Title: CSCI 4260MATH 4150
1CSCI 4260-MATH 4150
2Embedding graphs on surfaces
- Last lecture
- Graph G is planar G can be drawn on the plane in
a way that no two edges cross. - This drawing is called an embedding of G in the
plane - Today
- Embeddings of non-planar graphs
- If G is non-planar, it can not be embedded in the
plane
3Intuition
- Lets look at K(5), which we know is non-planar
4The main idea
- We will start with a basic surface
- Add handles as necessary
- The basic surface can be the plane
- But it is easier to visualize certain properties
on the sphere - In fact, in terms of embedability the plane and
sphere is equivalent
5Plane-sphere equivalence
- Embedding on the plane ? embedding on the sphere
- How about the other direction
- Can you convert an embedding on the sphere to an
embedding on the plane
6From http//www.heartofmath.com/first_edition/pdf
s/pg360.pdf
7Torus a sphere with a handle
8Constructing a torus
- http//www.cmis.brighton.ac.uk/jt40/MapleAnimatio
ns/Torus.html - Local copy
9Back to K(5)
10Surface genus number of handles
- k-handles
- Surface of genus k
- S(k)
- S(0) sphere
- S(1) torus
11Embedding graphs
- Given G, draw it on the sphere
- If two edges cross, add a handle
- Note we can draw the graph in a way that only
two edges cross at any intersection
12Genus of a graph G
- The smallest k such that G can be embedded on
S(k) - Lets see if/how the properties of planar graphs
generalize
13Remember Euler Equation
- n-mr 2 for connected, planar graphs
- In other words, number of regions is the same for
any planar embedding
14Three embeddings of K(4) on a torus
ERROR IN THE BOOK
? regions
4 regions
? regions
15Number of regions
- Depends on the embedding
- The problem is that there is too much variety
- We need the notion of a canonical embedding
162-cells
- A region is called a 2-cell if any closed curve
that is drawn inside the region can be
continuously shrunk to a single point
17Euler Equation
R2
R1
n 3, m 3, r 2 n-mr 2 Both R1 and R2 are
2-cell. Why?
18Euler Equation for diconnected planar graphs?
R3
R1
R2
n 6, m 6, r 3 n-mr 3 the equation is not
true. Is R3 a 2-cell region?
19Which of these embeddings of K(4) is 2-cell?
20Theorem
- Let G be a connected graph that is 2-cell
embedded on a surface of genus k. Then, - n m r 2 2k
- where n orderm orderr number of regions
21Proof (by induction on k)
- Basis k 0.
- G must be planar
- n m r 2 2k
- n m r 2
- Eulers equation
22Inductive step
- Suppose the theorem is true for all graphs that
can be 2-cell embedded on surfaces with k or less
handles - Let G be a graph 2-cell embedded on S(k1)
- Let H be one of the handles
23A closer look at H
- We can always embed in a way that there are no
vertices on handles - Case 1 there are no edges on the handle.
- (ind. hyp. holds)
- Now suppose there are edges along the handle
24A closer look at H
- Draw a curve C around H
- Let t be the of intersections with edges
- Insert a new vertex at every intersection
- Insert two more vertices to prevent parallel
edges or loops
C
25We now a have a new graph
- G1
- n1 n t 2
- m1 m 2t 2
- r1 r t
C
26Now cut the handle
- G1
- n1 n t 2
- m1 m 2t 2
- r1 r t
- G2
- n2 n1 t 2
- m2 m1 t 2
- r2 r1 2
27We can now apply the inductive hypothesis
- (n2t4) (m3t4) (r t 2) 2 2k
- n m r 2 2 - 2k
- n m r 2 2(k1)
- done.
28Recall the genus of a graph
- ??(G) the min. k such that G can be embedded in
S(k) - Theorem (w/o proof)
- If we embed G on S(?(G)), the resulting embedding
is 2-cell.
29Corollary
- ? Let G be a connected graph that is embedded on
a surface of genus ?(G). Then, - n m r 2 2 ?(G)
30Recall
- For planar graphs we have m 3n 6
- Similarly, we have
31Corollary
- If G is a connected graph of order n 3 and size
m, then
32Proof
- Similar to the planar case, we need to get rid of
r from the equation - n m r 2 2 ?(G)
- Let
- R1,, Rr be the regions, and
- m1,,mr be the number of edges on the boundaries
of respective regions.
33Proof (cont)
- We have
- mi 3
- Every edge is counted at most twice
- Which gives us 3r 2m
34Using 3r 2m
- 2 2 ?(G) n m r
- 6 6 ?(G) 3n 3m 3r 3n 3m 2m
3n m - Rearranging, we get
35A final remark
- Recall the argument about starting with a
embedding on the sphere and adding handles at the
intersection - This suggests that the genus of a graph of order
m is O(n2) - Can it ?be ?(n2)?
36Theorem
37A note
- I am traveling this week until Sunday
- No office hour on Thursday
- I am maybe slow in responding to email until
Sunday Night - Nikhil will hold his regular office hours and
review matchings on Thursday lecture.