61
Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture Algebraic combinatorics applied to finite geometry John Bamberg Centre for the Mathematics of Symmetry and Computation, The University of Western Australia December 1, 2011 , =2

Algebraic combinatorics applied to finite geometry - SymOmega

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Algebraic combinatorics applied tofinite geometry

John Bamberg

Centre for the Mathematics of Symmetry and Computation,The University of Western Australia

December 1, 2011

,

= 2

Page 2: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Graphs vs Linear algebra

Euler’s Theorem on latin squares

Finite geometry

The bigger picture

Page 3: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Graphs

• Vertices

• Edges: pairs of vertices (u, v)

Degree: 3

Page 4: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Graphs

• Vertices

• Edges: pairs of vertices (u, v)

Degree: 3

Page 5: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Good labellings

• Assign a real number to each vertex.

• For each vertex, sum the values of adjacent vertices.

• Goal: Sum at each vertex should be a common multiple of thevalue at the vertex.

?1

?

1?

1

?

1

?

1

Page 6: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Good labellings

• Assign a real number to each vertex.

• For each vertex, sum the values of adjacent vertices.

• Goal: Sum at each vertex should be a common multiple of thevalue at the vertex.

?1

?

1?

1

?

1

?

1

Page 7: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Good labellings

• Assign a real number to each vertex.

• For each vertex, sum the values of adjacent vertices.

• Goal: Sum at each vertex should be a common multiple of thevalue at the vertex.

?1

?

1?

1

?

1

?

1

Page 8: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Good labellings

• Assign a real number to each vertex.

• For each vertex, sum the values of adjacent vertices.

• Goal: Sum at each vertex should be a common multiple of thevalue at the vertex.

?1

?

1?

1

?

1

?

1

Page 9: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Good labellings

• Assign a real number to each vertex.

• For each vertex, sum the values of adjacent vertices.

• Goal: Sum at each vertex should be a common multiple of thevalue at the vertex.

−11

−1

1−1

1

−1

1

−1

1

Page 10: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

1

1

?

?

??

?

1

1 ?

Page 11: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

1

1

-23

-23

-23-23

-23

1

1 -23

Page 12: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Linear algebra

Adjacency relation

Let V be the vertex-set of a graph.u ∼ v

Adjacency operator A on RV

Given f : V → R, we define

Af : V → R

: v 7→∑u∼v

f (u)

Page 13: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Eigenvectors of A

11

1

1

11

1

1

1

1

−11

−1

1

−11

−1

1

−1

1

1

1

- 23

- 23

- 23

- 23

- 23

1

1 - 23

3 1 −2

Page 14: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Algebraic graph theory

• Subject grew in the 1950’s and ‘60’s:

• Graph is regular if 1 is an eigenvector.

• #Edges = 12

∑λ2i

• #Triangles = 16

∑λ3i

• 3 distinct eigenvalues −→ strongly regular

• Smallest eigenvalue −→ independence and chromatic numbers

• Second largest eigenvalue −→ expansion and randomness properties

• Interlacing −→ substructures.

Page 15: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Algebraic graph theory

• Subject grew in the 1950’s and ‘60’s:

• Graph is regular if 1 is an eigenvector.

• #Edges = 12

∑λ2i

• #Triangles = 16

∑λ3i

• 3 distinct eigenvalues −→ strongly regular

• Smallest eigenvalue −→ independence and chromatic numbers

• Second largest eigenvalue −→ expansion and randomness properties

• Interlacing −→ substructures.

Page 16: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Algebraic graph theory

• Subject grew in the 1950’s and ‘60’s:

• Graph is regular if 1 is an eigenvector.

• #Edges = 12

∑λ2i

• #Triangles = 16

∑λ3i

• 3 distinct eigenvalues −→ strongly regular

• Smallest eigenvalue −→ independence and chromatic numbers

• Second largest eigenvalue −→ expansion and randomness properties

• Interlacing −→ substructures.

Page 17: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Algebraic graph theory

• Subject grew in the 1950’s and ‘60’s:

• Graph is regular if 1 is an eigenvector.

• #Edges = 12

∑λ2i

• #Triangles = 16

∑λ3i

• 3 distinct eigenvalues −→ strongly regular

• Smallest eigenvalue −→ independence and chromatic numbers

• Second largest eigenvalue −→ expansion and randomness properties

• Interlacing −→ substructures.

Page 18: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Algebraic graph theory

• Subject grew in the 1950’s and ‘60’s:

• Graph is regular if 1 is an eigenvector.

• #Edges = 12

∑λ2i

• #Triangles = 16

∑λ3i

• 3 distinct eigenvalues −→ strongly regular

• Smallest eigenvalue −→ independence and chromatic numbers

• Second largest eigenvalue −→ expansion and randomness properties

• Interlacing −→ substructures.

Page 19: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Using the spectral decomposition

Spectral Theorem

The eigenspaces of A form an orthogonal decomposition of RV .

1

1

- 23

- 23

- 23- 2

3

- 23

1

1 - 23

,

−11

−1

1−1

1

−1

1

−1

1

= 0

Page 20: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing sets

Corollary

Intriguing sets X and Y associated to different eigenvalues satisfy

|X ∩ Y | =|X ||Y ||V |

.

1

1

0

0

00

0

1

1 0

,

10

1

01

0

1

0

1

0

= 2

Page 21: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing maps

f : V → R is intriguing ⇐⇒ for some α, β ∈ R

Af = α · f + β · 1.

(In fact, α is an eigenvalue of A.)

1

0

1

0

1

0

1

0

1

0

2

1

2

1

2

1

2

1

2

1

f

Af = f + 1

Page 22: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing maps

f : V → R is intriguing ⇐⇒ for some α, β ∈ R

Af = α · f + β · 1.

(In fact, α is an eigenvalue of A.)

1

0

1

0

1

0

1

0

1

0

2

1

2

1

2

1

2

1

2

1

f

Af = f + 1

Page 23: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing maps

f : V → R is intriguing ⇐⇒ for some α, β ∈ R

Af = α · f + β · 1.

(In fact, α is an eigenvalue of A.)

1

0

1

0

1

0

1

0

1

0

2

1

2

1

2

1

2

1

2

1

f

Af = f + 1

Page 24: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing maps

f : V → R is intriguing ⇐⇒ for some α, β ∈ R

Af = α · f + β · 1.

(In fact, α is an eigenvalue of A.)

1

0

1

0

1

0

1

0

1

0

2

1

2

1

2

1

2

1

2

1

f Af = f + 1

Page 25: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing maps

0

−1

1

0

1

0

1

0

0 0

1

3

−1

1

−1

1

−1

1

1 1

f

Af = −2 · f + 1

Page 26: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing maps

0

−1

1

0

1

0

1

0

0 0

1

3

−1

1

−1

1

−1

1

1 1

f

Af = −2 · f + 1

Page 27: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Intriguing maps

0

−1

1

0

1

0

1

0

0 0

1

3

−1

1

−1

1

−1

1

1 1

f Af = −2 · f + 1

Page 28: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Af = αf f + βf 1, Ag = αg f + βg1, A1 = k1.

Corollary

Intriguing maps f and g associated to different eigenvalues satisfy

〈f , g〉 =〈f ,1〉〈g ,1〉|V |

.

Proof.Eigenvectors Eigenvalue

(k − αf )f − βf 1 αf

(k − αg )g − βg1 αg

1 k

(k − αf )(k − αg )〈f , g〉−βg (k − αf )〈f ,1〉−βf (k − αg )〈1, g〉+βf βg |V | = 0

Page 29: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Af = αf f + βf 1, Ag = αg f + βg1, A1 = k1.

Corollary

Intriguing maps f and g associated to different eigenvalues satisfy

〈f , g〉 =〈f ,1〉〈g ,1〉|V |

.

Proof.Eigenvectors Eigenvalue

(k − αf )f − βf 1 αf

(k − αg )g − βg1 αg

1 k

(k − αf )(k − αg )〈f , g〉−βg (k − αf )〈f ,1〉−βf (k − αg )〈1, g〉+βf βg |V | = 0

Page 30: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Af = αf f + βf 1, Ag = αg f + βg1, A1 = k1.

Corollary

Intriguing maps f and g associated to different eigenvalues satisfy

〈f , g〉 =〈f ,1〉〈g ,1〉|V |

.

Proof.Eigenvectors Eigenvalue

(k − αf )f − βf 1 αf

(k − αg )g − βg1 αg

1 k

(k − αf )(k − αg )〈f , g〉−βg (k − αf )〈f ,1〉−βf (k − αg )〈1, g〉+βf βg |V | = 0

Page 31: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Euler’s Theorem on latin squares

Cyclic latin squares1 2 3 4 5

2 4 1 5 3

3 5 4 2 1

4 1 5 3 2

5 3 2 1 4

Transversal1 2 3 4 5

2 4 1 5 3

3 5 4 2 1

4 1 5 3 2

5 3 2 1 4

Page 32: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Euler’s Theorem on latin squares

Cyclic latin squares1 2 3 4 5

2 4 1 5 3

3 5 4 2 1

4 1 5 3 2

5 3 2 1 4

Transversal1 2 3 4 5

2 4 1 5 3

3 5 4 2 1

4 1 5 3 2

5 3 2 1 4

Page 33: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Theorem (Euler, 1782)

An n × n cyclic latin square does not have a transversal whenn is even.

The strongly regular graph

Vertices: Cells of the latin squareAdjacency: Same row, same column, or same symbolEigenvalues: 3(n − 1), n − 3, −3.

1 2 3 4 52 4 1 5 33 5 4 2 14 1 5 3 25 3 2 1 4

Page 34: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Theorem (Euler, 1782)

An n × n cyclic latin square does not have a transversal whenn is even.

The strongly regular graph

Vertices: Cells of the latin squareAdjacency: Same row, same column, or same symbolEigenvalues: 3(n − 1), n − 3, −3.

1 2 3 4 52 4 1 5 33 5 4 2 14 1 5 3 25 3 2 1 4

Page 35: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Transversals are intriguing

11 20 30 40 50

20 40 10 51 30

30 50 41 20 10

40 10 50 30 21

50 31 20 10 40

10 23 33 43 53

23 43 13 50 33

33 53 40 23 13

43 13 53 33 20

53 30 23 13 43

1T

A1T = −31T + 31

Page 36: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Transversals are intriguing

11 20 30 40 50

20 40 10 51 30

30 50 41 20 10

40 10 50 30 21

50 31 20 10 40

10 23 33 43 53

23 43 13 50 33

33 53 40 23 13

43 13 53 33 20

53 30 23 13 43

1T

A1T = −31T + 31

Page 37: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Transversals are intriguing

11 20 30 40 50

20 40 10 51 30

30 50 41 20 10

40 10 50 30 21

50 31 20 10 40

10 23 33 43 53

23 43 13 50 33

33 53 40 23 13

43 13 53 33 20

53 30 23 13 43

1T A1T = −31T + 31

Page 38: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

A magical intriguing map µ

11 21 31 41 51

21 40 14 51 34

31 50 42 25 17

41 15 52 35 27

51 34 26 18 46

132 232 332 432 532

232 430 138 532 338

332 530 434 240 144

432 140 534 340 244

532 338 242 146 442

µ

Aµ = 2 · µ+ 30 · 1

Page 39: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

A magical intriguing map µ

11 21 31 41 51

21 40 14 51 34

31 50 42 25 17

41 15 52 35 27

51 34 26 18 46

132 232 332 432 532

232 430 138 532 338

332 530 434 240 144

432 140 534 340 244

532 338 242 146 442

µ

Aµ = 2 · µ+ 30 · 1

Page 40: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

A magical intriguing map µ

11 21 31 41 51

21 40 14 51 34

31 50 42 25 17

41 15 52 35 27

51 34 26 18 46

132 232 332 432 532

232 430 138 532 338

332 530 434 240 144

432 140 534 340 244

532 338 242 146 442

µ Aµ = 2 · µ+ 30 · 1

Page 41: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Two opposing intriguing maps

1T : 〈1T ,1〉 = nµ: 〈µ,1〉 = n2(n + 1)/2

So

〈1T , µ〉 =〈1T ,1〉〈µ,1〉

n2=

n · n2(n + 1)/2

n2=

n(n + 1)

2

But if n is even, then

n(n + 1)

2≡ n

2(mod n).

A contradiction for a CYCLIC latin square!

Page 42: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Two opposing intriguing maps

1T : 〈1T ,1〉 = nµ: 〈µ,1〉 = n2(n + 1)/2

So

〈1T , µ〉 =〈1T ,1〉〈µ,1〉

n2=

n · n2(n + 1)/2

n2=

n(n + 1)

2

But if n is even, then

n(n + 1)

2≡ n

2(mod n).

A contradiction for a CYCLIC latin square!

Page 43: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Two opposing intriguing maps

1T : 〈1T ,1〉 = nµ: 〈µ,1〉 = n2(n + 1)/2

So

〈1T , µ〉 =〈1T ,1〉〈µ,1〉

n2=

n · n2(n + 1)/2

n2=

n(n + 1)

2

But if n is even, then

n(n + 1)

2≡ n

2(mod n).

A contradiction for a CYCLIC latin square!

Page 44: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Finitely many points and lines

Generalised quadrangle

Given a point P and ` which are not incident, there is a unique linem on P concurrent with `.

`

P

Order (s, t)

s + 1 points on a line, t + 1 lines through a point

Page 45: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Finitely many points and lines

Generalised quadrangle

Given a point P and ` which are not incident, there is a unique linem on P concurrent with `.

`

P

Order (s, t)

s + 1 points on a line, t + 1 lines through a point

Page 46: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

m-coversA set of lines M of a generalised quadrangle is an m-cover ifevery point lies on m elements of M.

Figure: A 2-cover of W(3, 2).

Page 47: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

• Known m-covers:W(3, q) Not too manyQ(4, q), q odd Many, m evenQ−(5, q) Many!H(4, q2) Many found (recent), m > 1H(3, q2) Hemisystems, q odd

• Segre (1965):An m-cover of H(3, q2), q odd, has m = q+1

2 (a hemisystem).

• Bruen & Hirschfeld (1978):No m-cover exists of H(3, q2), q even.

Page 48: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

• Known m-covers:W(3, q) Not too manyQ(4, q), q odd Many, m evenQ−(5, q) Many!H(4, q2) Many found (recent), m > 1H(3, q2) Hemisystems, q odd

• Segre (1965):An m-cover of H(3, q2), q odd, has m = q+1

2 (a hemisystem).

• Bruen & Hirschfeld (1978):No m-cover exists of H(3, q2), q even.

Page 49: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

• Known m-covers:W(3, q) Not too manyQ(4, q), q odd Many, m evenQ−(5, q) Many!H(4, q2) Many found (recent), m > 1H(3, q2) Hemisystems, q odd

• Segre (1965):An m-cover of H(3, q2), q odd, has m = q+1

2 (a hemisystem).

• Bruen & Hirschfeld (1978):No m-cover exists of H(3, q2), q even.

Page 50: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

• J. A. Thas (1981):

Hemisystem of H(3, q2) −→ partial quadrangle andstrongly regular graph

• J. A. Thas (1989):An m-cover of a GQ of order (q2, q), q odd, has m = q+1

2 .

• m-covers are intriguing.

Af = (−s + 1) · f + m(s + 1) · 1

Page 51: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

• J. A. Thas (1981):

Hemisystem of H(3, q2) −→ partial quadrangle andstrongly regular graph

• J. A. Thas (1989):An m-cover of a GQ of order (q2, q), q odd, has m = q+1

2 .

• m-covers are intriguing.

Af = (−s + 1) · f + m(s + 1) · 1

Page 52: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

• J. A. Thas (1981):

Hemisystem of H(3, q2) −→ partial quadrangle andstrongly regular graph

• J. A. Thas (1989):An m-cover of a GQ of order (q2, q), q odd, has m = q+1

2 .

• m-covers are intriguing.

Af = (−s + 1) · f + m(s + 1) · 1

Page 53: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

A magical intriguing map (Bamberg, Devillers & Schillewaert)

Suppose we have two disjoint lines ` and m. Then

µ := s · 1{`,m}⊥⊥ + t · 1{`,m}⊥

is intriguing.

`

m

Page 54: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Two opposing intriguing maps 1

In the case that |{`,m}⊥⊥| = t2/s + 1 ...

1m-cover 〈1m-cover,1〉 = m(st + 1)µ 〈µ,1〉 = (s + t)(t + 1)

So

〈1m-cover, µ〉 =〈1m-cover,1〉〈µ,1〉

(t + 1)(st + 1)= m(s + t)

1µ = s · 1{`,m}⊥⊥ + t · 1{`,m}⊥

Page 55: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Two opposing intriguing maps 1

In the case that |{`,m}⊥⊥| = t2/s + 1 ...

1m-cover 〈1m-cover,1〉 = m(st + 1)µ 〈µ,1〉 = (s + t)(t + 1)

So

〈1m-cover, µ〉 =〈1m-cover,1〉〈µ,1〉

(t + 1)(st + 1)= m(s + t)

1µ = s · 1{`,m}⊥⊥ + t · 1{`,m}⊥

Page 56: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Double Counting: Fix ` not in m-cover M. Count concurrent pairs(r , s) ∈M, such that r , s are concurrent with `.

Result: m = t+12 .

Generalisation: by Frederic Vanhove to regular near polygons.

Page 57: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Double Counting: Fix ` not in m-cover M. Count concurrent pairs(r , s) ∈M, such that r , s are concurrent with `.

Result: m = t+12 .

Generalisation: by Frederic Vanhove to regular near polygons.

Page 58: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

Double Counting: Fix ` not in m-cover M. Count concurrent pairs(r , s) ∈M, such that r , s are concurrent with `.

Result: m = t+12 .

Generalisation: by Frederic Vanhove to regular near polygons.

Page 59: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

The bigger picture

Instead of regular graphs:Association schemes and the Bose-Mesner algebra.

More than inner products:Inner/outer distribution, Krein parameters,MacWilliams Transform.

More than latin squares and generalised quadrangles:Projective spaces, polar spaces, partial geometries.

Page 60: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

The bigger picture

Instead of regular graphs:Association schemes and the Bose-Mesner algebra.

More than inner products:Inner/outer distribution, Krein parameters,MacWilliams Transform.

More than latin squares and generalised quadrangles:Projective spaces, polar spaces, partial geometries.

Page 61: Algebraic combinatorics applied to finite geometry - SymOmega

Graphs vs Linear algebra Euler’s Theorem on latin squares Finite geometry The bigger picture

The bigger picture

Instead of regular graphs:Association schemes and the Bose-Mesner algebra.

More than inner products:Inner/outer distribution, Krein parameters,MacWilliams Transform.

More than latin squares and generalised quadrangles:Projective spaces, polar spaces, partial geometries.