Delsarte designs . . . from finite polar spaces John Bamberg The - - PowerPoint PPT Presentation

delsarte designs from finite polar spaces
SMART_READER_LITE
LIVE PREVIEW

Delsarte designs . . . from finite polar spaces John Bamberg The - - PowerPoint PPT Presentation

Delsarte designs . . . from finite polar spaces John Bamberg The University of Western Australia quote This motivates the present general definition, the conjecture being that T-designs will often have interesting properties..


slide-1
SLIDE 1

Delsarte designs . . . from finite polar spaces

John Bamberg The University of Western Australia

slide-2
SLIDE 2

quote

“This motivates the present general definition, the “conjecture” being that T-designs will often have interesting properties.”. – P. Delsarte’s thesis, p. 33

quote

“If it isn’t in Delsarte’s thesis, it’s in Haemers’ thesis”. – A. B.

slide-3
SLIDE 3

t − (v, k, λ) design

k-subsets B (blocks) of a v-set such that every t-subset is contained in exactly λ blocks. inclusion matrices: k versus i Ii(b, y) =

  • 1

y ⊆ b

  • therwise

      

{1} {2} {3} {4} {1,2}

1 1

{1,3}

1 1

{1,4}

1 1

{2,3}

1 1

{2,4}

1 1

{3,4}

1 1       

designs revisited I

χB(b) =

  • 1

b ∈ B

  • therwise

χBIi = |B| v jIi, ∀0 i t

slide-4
SLIDE 4

1 3 5 2 4 6 7 B = {123, 145, 167, 246, 257, 347, 356} All 3-subsets: {123, 124, 125, 126, 127, 134, . . . , 567} χB = (1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0) χBI0 = (7) χBI1 = (3, 3, 3, 3, 3, 3, 3) χBI2 = (1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) χBI3 = χB

slide-5
SLIDE 5

the Johnson scheme J(7, 3)

two 3-subsets can interact in four different ways: equal A0 = I differ in 1 element A1 differ in 2 elements A2 differ in 3 elements A3 adjacency matrices A0 A1 A2 A3 (simultaneous) eigenspaces E0 E1 E2 E3 dimensions 1 6 14 14 inclusion matrices I0 I1 I2 I3 ranks 1 7 21 35 n.b., I0 =

  • j⊤

and I3 = I.

slide-6
SLIDE 6

eigenspaces: {E0, E1, . . . , Ed}. There is an ordering so that RowSpace(Ii) =

i

  • j=0

Ej

projections

The projection Ei to Ei is a minimal idempotent for the Johnson scheme. χBIi = |B| v jIi ∀0 i t χBEi = |B| v jEi ∀0 i t = 0 ∀0 < i t n.b., E0 = 1

v J

slide-7
SLIDE 7

designs revisited II

χBEi = 0, ∀0 < i t

slide-8
SLIDE 8

what if you change the scheme?

Hamming scheme: orthogonal arrays Grassmann scheme: q-designs Sn-scheme: λ-transitive sets of permutations

T-design, T ⊆ {1, . . . , d}

χBEi = 0, ∀i ∈ T

slide-9
SLIDE 9

Remark

Since E 2

i = Ei and each Ei is positive semidefinite, we have

χBEi = 0 ⇐ ⇒ χBEi = 0 ⇐ ⇒ (χBEi)(χBEi)⊤ = 0 ⇐ ⇒ χB EiE ⊤

i χ⊤ B = 0

⇐ ⇒ χB Ei χ⊤

B = 0.

Analogue (coclique): χB Ai χ⊤

B = 0.

slide-10
SLIDE 10

strongly regular graphs

k-regular graph such that

◮ every pair of adjacent vertices has λ common neighbours ◮ every pair of non-adjacent vertices has µ common neighbours

λ µ

A2 = (k − µ)I + (λ − µ)A + µJ three distinct eigenvalues k θ τ three minimal idempotents Ek = 1

nJ

Eθ Eτ

slide-11
SLIDE 11

Take a minimal idempotent E = 1

nJ and suppose χDE = 0.

A = βI + γJ + αE χD A = βχD + γ|D| j = h1 · χD + h2 · (j − χD)

Theorem

χDE = 0 = ⇒ there exist constants h1, h2, |Γ(v) ∩ D| =

  • h1

if v ∈ D h2 if v / ∈ D That is, D is intriguing. h2 h1

d

slide-12
SLIDE 12

Theorem (Haemers 1995)

Let Γ be a connected k-regular graph, and Y ⊆ V Γ. Let N be the number

  • f ordered pairs of adjacent vertices of Y . Then

θmin|Y | + k − θmin n |Y |2 N θmax|Y | + k − θmax n |Y |2 Moreover, if equality holds in one of the inequalities above, then Y is intriguing. generalised by J¨

  • rg Eisfeld to association schemes1

explored further by De Bruyn and Suzuki2

1Theorem 2 of ‘Subsets of association schemes corresponding to eigenvectors of the

Bose-Mesner algebra’. Bull. Belg. Math. Soc. Simon Stevin 5 (1998).

2‘Intriguing sets of vertices of regular graphs’. Graphs & Combinatorics 26 (2010).

slide-13
SLIDE 13

polar spaces

geometry of totally singular subspaces of a sesquilinear or quadratic form introduced by Freudenthal, Tits, Veldkamp (1950’s) collinearity graph is strongly regular

classical

symplectic

  • rthogonal (elliptic, parabolic, hyperbolic)

Hermitian W(d, q) Q−(d, q), Q(d, q), Q+(d, q) H(d, q)

slide-14
SLIDE 14

eigenvalues: degree positive one negative one eigenspaces: j E+ E− interesting cases: – {1}-designs {2}-designs

a {2}-design: points in a generator π (dimension d)

χπA =

  • qd − 1
  • χπ + qd − 1

q − 1 j The χπ span all of j ⊕ E+.

slide-15
SLIDE 15

anti-designs

Theorem (Roos 1982)

Let D be a T-design and let A be a U-design with T ∩ U = ∅. Then |D ∩ A| = |D||A| n .

Proof.

0 = (χDEi)(χAEi)⊤ = χDEiχ⊤

A

= χD(I − 1

nJ)χ⊤ A

= χDχ⊤

A − 1 nχDJχ⊤ A

= |D ∩ A| − 1 n |D||A|

slide-16
SLIDE 16

{1}-designs are m-ovoids

must meet {2}-designs in (product of sizes divided by n) = ⇒ must meet sets of points in a generator in a constant = ⇒ m-ovoid {1}-designs {2}-designs i-tight sets m-ovoids

intriguing sets

χSA = h1 · χS + h2 · (j − χS)

Theorem

An m-ovoid and an i-tight set meet in m · i elements.

slide-17
SLIDE 17

Quote

“The definition of T-designs in a symmetric association scheme (X, R) can be extended so as to admit the possibility of ‘repeated points’.” – P. Delsarte’s thesis, p. 34

Generalisation

Replace χD with x ∈ Rn (or Cn). T-design xEi = 0, ∀i ∈ T Roos’ Theorem T ∩ U = 0 = ⇒ x · y = (x·j)(y·j)

n

slide-18
SLIDE 18

Some applications

Vanhove (2009)

A partial spread of H(2d − 1, q2), d odd, has size at most qd + 1.

De Bruyn & Vanhove (2013)

No generalised hexagon of order (s, s3), s 2, can have 1-ovoids.

B., De Beule, Ihringer (2017)

No ovoids of H(2r − 1, q2) can exist for r > q3 − q2 + 2. Excellent resources Ferdinand Ihringer’s Masters Thesis: Faszinierende Mengen in Polarr¨ aumen Fr´ ed´ eric Vanhove’s PhD Thesis: Incidence geometry from an algebraic graph theory point of view Jan De Beule’s talk (2014): Do i-tight sets and m-ovoids hate each other?

slide-19
SLIDE 19
  • ther interesting designs and antidesigns, and connections

m-systems of polar spaces distance-j-ovoids of generalised hexagons SPG-reguli pseudo-ovoids of PG(4m + 3, q) Cameron-Liebler line classes of PG(3, q) Erd˝

  • s-Ko-Rado sets

partial quadrangles cometric association schemes

slide-20
SLIDE 20

what do we know about m-ovoids?

underlying points of an m-system is an m′-ovoid (Shult + Thas 1994) hemisystems sometimes give rise to two-character sets

slide-21
SLIDE 21

m-ovoids of symplectic spaces

W(3, q), q even Many W(3, q), q odd Many, m even W(5, q), q even m = q + 1, (Cossidente, Pavese) W(5, q), q odd Various W(2r − 1, q), r > 2, q odd m −3+√9+4qr

2q−2

, (B., Kelly, Law, Penttila)

Problem

Improve the lower bound on m, when q is odd and r > 2. Conjecture? m qr−2

slide-22
SLIDE 22

hermitian spaces

H(3, q2) many H(4, q2) none known! H(5, q2) various H(6, q2) none known!

conjecture

For all 1 m q3, there are no m-ovoids of H(4, q2).

known

Thas 1981: no examples for m = 1 B., Kelly, Law, Penttila 2007: m < √q. B., Devillers, Schillewaert 2012: m < −3q−3+√

4q5−4q4+5q2−2q+1 2(q2−q−2)

slide-23
SLIDE 23
  • rthogonal spaces

Q(4, q) interesting Q(6, q) many? Q−(5, q) hemisystems Q−(7, q) from 1-systems; m = q + 1 Q+(5, q) Many Q+(7, q) ?

question

Do there exist 2-ovoids of Q(4, q) for q > 5?

slide-24
SLIDE 24

non-embeddable GQ

DH(4, q2): constructions for many m, lower bound m f (q)? Flock GQ: hemisystems (Cameron, Goethals, Seidel 1979) Payne derived GQ: some examples (B., Devillers, Schillewaert 2012) T3(O), q even: don’t exist T2(O), q even: ?

slide-25
SLIDE 25

what do we know about i-tight sets?

sometimes gives rise to two-character sets i-tight sets of orthogonal (hyperbolic) spaces ≡ Cameron-Liebler line classes of PG(3, q) reducible examples: union of generators Baer examples: union of Baer-subgeometries (e.g., W(2n + 1, q) ֒ → H(2n + 1, q2))

slide-26
SLIDE 26

B., Kelly, Law, Penttila 2007: no irreducible i-tight sets in W(2r − 1, q), H(2r − 1, q2), or Q+(2r − 1, q), for i small compared to q De Beule-Govaerts-Hallez-Storme 2009:

H(2n + 1, q)

i < q5/8/ √ 2 + 1 = ⇒ reducible or Baer. Beukemann & Metsch 2013:

Q+(2n + 1, q)

1 n 3 and i q, or n 4 and i < q and q 71 = ⇒ reducible.

slide-27
SLIDE 27

Metsch 2016: reducibility in the ‘other’ polar spaces De Beule & Metsch 2017: i-tight set of H(4, q) and H(6, q) is reducible if i is small compared to q. Naki´ c and Storme 2017

H(2n + 1, q)

i < (q2/3 − 1)/2 = ⇒ reducible or Baer.

slide-28
SLIDE 28

reducibility

sometimes an irreducible i-tight set is the complement of a reducible i′-tight set. S7 acting on H(3, 32)

  • rbit lengths on points: 70, 210

the set of size 210 is an irreducible 21-tight set but it is the complement of a partial spread of 7 lines.

proposal

study minimal partitions of the points into i-tight sets.

slide-29
SLIDE 29
  • ther things

m-ovoids of regular near polygons (Jesse Lansdown’s talk) relative m-ovoids of polar spaces (Francesco Pavese’s talk) i-tight sets of hyperbolic quadrics (Sasha Gavrilyuk’s talk)

research directions

interplay of tight sets with m-ovoids to prove non-existence results constructions of examples re-contextualisation of old results