2 E2
Petr Hliněný, Eurocomb, Pisa, 2013 9 / 16 Planar Emulators Conjecture for Cubic . . .
Among the projective list

E 20 K3,5
NO emulators (proved)
• the case of “two disjoint k-graphs”,

Petr Hliněný, Eurocomb, Pisa, 2013 10 / 16 Planar Emulators Conjecture for Cubic . . .
Among the projective list

E 20 K3,5
NO emulators (proved)
• the case of “two disjoint k-graphs”,
• a sporadic proof for K3,5 extends as well (nontrivial),
but none of the other proofs from covers works for emulators.

Petr Hliněný, Eurocomb, Pisa, 2013 10 / 16 Planar Emulators Conjecture for Cubic . . .
Among the projective list

E 20 K3,5
NO emulators (proved)
• the case of “two disjoint k-graphs”,
• a sporadic proof for K3,5 extends as well (nontrivial),
but none of the other proofs from covers works for emulators.
YES emulators
• all projective-planar graphs, but those are the trivial ones,

Petr Hliněný, Eurocomb, Pisa, 2013 10 / 16 Planar Emulators Conjecture for Cubic . . .
Among the projective list

E 20 K3,5
NO emulators (proved)
• the case of “two disjoint k-graphs”,
• a sporadic proof for K3,5 extends as well (nontrivial),
but none of the other proofs from covers works for emulators.
YES emulators
• all projective-planar graphs, but those are the trivial ones,
• K1,2,2,2 and K4,5 −4K2 by [Rieck and Yamashita, 2008],

Petr Hliněný, Eurocomb, Pisa, 2013 10 / 16 Planar Emulators Conjecture for Cubic . . .
Among the projective list

E 20 K3,5
NO emulators (proved)
• the case of “two disjoint k-graphs”,
• a sporadic proof for K3,5 extends as well (nontrivial),
but none of the other proofs from covers works for emulators.
YES emulators
• all projective-planar graphs, but those are the trivial ones,
• K1,2,2,2 and K4,5 −4K2 by [Rieck and Yamashita, 2008],
• C 4 and E 2 by [PH and Chimani, 2009],
and hence consequently the whole rich “family of K1,2,2,2 ”,

Petr Hliněný, Eurocomb, Pisa, 2013 10 / 16 Planar Emulators Conjecture for Cubic . . .
Among the projective list

E 20 K3,5
NO emulators (proved)
• the case of “two disjoint k-graphs”,
• a sporadic proof for K3,5 extends as well (nontrivial),
but none of the other proofs from covers works for emulators.
YES emulators
• all projective-planar graphs, but those are the trivial ones,
• K1,2,2,2 and K4,5 −4K2 by [Rieck and Yamashita, 2008],
• C 4 and E 2 by [PH and Chimani, 2009],
and hence consequently the whole rich “family of K1,2,2,2 ”,
• and K7 −C4 and its whole family by [Klusáček, 2011].
Petr Hliněný, Eurocomb, Pisa, 2013 10 / 16 Planar Emulators Conjecture for Cubic . . .
Graphically

Petr Hliněný, Eurocomb, Pisa, 2013 11 / 16 Planar Emulators Conjecture for Cubic . . .
4 The cubic case
• Although we do not much understand the whole class of non-
projective planar-emulable graphs
— is it essentially finite or infinite ? —

Petr Hliněný, Eurocomb, Pisa, 2013 12 / 16 Planar Emulators Conjecture for Cubic . . .
4 The cubic case
• Although we do not much understand the whole class of non-
projective planar-emulable graphs
— is it essentially finite or infinite ? —
• it appears significant that no such cubic graph has been found.

Petr Hliněný, Eurocomb, Pisa, 2013 12 / 16 Planar Emulators Conjecture for Cubic . . .
4 The cubic case
• Although we do not much understand the whole class of non-
projective planar-emulable graphs
— is it essentially finite or infinite ? —
• it appears significant that no such cubic graph has been found.
• We can thus use this easier ground to perhaps train our techniques
before attacking the full problem. . .

Theorem. If a cubic nonprojective graph H has a finite planar emulator,

then H is a planar expansion of one of the following two graphs:

Petr Hliněný, Eurocomb, Pisa, 2013 12 / 16 Planar Emulators Conjecture for Cubic . . .
Proof:
G1 G2

• The two graphs – [Glover and Huneke, 1975]

– two out of all six cubic obstructions for projective embeddability;
while the other four have two disjoint k-graphs.

Petr Hliněný, Eurocomb, Pisa, 2013 13 / 16 Planar Emulators Conjecture for Cubic . . .
Proof:
G1 G2

• The two graphs – [Glover and Huneke, 1975]

– two out of all six cubic obstructions for projective embeddability;
while the other four have two disjoint k-graphs.
• Any nonproj. planar emul. graph H must contain a subdiv. of those.

Petr Hliněný, Eurocomb, Pisa, 2013 13 / 16 Planar Emulators Conjecture for Cubic . . .
Proof:
G1 G2

• The two graphs – [Glover and Huneke, 1975]

– two out of all six cubic obstructions for projective embeddability;
while the other four have two disjoint k-graphs.
• Any nonproj. planar emul. graph H must contain a subdiv. of those.

• Cubic H ⇒ H = Gi + bridges
where all the bridge legs subdivide the edges of Gi .

Petr Hliněný, Eurocomb, Pisa, 2013 13 / 16 Planar Emulators Conjecture for Cubic . . .
Proof:
G1 G2

• The two graphs – [Glover and Huneke, 1975]

– two out of all six cubic obstructions for projective embeddability;
while the other four have two disjoint k-graphs.
• Any nonproj. planar emul. graph H must contain a subdiv. of those.

• Cubic H ⇒ H = Gi + bridges
where all the bridge legs subdivide the edges of Gi .
• Even a single bridge having legs on non-incident edges of Gi
⇒ two disjoint k-graphs or a K3,5 minor. [computer] OK

Petr Hliněný, Eurocomb, Pisa, 2013 13 / 16 Planar Emulators Conjecture for Cubic . . .
All bridges trivial (i.e., on incident legs)
H = Gi + bridges

Petr Hliněný, Eurocomb, Pisa, 2013 14 / 16 Planar Emulators Conjecture for Cubic . . .
All bridges trivial (i.e., on incident legs)
H = Gi + bridges

• Assign each trivial bridge to its Gi -vertex or edge. . .

Petr Hliněný, Eurocomb, Pisa, 2013 14 / 16 Planar Emulators Conjecture for Cubic . . .
All bridges trivial (i.e., on incident legs)
H = Gi + bridges

• Assign each trivial bridge to its Gi -vertex or edge. . .

• Must handle overlapping – conflict of bridges:

Petr Hliněný, Eurocomb, Pisa, 2013 14 / 16 Planar Emulators Conjecture for Cubic . . .
All bridges trivial (i.e., on incident legs)
H = Gi + bridges

• Assign each trivial bridge to its Gi -vertex or edge. . .

• Must handle overlapping – conflict of bridges:

Any conflict ⇒ another subdivision gives a non-trivial bridge.

Petr Hliněný, Eurocomb, Pisa, 2013 14 / 16 Planar Emulators Conjecture for Cubic . . .
All bridges trivial (i.e., on incident legs)
H = Gi + bridges

• Assign each trivial bridge to its Gi -vertex or edge. . .

• Must handle overlapping – conflict of bridges:

Any conflict ⇒ another subdivision gives a non-trivial bridge.

• Hence the assign. of bridges to vert./edges of Gi is rigorous. OK

Petr Hliněný, Eurocomb, Pisa, 2013 14 / 16 Planar Emulators Conjecture for Cubic . . .
Trivial bridges are local
H = Gi + bridges

Petr Hliněný, Eurocomb, Pisa, 2013 15 / 16 Planar Emulators Conjecture for Cubic . . .
Trivial bridges are local
H = Gi + bridges

• Non-conflicting trivial bridges at any given vertex;

Petr Hliněný, Eurocomb, Pisa, 2013 15 / 16 Planar Emulators Conjecture for Cubic . . .
Trivial bridges are local
H = Gi + bridges

• Non-conflicting trivial bridges at any given vertex;

either planar expansion (left), or K2,3 in the fragment (twice right),

Petr Hliněný, Eurocomb, Pisa, 2013 15 / 16 Planar Emulators Conjecture for Cubic . . .
Trivial bridges are local
H = Gi + bridges

• Non-conflicting trivial bridges at any given vertex;

either planar expansion (left), or K2,3 in the fragment (twice right),

⇒ two disjoint k-graphs. OK

• End of proof. 2

Petr Hliněný, Eurocomb, Pisa, 2013 15 / 16 Planar Emulators Conjecture for Cubic . . .
5 Conclusion
• Resolve the two cubic graphs (everything else is now finished!).

Petr Hliněný, Eurocomb, Pisa, 2013 16 / 16 Planar Emulators Conjecture for Cubic . . .
5 Conclusion
• Resolve the two cubic graphs (everything else is now finished!).

• Any idea for a new hypothesis concerning (non-cubic) graphs?

Petr Hliněný, Eurocomb, Pisa, 2013 16 / 16 Planar Emulators Conjecture for Cubic . . .
5 Conclusion
• Resolve the two cubic graphs (everything else is now finished!).

• Any idea for a new hypothesis concerning (non-cubic) graphs?

• Prove that the required fold number is finite for planar emulators?

Petr Hliněný, Eurocomb, Pisa, 2013 16 / 16 Planar Emulators Conjecture for Cubic . . .
5 Conclusion
• Resolve the two cubic graphs (everything else is now finished!).

• Any idea for a new hypothesis concerning (non-cubic) graphs?

• Prove that the required fold number is finite for planar emulators?

• And, of course, do not forget still open Negami’s conjecture!

Petr Hliněný, Eurocomb, Pisa, 2013 16 / 16 Planar Emulators Conjecture for Cubic . . .
5 Conclusion
• Resolve the two cubic graphs (everything else is now finished!).

• Any idea for a new hypothesis concerning (non-cubic) graphs?

• Prove that the required fold number is finite for planar emulators?

• And, of course, do not forget still open Negami’s conjecture!

• Lastly, a structural question – for which graph class G;


G-emulable/coverable =G ?

Petr Hliněný, Eurocomb, Pisa, 2013 16 / 16 Planar Emulators Conjecture for Cubic . . .
5 Conclusion
• Resolve the two cubic graphs (everything else is now finished!).

• Any idea for a new hypothesis concerning (non-cubic) graphs?

• Prove that the required fold number is finite for planar emulators?

• And, of course, do not forget still open Negami’s conjecture!

• Lastly, a structural question – for which graph class G;


G-emulable/coverable =G ?

– Holds true for G = outerplanar (so, k-outerplanar?),

Petr Hliněný, Eurocomb, Pisa, 2013 16 / 16 Planar Emulators Conjecture for Cubic . . .
5 Conclusion
• Resolve the two cubic graphs (everything else is now finished!).

• Any idea for a new hypothesis concerning (non-cubic) graphs?

• Prove that the required fold number is finite for planar emulators?

• And, of course, do not forget still open Negami’s conjecture!

• Lastly, a structural question – for which graph class G;


G-emulable/coverable =G ?

– Holds true for G = outerplanar (so, k-outerplanar?),

– Negami ⇐⇒ true for G =projective (and so false with emul.),
– other classes, e.g., G = other nonorientable surface?

Petr Hliněný, Eurocomb, Pisa, 2013 16 / 16 Planar Emulators Conjecture for Cubic . . .