30
1 Developments of Algebraic Combinatorics a personal view Eiichi Bannai Shanghai Jiao Tong University Sept 13, 2013 Talk at Egawa 60 conference Tokyo University of Science Tokyo Japan

Developments of Algebraic Combinatorics

  • Upload
    others

  • View
    3

  • Download
    0

Embed Size (px)

Citation preview

1

Developments of Algebraic Combinatorics

a personal view

Eiichi Bannai

Shanghai Jiao Tong University

Sept 13, 2013

Talk at Egawa 60 conference

Tokyo University of ScienceTokyo Japan

2

Finite group theory in 1960’s (−→ 1970’s)

• The attempt to classify finite simple groups had been started.( Feit-Thompson, M. Suzuki, Gorenstein, Aschbacher, K. Harada, . . . . . . . . . )

• Many new finite simple groups were discovered.(Janko, Higman-Sims, Conway, Fisher, . . . . . . . . . )

Finite simple groups

Zp cyclic groups of order pAn alternating groupssimple groups of Lie typesporadic simple groups (26 of them)

• Finite permutation groups were very much sterdied.

• Multiply transitive permutation groups. (Wielandt, M. Hall, . . . )• Rank 3 permutation groups. (Higman-Sims, . . . )• Primitive permutation groups (D. G. Higman, . . . )

(permutation groups of maximum diameter = distance transitive graphs)

3

I was fascinated with finite group theory.

The development of finite group theory was very dynamic !

Japanese group theorists at that time

in USA in Japan

M. Suzuki (Univ. Illinois) Tokyo (Iwahori,. . . )

N. Ito (Chicago Circle) Osaka (Nagao,. . . )

K. Harada (Ohio State) Hokkaido (Tsuzuku,. . . )

4Ohio State University

Harada 1968 1972 -

Bannai -1974 1976 1978 -1989

Egawa -1977 1981

Y. Egawa: PhD in 1980 (adviser K. Harada)Standard component problem (in pure group theory)

In late 1970’s, there was a move from“ group theory ” to “algebraic combinatorics ”

5rank 3 perm. group −→ strongly regular graph

G y Ω transitively (v, k, λ, µ)

α ∈ Ω

v = |Ω| = |V |,k = |Γ1(α)|

Gα has 3 orbits

α, Γ1(α), Γ2(α)

SRG of type (10, 3, 0, 1)v, k, λ, µ

Tsuchiya
楕円
Tsuchiya
四角形
Tsuchiya
楕円
Tsuchiya
四角形
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
多角形
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
タイプライタ
x
Tsuchiya
タイプライタ
y
Tsuchiya
タイプライタ
z
Tsuchiya
タイプライタ
λ
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
z
Tsuchiya
タイプライタ
μ
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
鉛筆
Tsuchiya
タイプライタ
k

6

distance transitive graph

(permutation gp. of max. diameter)

x, y, z, w ∈ V

d(x, y) = d(z, w)

⇒ ∃g ∈ G

s.t. z = xg, w = yg

distance regular graph

0 1

k a1 c2

b1 a2 cd−1

b2 ad−1 cd

bd−1 ad

Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
四角形
Tsuchiya
楕円
Tsuchiya
四角形
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
Γ1(x)
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
Γ2(x)
Tsuchiya
楕円
Tsuchiya
タイプライタ
Γi-1(x)
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
タイプライタ
Γi(x)
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
Tsuchiya
タイプライタ
ci
Tsuchiya
タイプライタ
y
Tsuchiya
鉛筆
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
ai
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
Tsuchiya
タイプライタ
bi
Tsuchiya
タイプライタ
Γi+1(x)
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
Γd(x)
Tsuchiya
矢印

7

transitive perm. group

of subdegrees

1, k, k(k − 1), k(k − 1)2,

· · · , k(k − 1)d

−→ Moore graph

0 1k 0 1

k − 1 0k − 1

10 1

k − 1 k − 1

Hoffman-Singleton

d = 2 =⇒ k = 2, 3, 7 or 57

k = 57 =⇒ still open

d ≥ 3, k ≥ 3 =⇒ Non-existence.

Tsuchiya
テキスト ボックス
-1

8More generally,

transitive perm. proup −→ association scheme

multiplicity free trans.

perm. group −→ com. association scheme

a perm. group −→ coherent configuration

(not necessarily transitive)

9Association schemes

X = (X, Ri0≤i≤d).

Ri ⊂ X × X

1. R0 = (x, x) | x ∈ X,

2. R0, R1, . . . , Rd gives a partition of X × X,

3. for each i ∈ 0, 1, . . . , d, tRi = Ri′

with some i′ ∈ 0, 1, . . . , d,

where tRi = (y, x) | (x, y) ∈ Ri,

4. For i, j, k ∈ 0, 1, . . . , d,

|z ∈ X | (x, z) ∈ Ri, (z, y) ∈ Rj| = pki,j.

A0, A1, . . . , Ad are the adjacency matrices

for the relations R0, R1, . . . , Rd

10We want to classify distance transitive graphs

and distance regular graphs

Question 1.

Which parameters are possible for distance regular graphs ?

Question 2.

Can we characterize distance regular graphs with

given set of parameters ?

Y. Egawa: Characterization of H(n, q) by the parameters,

J.C.T.(A), 1981.(Characterization of H(n, 2) by the parameters was done earlier

by H. Enomoto)

Y. Egawa: Association schemes of quadratic forms,

J.C.T.(A), 1985.

11The set X of all quadratic forms of degree n over Fq

We can get a structure of distance regular graphs by combining

several relations together.

This means that there is an advantage of “considering associa-

tion schemes ” over “considering just the action of the group”.

After this work, Y. Egawa (also H. Enomoto) started to study

graph theory.

Tsuchiya
楕円
Tsuchiya
多角形
Tsuchiya
多角形
Tsuchiya
多角形
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
多角形
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
0
Tsuchiya
タイプライタ
1
Tsuchiya
タイプライタ
2
Tsuchiya
タイプライタ
3
Tsuchiya
タイプライタ
4
Tsuchiya
タイプライタ
n-1
Tsuchiya
タイプライタ
n
Tsuchiya
タイプライタ
|X|=q
Tsuchiya
タイプライタ
n(n+1) 2
Tsuchiya
直線
Tsuchiya
タイプライタ
d=[ ]
Tsuchiya
タイプライタ
d+1 2
Tsuchiya
直線
Tsuchiya
タイプライタ
(q=odd, n=even)

12What is Algebraic Combinatorics?E. Bannai-T. Ito: Algebraic Combinatorics, I

(Benjamin/Cummings, 1984).

Our philosophy: ”group theory without groups”.

Let X = (X, Ri(0≤i≤d)) be a symmetric association scheme.

Let ⟨A0, A1, . . . , Ad⟩ = ⟨E0, E1, . . . , Ed⟩ be the Bose-Mesner

algebra.

X is a P-poly. ⇐⇒ ∃vi(x) = poly. of degree exactly i,

assoc. scheme 0 ≤ i ≤ d , such that Ai = vi(A1)

⇐⇒ (X, R1) is a distance regular graph

X is a Q-poly. ⇐⇒ ∃v∗i (x) = poly. of degree exactly i,

assoc. scheme 0 ≤ i ≤ d , such that |X|Ei = v∗i (|X|E1)

13Our main theme:Classify P- and Q-polynomial association schemes.• X = P- and Q-poly. assoc. schme

=⇒ both vi(x) (0 ≤ i ≤ d) and v∗i (x) (0 ≤ i ≤ d)

are orthgonal polynomials (they are mutually related).

=⇒ They are expressed by Ashkey-Wilson polynomials,

or by their special cases or limiting cases. (Leonard 1982)q = ±1 q −→ 1 q −→ −1

Case I Case II Case III

Recently, this class of poly-nomials have very much stud-ied in the theory og orthogonalpolynomials. (Tsujimoto-Vinet-Zhedanov [20] 2012)

Tsuchiya
テキスト ボックス
[20]

14

T. Ito-P. Terwilliger determined irreducible representations of

Terwilliger algebra (for case I).

Now we should come back to the study of P- and Q- polynomial

association schemes using their results. (For example, charac-

terizations by parameters.)

15Next steps

multiplicity free perm. groups(= Gelfand pairs )

−→ commutative assoc. schemes(Cf. Martin-Tanaka [19] 2009)

any finite groups −→ group association schemes(commutative assoc. schemes)

pairs of a finite group and a max-imal subgroup

−→ primitive association schemes

finite simple groups −→ primitive group assoc. schemes(commutative)

We would like to study finite simple groups from the view pointof association schemes and algebraic combinatorics.At present, this is not a realistic problem. But we should try,and should not give up.

Tsuchiya
テキスト ボックス
[19]

Another source of Algebraic Combinatorics 16(Delsarte Theory)

What are codes and designs ?

M = a set (finite or infinite metric space)

e.g.(V

k

)(= set of a k elements subset of V ),

Sn−1 = (x1, . . . , xn) | x21 + · · · + x2

n = 1

Johnson association scheme J(v, k) is defined on(V

k

).

Hamming association scheme H(n, q) is defined as follows.X = F n, where F is a finite set of q elements.For x = (x1, . . . , xn), y = (y1, . . . , yn) ∈ X we define(x, y) ∈ Ri ⇐⇒ |ν | xν = yν| = i, for i = 0, 1, . . . , n

Tsuchiya
タイプライタ
In
Tsuchiya
タイプライタ
X=( ),
Tsuchiya
タイプライタ
V k
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
タイプライタ
x
Tsuchiya
タイプライタ
y
Tsuchiya
タイプライタ
k
Tsuchiya
タイプライタ
k
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
k-i
Tsuchiya
鉛筆
Tsuchiya
タイプライタ
Tsuchiya
タイプライタ
(x,y)∈Ri(i=0,1, ,d)
Tsuchiya
直線

17Purpose of coding theory

Find subset Y ⊂ M

such that

mind(x, y); x, y ∈ Y, x = yis maximum

M = H(n, q) =⇒usual coding theory.

M = Sn−1 =⇒optimal spherical code

Purpose of design theory

Find subset Y ⊂ M which approximates the whole

space M well.

combinatorial t-design

(t-(v, k, λ)- design, 1 ≤ t ≤ k ≤ v)

X =(V

k

)Y ⊂ X is a t-(v, k, λ)- design

⇐⇒ |y ∈ Y | T ⊂ y| = λ (constant) for ∀T ⊂(V

t

).

18M = Sn−1

spherical t-design

Y ⊂ Sn−1

1|Sn−1|

∫Sn−1 f(x)dσ(x) = 1

|Y |∑

y∈Y f(y)

∀polynomial f(x) = f(x1, . . . , xn) of degree ≤ t.

regular polyhedron # of vertices t-design

tetrahedron 4 2

cube 8 3

octahedron 6 3

icosahedron 12 5

dodecahedron 20 5240 roots of type E8(⊂ R8) is a 7-design

196560 min. vectors of Leech lattice (⊂ R24) is an 11-design

19

Delsarte (1973) unified the study of codes and designsin the frame work of assoc. schemes.

An algebraic approach to the theory of association schemes of

the coding theory [10] (1973)

P-poly. assoc. scheme −→ error correcting e-codes.

(H(n, q) −→ usual coding theory)

Q-poly. assoc. scheme −→ t-designs.

(J(v, k) =(V

k

)−→ usual design theory)

Delsare-Goethals-Seidel [12] (1977) studied codes and designs

on the sphere Sn−1, in a similar way as codes and designs were

studied in association schemes.

Tsuchiya
テキスト ボックス
[10]
Tsuchiya
テキスト ボックス
[12]

A natural way to find t -(v, k, λ) designs is to take a finite 20subgroup G ⊂ Sv, and take the orbit Y = yG of G on y ∈

(Vk

).

Also, a natural way to find spherical t-designs is to take a finite

subgroup G ⊂ O(n) and take an orbit Y = yG of G on y ∈ Sn−1.

However, it seems that if such Y is a t-design (i.e., t-(v, k, λ)

design or spherical t-design on Sn−1 with n ≥ 3), then t must

be bounded.

So, we cannot construct t-(v, k, λ) designs or spherical t-designs

for large t, from groups.

However, t-(v, k, λ) design exist for any t (Teirlinck, 1987) (forsome v, k and λ), and spherical t-designs in Sn−1 exist for any tand n. (Seymour-Zaslavsky, 1984). See also, recent arXiv papers:(1) Kuperberg-Lovett-Peled, arXiv:1302.4295,(2) Fazeli-Lovett-Vardy, arXiv:1306.2088,(3) Bondarenko-Radchenko-Viazovska, Ann of Math.(2013), and

arXiv:1303.5991.

21

Note that Delsarte’s definition of t-design for any Q-polynomial

association scheme X = (X, Ri(0≤i≤d)) is given as follows.

Let Y be a subset of X, and let φY be the characteristic vector

(function) of Y . Then, Y is a t-design, if and only if

EiφY = 0, for i = 1, 2, . . . , t.

Then we have the following Fisher type inequalities. (For sim-

plicity, we assume t = 2e.)

22• If t = 2e and Y be a 2e-design in

(Vk

)(i.e., in Johnson assoc.

scheme J(v, k), or more generally for any Q-polynomial assoc.

scheme X = (X, Ri0≤i≤d)), then

|Y | ≥ me + me−1 + · · · + m1 + m0,

where mi = rank of Ei.

Namely, the RHS is(v

e

)for J(v, k),

and is equal to(v

e

)+

( ve−1

)+ · · · +

(v0

)for H(n, 2),

since mi =(v

i

)−

( vi−1

)for J(v, k),

and mi =(v

i

), for H(n, 2).

• If t = 2e and Y is a spherical 2e-design in Sn−1, then

|Y | ≥ me + me−1 + · · · + m1 + m0 =(n−1+e

e

)+

(n−1+e−1e−1

),

where mi =(n−1+i

i

)−

(n−1+i−2i−2

).

( Delsarte-Goethals-Seidel, 1977).

23So, we are interested in t-designs of possible smallest cardinality

|Y |. (Those which satisfy the equality in Fisher type inequality

are called tight t-designs.The classification of tight 2e-designs in J(v, k).• e = 1 =⇒ symmetric 2-designs. (many examples).• e = 2 =⇒ 4-(23, 7, 1) design or 4-(23, 16, 52) design.(Enomoto-Ito-Noda, 1979.)• e = 3 =⇒ Non-existence (Peterson, 1977)• e ≥ 4 =⇒ for each e there are only finitely many tight 2e-designs, (B, 1977)• 5 ≤ e ≤ 9 =⇒ Non-existence (Dukes and Short-Gershmen, 2013),• e = 4 =⇒ Non-existence (Z. Xiang, 2012 (unpublished))(Open for e ≥ 10.)

The classification of tight spherical t-designs in Sn−1 (we assume n ≥ 3).• If there is a tight t -design =⇒ t ∈ 1, 2, 3, 4, 5, 7, 11 .

(Bannai-Damerell, 1979/80, Bannai-Sloane, 1981)Now, the classification is open only for t = 4, 5, 7.(Cf. Bannai-Munemasa-Venkov (2004, Algebra i Analiz),Nebe-Venkov (2012, Algebra i Analiz).)

24Generalizations of the concept of t-designs

There are many generalizations of t-(v, k, λ) designs and spher-

ical t-designs.

1. t-designs in other (Q-polynomial) association schemes,

and t-designs in other compact symmetric spaces of rank 1

(i.e. projective spaces over R, C, H, O.) That is,

Y ⊂ X, EiφY = 0, for i = 1, 2, . . . , t,

or

Y ⊂ M,1

|M |

∫M

f(x)dσ(x) =1

|Y |∑y∈Y

f(y).

for appropriate f(x).

252. Allow weight function w : Y −→ R>0

Namely, let (Y, w) satisfy:

either ∑T⊂y∈Y

w(y) = λi (constant), for all T ∈ V (i)

for i = 1, 2, . . . , t. (in t-(v, k, λ) design case),

or1

|Sn−1|

∫Sn−1

f(x)dσ(x) =∑y∈Y

w(y)f(y)

for any polynomials f(x) of degree at most t

(in spherical t-design case).

263. Allow different sizes of blocks (in t-(v, k, λ) design case), or

different radii of spheres (in spherical case). Namely, either

Y ⊂ V (k1) ∪ V (k2) ∪ · · · ∪ V (kp)

or

Y ⊂ Sn−1(r1) ∪ Sn−1(r2) ∪ · · · ∪ Sn−1(rp).

Tsuchiya
タイプライタ
|
Tsuchiya
タイプライタ
|
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V k1
Tsuchiya
タイプライタ
|
Tsuchiya
タイプライタ
|
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V k2
Tsuchiya
タイプライタ
|
Tsuchiya
タイプライタ
|
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V kp
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
多角形
Tsuchiya
直線
Tsuchiya
直線
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V 0
Tsuchiya
タイプライタ
,( )
Tsuchiya
タイプライタ
V 1
Tsuchiya
タイプライタ
,( ),
Tsuchiya
タイプライタ
V 2
Tsuchiya
タイプライタ
,( ),
Tsuchiya
タイプライタ
V i
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V k
Tsuchiya
タイプライタ
J(v,i)
Tsuchiya
タイプライタ
H(n,2)(=2 )
Tsuchiya
タイプライタ
V
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
楕円
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V k1
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V k2
Tsuchiya
タイプライタ
( )
Tsuchiya
タイプライタ
V kp

27Remark. The generalization of two steps 2 and 3of t-(v.k.λ) design (=t-design in J(v, k),) is called

”(weighted) regular t-wise balanced design”,

and this concept is equivalent to the concept of

”(weighted) relative t-design” in H(n, 2)

in the sense of Delsarte (1977).

On the other hand, the generalization of two steps 2 and 3of spherical t-designs are called Euclidean t-designs (on p shells).

28Fisher type lower bound are known for relativet-designs in H(n, 2) and Euclidean t-designs.

(Here, we assume t = 2e for simplicity.)

• If t = 2e and (Y, w) is a (weighted) relative 2e-design in H(n, 2)

and if Y is on p-shells (i.e. if Y ⊂ V (k1) ∪ V (k2) ∪ · · · ∪ V (kp)),

then

|Y | ≥ me + me−1 + · · · + me−p+1,

where mi =(v

i

). (Z. Xiang (JCT(A),2012).)

• If t = 2e and (Y, w) is an Euclidean 2e-design in Rn on p-shells,

then

|Y | ≥ me + me−1 + · · · + me−p+1

where mi =(n−1+i

i

)−

(n−1+i−2i−2

). (Muller, 1970’s).

29• The concept of Euclidean t-design was defined by Neumaier-Seidel (1988)and also studied by Delsarte-Seidel (1989). (Similar concepts have been stud-ied in other areas, e.g. numerical analysis, statistics, etc.)We have studied ”tight” Euclidean t-designs extensively for the last 10 years.(There are many interesting examples, and some classification results, thoughthey are still partial.)

• The concept of relative t-design (in Q-polynomial) association schemes) wasintroduced by Delsarte (1977) [13]. The study of tight relative t-designs hasjust started. See for our recent results,Bannai-Bannai-Suda-Tanaka [6] (arXiv: 1303.7163),Bannai-Bannai-Bannai [4] (arXiv:1304.5769).

I believe that now it is the time to start the study of tight relative t-designs

in general (Q-polynomial) association schemes more systematically.

Tsuchiya
テキスト ボックス
[13]
Tsuchiya
テキスト ボックス
[6]
Tsuchiya
テキスト ボックス
[4]

30

Thank You