• No results found

A solution to Alspach’s problem for complete graphs of large odd order

N/A
N/A
Protected

Academic year: 2021

Share "A solution to Alspach’s problem for complete graphs of large odd order"

Copied!
69
0
0

Loading.... (view fulltext now)

Full text

(1)

A solution to Alspach’s problem for complete graphs of large odd order

Daniel Horsley

Joint work with Darryn Bryant (University of Queensland)

(2)

Cycle Decompositions

(3)

Cycle Decompositions

(4)

Cycle Decompositions of Complete Graphs

(4, 4, 6, 7)-decomposition of K

7

(5)

Cycle Decompositions of Complete Graphs

(4, 4, 6, 7)-decomposition of K

7

(6)

If there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

then (1) n is odd;

(2) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and (3) m

1

+ m

2

+ · · · + m

t

= |E (K

n

)|.

Alspach’s Cycle Decomposition Problem (1981): Prove (1), (2) and (3) are also sufficient for an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

.

Alspach also posed the equivalent problem for K

n

− I when n is even.

If a list satisfies (2) and (3) for some odd n we call it n-admissible.

(7)

If there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

then (1) n is odd;

(2) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and (3) m

1

+ m

2

+ · · · + m

t

= |E (K

n

)|.

Alspach’s Cycle Decomposition Problem (1981): Prove (1), (2) and (3) are also sufficient for an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

.

Alspach also posed the equivalent problem for K

n

− I when n is even.

If a list satisfies (2) and (3) for some odd n we call it n-admissible.

(8)

If there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

then (1) n is odd;

(2) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and (3) m

1

+ m

2

+ · · · + m

t

= |E (K

n

)|.

Alspach’s Cycle Decomposition Problem (1981): Prove (1), (2) and (3) are also sufficient for an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

.

Alspach also posed the equivalent problem for K

n

− I when n is even.

If a list satisfies (2) and (3) for some odd n we call it n-admissible.

(9)

If there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

then (1) n is odd;

(2) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and (3) m

1

+ m

2

+ · · · + m

t

= |E (K

n

)|.

Alspach’s Cycle Decomposition Problem (1981): Prove (1), (2) and (3) are also sufficient for an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

.

Alspach also posed the equivalent problem for K

n

− I when n is even.

If a list satisfies (2) and (3) for some odd n we call it n-admissible.

(10)

I think this is a hard problem

I

The cycles can have as many as n vertices.

I

We have to find a very large number of decompositions for each value of n

(11)

I think this is a hard problem

I

The cycles can have as many as n vertices.

I

We have to find a very large number of decompositions for each value of n

(12)

Existing Results

There exists a (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

when (1), (2), (3) and any of the following hold.

I

m

1

= m

2

= · · · = m

t

Alspach, Gavlas, ˇ Sajna

I

each m

i

∈ {3, 4, 5} Balister

I

each m

i

∈ {3, n} Bryant, Maenhaut

I

each m

i

∈ {2

k

, 2

k+1

} for some k ≥ 2 Heinrich, Hor´ ak, Rosa

I

n is large and each m

i

≤ b

n−11220

c Balister

I

etc.

I

each m

i

n+52

Bryant, Horsley

I

each m

i

n−12

and m

t

≤ 2m

t−1

Bryant, Horsley

(13)

Existing Results

There exists a (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

when (1), (2), (3) and any of the following hold.

I

m

1

= m

2

= · · · = m

t

Alspach, Gavlas, ˇ Sajna

I

each m

i

∈ {3, 4, 5} Balister

I

each m

i

∈ {3, n} Bryant, Maenhaut

I

each m

i

∈ {2

k

, 2

k+1

} for some k ≥ 2 Heinrich, Hor´ ak, Rosa

I

n is large and each m

i

≤ b

n−11220

c Balister

I

etc.

I

each m

i

n+52

Bryant, Horsley

I

each m

i

n−12

and m

t

≤ 2m

t−1

Bryant, Horsley

(14)

Existing Results

There exists a (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

when (1), (2), (3) and any of the following hold.

I

m

1

= m

2

= · · · = m

t

Alspach, Gavlas, ˇ Sajna

I

each m

i

∈ {3, 4, 5} Balister

I

each m

i

∈ {3, n} Bryant, Maenhaut

I

each m

i

∈ {2

k

, 2

k+1

} for some k ≥ 2 Heinrich, Hor´ ak, Rosa

I

n is large and each m

i

≤ b

n−11220

c Balister

I

etc.

I

each m

i

n+52

Bryant, Horsley

I

each m

i

n−12

and m

t

≤ 2m

t−1

Bryant, Horsley

(15)

Our Result

Theorem: There is an integer N such that for all odd n ≥ N there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

if and only if

(1) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and

(2) m

1

+ m

2

+ · · · + m

t

=

n2

.

(16)

Our Result

Theorem: There is an integer N such that for all odd n ≥ N there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

if and only if

(1) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and

(2) m

1

+ m

2

+ · · · + m

t

=

n2

.

(17)

Equalise Cycles Lemma

Suppose there exists an (m

1

, m

2

, . . . , m

t

, x , y )-decomposition of K

n

where x < y and x + y ≥ n + 2. Then there exists an

(m

1

, m

2

, . . . , m

t

, x + 1, y − 1)-decomposition of K

n

.

Example: (m

1

, m

2

, . . . , m

t

, 10, 15)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 11, 14)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 12, 13)-decomposition of K

23

(18)

Equalise Cycles Lemma

Suppose there exists an (m

1

, m

2

, . . . , m

t

, x , y )-decomposition of K

n

where x < y and x + y ≥ n + 2. Then there exists an

(m

1

, m

2

, . . . , m

t

, x + 1, y − 1)-decomposition of K

n

.

Example: (m

1

, m

2

, . . . , m

t

, 10, 15)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 11, 14)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 12, 13)-decomposition of K

23

(19)

Equalise Cycles Lemma

Suppose there exists an (m

1

, m

2

, . . . , m

t

, x , y )-decomposition of K

n

where x < y and x + y ≥ n + 2. Then there exists an

(m

1

, m

2

, . . . , m

t

, x + 1, y − 1)-decomposition of K

n

.

Example: (m

1

, m

2

, . . . , m

t

, 10, 15)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 11, 14)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 12, 13)-decomposition of K

23

(20)

Join Cycles Lemma

Suppose there exists an (m

1

, m

2

, . . . , m

t

, c, x , y )-decomposition of K

n

where (1) c ≥

12

(x + y ); and

(2) x + y + c ≤ n + 1.

Then there exists an (m

1

, m

2

, . . . , m

t

, c, x + y )-decomposition of K

n

.

Example: (m

1

, m

2

, . . . , m

t

, 10, 7, 6)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 10, 13)-decomposition of K

23

(21)

Join Cycles Lemma

Suppose there exists an (m

1

, m

2

, . . . , m

t

, c, x , y )-decomposition of K

n

where (1) c ≥

12

(x + y ); and

(2) x + y + c ≤ n + 1.

Then there exists an (m

1

, m

2

, . . . , m

t

, c, x + y )-decomposition of K

n

.

Example: (m

1

, m

2

, . . . , m

t

, 10, 7, 6)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 10, 13)-decomposition of K

23

(22)

Join Cycles Lemma

Suppose there exists an (m

1

, m

2

, . . . , m

t

, c, x , y )-decomposition of K

n

where (1) c ≥

12

(x + y ); and

(2) x + y + c ≤ n + 1.

Then there exists an (m

1

, m

2

, . . . , m

t

, c, x + y )-decomposition of K

n

.

Example: (m

1

, m

2

, . . . , m

t

, 10, 7, 6)-decomposition of K

23

⇒ (m

1

, m

2

, . . . , m

t

, 10, 13)-decomposition of K

23

(23)

A Reduction of Alspach’s Problem

m

1

, m

2

, . . . , m

t

, x , y

3 < x ≤ y < n, x + y ≥ n + 2

m

1

, m

2

, . . . , m

t

, x − 1, y + 1

⇑ .. .

m

1

, m

2

, . . . , m

s−1

, m

s

, c, n, n, . . . , n

mi ≤ c, mi+ c ≤ n + 1

m

1

, m

2

, . . . , m

s−1

, 3, m

s

− 3, c, n, n, . . . , n

(if ms∈ {3, 4, 5})/

⇑ .. .

3, 3, . . . , 3, 4, 4, . . . , 4, 5, 5, . . . , 5, c, n, n, . . . , n

(24)

A Reduction of Alspach’s Problem

m

1

, m

2

, . . . , m

t

, x , y

3 < x ≤ y < n, x + y ≥ n + 2

m

1

, m

2

, . . . , m

t

, x − 1, y + 1

⇑ .. .

m

1

, m

2

, . . . , m

s−1

, m

s

, c, n, n, . . . , n

mi ≤ c, mi+ c ≤ n + 1

m

1

, m

2

, . . . , m

s−1

, 3, m

s

− 3, c, n, n, . . . , n

(if ms∈ {3, 4, 5})/

⇑ .. .

3, 3, . . . , 3, 4, 4, . . . , 4, 5, 5, . . . , 5, c, n, n, . . . , n

(25)

A Reduction of Alspach’s Problem

m

1

, m

2

, . . . , m

t

, x , y

3 < x ≤ y < n, x + y ≥ n + 2

m

1

, m

2

, . . . , m

t

, x − 1, y + 1

⇑ .. .

m

1

, m

2

, . . . , m

s−1

, m

s

, c, n, n, . . . , n

mi ≤ c, mi+ c ≤ n + 1

m

1

, m

2

, . . . , m

s−1

, 3, m

s

− 3, c, n, n, . . . , n

(if ms∈ {3, 4, 5})/

⇑ .. .

3, 3, . . . , 3, 4, 4, . . . , 4, 5, 5, . . . , 5, c, n, n, . . . , n

(26)

A Reduction of Alspach’s Problem

m

1

, m

2

, . . . , m

t

, x , y

3 < x ≤ y < n, x + y ≥ n + 2

m

1

, m

2

, . . . , m

t

, x − 1, y + 1

⇑ .. .

m

1

, m

2

, . . . , m

s−1

, m

s

, c, n, n, . . . , n

mi ≤ c, mi+ c ≤ n + 1

m

1

, m

2

, . . . , m

s−1

, 3, m

s

− 3, c, n, n, . . . , n

(if ms∈ {3, 4, 5})/

⇑ .. .

3, 3, . . . , 3, 4, 4, . . . , 4, 5, 5, . . . , 5, c, n, n, . . . , n

(27)

A Reduction of Alspach’s Problem

m

1

, m

2

, . . . , m

t

, x , y

3 < x ≤ y < n, x + y ≥ n + 2

m

1

, m

2

, . . . , m

t

, x − 1, y + 1

⇑ .. .

m

1

, m

2

, . . . , m

s−1

, m

s

, c, n, n, . . . , n

mi ≤ c, mi+ c ≤ n + 1

m

1

, m

2

, . . . , m

s−1

, 3, m

s

− 3, c, n, n, . . . , n

(if ms∈ {3, 4, 5})/

⇑ .. .

3, 3, . . . , 3, 4, 4, . . . , 4, 5, 5, . . . , 5, c, n, n, . . . , n

(28)

A Reduction of Alspach’s Problem

n-ancestor lists: n-admissible lists of the form

3, 3, . . . , 3, 4, 4, . . . , 4, 5, 5, . . . , 5, c, n, n, . . . , n

Theorem: If an (m

01

, m

20

, . . . , m

0t

)-decomposition of K

n

exists for each n-ancestor list m

10

, m

02

, . . . , m

0t

, then an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

exists for each n-admissible list m

1

, m

2

, . . . , m

t

.

Note: Everything I’ve said so far is also true for K

n

− I when n is even.

(29)

A Reduction of Alspach’s Problem

n-ancestor lists: n-admissible lists of the form

3, 3, . . . , 3, 4, 4, . . . , 4, 5, 5, . . . , 5, c, n, n, . . . , n

Theorem: If an (m

01

, m

20

, . . . , m

0t

)-decomposition of K

n

exists for each n-ancestor list m

10

, m

02

, . . . , m

0t

, then an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

exists for each n-admissible list m

1

, m

2

, . . . , m

t

.

Note: Everything I’ve said so far is also true for K

n

− I when n is even.

(30)

Main Result

Lemma: There is an integer N such that for all odd n ≥ N there exists an (m

01

, m

02

, . . . , m

0t

)-decomposition of K

n

for each n-ancestor list m

10

, m

02

, . . . , m

0t

. Proof: Fiddly and painful. Uses difference methods.

Theorem: There is an integer N such that for all odd n ≥ N there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

if and only if

(1) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and

(2) m

1

+ m

2

+ · · · + m

t

=

n2

.

(31)

Main Result

Lemma: There is an integer N such that for all odd n ≥ N there exists an (m

01

, m

02

, . . . , m

0t

)-decomposition of K

n

for each n-ancestor list m

10

, m

02

, . . . , m

0t

.

Proof: Fiddly and painful. Uses difference methods.

Theorem: There is an integer N such that for all odd n ≥ N there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

if and only if

(1) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and

(2) m

1

+ m

2

+ · · · + m

t

=

n2

.

(32)

Main Result

Lemma: There is an integer N such that for all odd n ≥ N there exists an (m

01

, m

02

, . . . , m

0t

)-decomposition of K

n

for each n-ancestor list m

10

, m

02

, . . . , m

0t

. Proof: Fiddly and painful. Uses difference methods.

Theorem: There is an integer N such that for all odd n ≥ N there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

if and only if

(1) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and

(2) m

1

+ m

2

+ · · · + m

t

=

n2

.

(33)

Main Result

Lemma: There is an integer N such that for all odd n ≥ N there exists an (m

01

, m

02

, . . . , m

0t

)-decomposition of K

n

for each n-ancestor list m

10

, m

02

, . . . , m

0t

. Proof: Fiddly and painful. Uses difference methods.

Theorem: There is an integer N such that for all odd n ≥ N there exists an (m

1

, m

2

, . . . , m

t

)-decomposition of K

n

if and only if

(1) 3 ≤ m

1

, m

2

, . . . , m

t

≤ n; and

(2) m

1

+ m

2

+ · · · + m

t

=

n2

.

(34)

Packings and Leaves

(6, 5, 3)−packing of K

7

(35)

Packings and Leaves

(6, 5, 3)−packing of K

7

(36)

Packings and Leaves

leave

(6, 5, 3)−packing of K

7

(37)

Proof of Equalise Lemma

Equalise Lemma:

Suppose there exists an (m

1

, m

2

, . . . , m

t

, x , y )-decomposition of K

n

where x < y and x + y ≥ n + 2. Then there exists an

(m

1

, m

2

, . . . , m

t

, x + 1, y − 1)-decomposition of K

n

.

(38)

Proof of Equalise Lemma

Equalise Lemma:

Suppose there exists an (m

1

, m

2

, . . . , m

t

, x , y )-decomposition of K

n

where x < y and x + y ≥ n + 2. Then there exists an

(m

1

, m

2

, . . . , m

t

, x + 1, y − 1)-decomposition of K

n

.

(39)

Proof of Equalise Lemma

Omit the x −cycle and y −cycle to form an (m

1

, m

2

, . . . , m

t

)−packing of K

n

. Picture the leave with the (long) y −cycle “outside” and the (short) x −cycle

“inside”.

Example:

(40)

Proof of Equalise Lemma

A sublemma:

(41)

Proof of Equalise Lemma

A sublemma:

inside + 1

outside - 1

(42)

Proof of Equalise Lemma

Theorem (Thomason): If a graph G has a decomposition into two cycles which share at least two vertices then there is another decomposition of G into two cycles which is distinct from the first.

Example:

(43)

Proof of Equalise Lemma

Theorem (Thomason): If a graph G has a decomposition into two cycles which share at least two vertices then there is another decomposition of G into two cycles which is distinct from the first.

Example:

(44)

Proof of Equalise Lemma

Theorem (Thomason): If a graph G has a decomposition into two cycles which share at least two vertices then there is another decomposition of G into two cycles which is distinct from the first.

Example:

green outside

inside red

(45)

Proof of Equalise Lemma

Theorem (Thomason): If a graph G has a decomposition into two cycles which share at least two vertices then there is another decomposition of G into two cycles which is distinct from the first.

Example:

outside green

red inside

(46)

Proof of Equalise Lemma

(47)

Proof of Equalise Lemma

(48)

Proof of Equalise Lemma

(49)

Proof of Equalise Lemma

(50)

Proof of Equalise Lemma

(51)

Proof of Equalise Lemma

(52)

Proof of Equalise Lemma

(53)

Proof of Equalise Lemma

(54)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

(55)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

green outside

inside red

(56)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

green outside

inside red

(57)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside green

inside red

(58)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside green

red inside

(59)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside green

red inside

(60)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside

green

red

inside

(61)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside

green

red

inside

(62)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside

green red

inside

(63)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside

red green

inside

(64)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside

red

green

inside

(65)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside

red

green

inside

(66)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside red

green inside

(67)

Proof of Equalise Lemma

So if red < outside we can get either

I

inside + 1, outside − 1; or

I

red + 1,

green

− 1.

outside red

green inside

(68)

Equalise Lemma

Suppose there exists an (m

1

, m

2

, . . . , m

t

, x , y )-decomposition of K

n

where x < y and x + y ≥ n + 2. Then there exists an

(m

1

, m

2

, . . . , m

t

, x + 1, y − 1)-decomposition of K

n

.

Proved!

(69)

The End

References

Related documents

For a Shannon board state represented by a graph G, the vertices of the Shannon reduced graph G  are the terminals and uncoloured vertices of G, with two vertices adjacent in G 

• after reweighting (assuming correctly classified points get zero

• dot-product kernels usually lead to convex learning problems. Usually you loose this guarantee for

we will consider general (not only convex) constrained optimization problems, start by case with only equalities Theorem: consider the problem. where the constraint gradients h i

Theorem 2 Given compatible combinatorial embeddings for two graphs G 1 and G 2 whose shared graph is connected and has n vertices, we can find planar drawings for G 1 and G 2

The coprime graph of a group G, denoted by Γ G , is a graph whose vertices are elements of G and two distinct vertices x and y are adjacent if and only if ( |x|, |y|) = 1.. In

The blocktransformation graph of a graph G is the graph, whose vertex set is the union of vertices and blocks of G, in which two vertices are adjacent whenever the

The Gallai fuzzy graph Γ(G) of a fuzzy graph G has the fuzzy edges of G as its fuzzy vertices and two distinct fuzzy edges of G are fuzzy incident in G, but do not span a fuzzy