Skew Hadamard Difference Sets Alexander Pott (with Cunsheng Ding and - - PowerPoint PPT Presentation

skew hadamard difference sets
SMART_READER_LITE
LIVE PREVIEW

Skew Hadamard Difference Sets Alexander Pott (with Cunsheng Ding and - - PowerPoint PPT Presentation

Skew Hadamard Difference Sets Alexander Pott (with Cunsheng Ding and Qi Wang) Otto-von-Guericke-University Magdeburg December 06, 2013 1 / 22 Difference set Subset D of a group G such that every g G , g = 0, has the same number of


slide-1
SLIDE 1

Skew Hadamard Difference Sets

Alexander Pott (with Cunsheng Ding and Qi Wang)

Otto-von-Guericke-University Magdeburg

December 06, 2013

1 / 22

slide-2
SLIDE 2

Difference set

Subset D of a group G such that every g ∈ G, g = 0, has the same number of difference representations d − d′ with d, d′ ∈ D.

Example

{1, 2, 4} ⊆ Z7.

2 / 22

slide-3
SLIDE 3

Construction of difference sets

◮ Use trivial additive sub-structures, interprete

multiplicatively.

◮ Use trivial multiplicative sub-structures, interprete

additively.

3 / 22

slide-4
SLIDE 4

Construction of difference sets

◮ Use trivial additive sub-structures, interprete

multiplicatively.

◮ Use trivial multiplicative sub-structures, interprete

additively.

Example

◮ trace(x) = 0 in F∗ 2n ◮ squares in Fq

3 / 22

slide-5
SLIDE 5

How can we generalize trace(x) = 0?

◮ GORDON-MILLS-WELCH (1962): Modify trace

4 / 22

slide-6
SLIDE 6

How can we generalize trace(x) = 0?

◮ GORDON-MILLS-WELCH (1962): Modify trace

Breakthrough: MASCHIETTI (1998) {x ∈ F∗

2n : trace(x) = 0} = {y2 + y : y ∈ F∗ 2n, y = 1}

Difference set is the image set of y2 + y in F∗

2n.

4 / 22

slide-7
SLIDE 7

How can we generalize trace(x) = 0?

◮ GORDON-MILLS-WELCH (1962): Modify trace

Breakthrough: MASCHIETTI (1998) {x ∈ F∗

2n : trace(x) = 0} = {y2 + y : y ∈ F∗ 2n, y = 1}

Difference set is the image set of y2 + y in F∗

2n.

Generalize this description: Use 2-to-1 mappings.

4 / 22

slide-8
SLIDE 8

Hyperovals

Maschietti used monomial hyperovals: {   1 x xd   : x ∈ F2n} ∪ {   1   ,   1  } is a hyperoval in PG(2, 2n) if and only if yd + y is 2-to-1.

5 / 22

slide-9
SLIDE 9

SIDELNIKOV

{x2 − 1 : x ∈ Fq} ⊆ F∗

q

“almost” difference set in F∗

q, yields sequences with optimal

autocorrelation properties.

6 / 22

slide-10
SLIDE 10

Generalizing Squares I

Cyclotomy: Unions of cosets of multiplicative subgroup. TAO FENG, KOJI MOMIHARA, QING XIANG use small subgroups.

7 / 22

slide-11
SLIDE 11

Generalizing Squares II

Squares are image set of a 2-to-1 mapping f : Fq → Fq! But in the additive group.

8 / 22

slide-12
SLIDE 12

Generalizing Squares II

Squares are image set of a 2-to-1 mapping f : Fq → Fq! But in the additive group. Consider the graph Gf = {(x, f(x)) : x ∈ Fq} If Gf has “nice” properties with respect to addition, then perhaps also the image set.

8 / 22

slide-13
SLIDE 13

Planar functions

f : Fq → Fq is planar if f(x + a) − f(x) is a permutation for all a = 0.

Example

f(x) = x2: (x + a)2 − x2 = 2xa + a2 is a permutation on Fq if q odd. Hence: Squares are image sets of a class of planar functions!

9 / 22

slide-14
SLIDE 14

Squares in Fq are nice

The set of squares are a difference set: d − d′ = x has q−3

4

solutions with d, d′ ∈ D for all x,

10 / 22

slide-15
SLIDE 15

Squares in Fq are nice

The set of squares are a difference set: d − d′ = x has q−3

4

solutions with d, d′ ∈ D for all x, and D ∪ (−D) ∪ {0} = Fq (∗)

10 / 22

slide-16
SLIDE 16

Squares in Fq are nice

The set of squares are a difference set: d − d′ = x has q−3

4

solutions with d, d′ ∈ D for all x, and D ∪ (−D) ∪ {0} = Fq (∗)

Example (q = 7)

{1, 2, 4} ∪ {3, 5, 6} ∪ {0} = F7 skew Hadamard difference sets Hadamard difference set: without (∗).

10 / 22

slide-17
SLIDE 17

Are there others?

Brilliant idea due to DING and YUAN (2006): Try other planar functions!

11 / 22

slide-18
SLIDE 18

Are there others?

Brilliant idea due to DING and YUAN (2006): Try other planar functions! Exactly one gives new example: f(x) = x10 + x6 − x2 in F3n COULTER, MATTHEWS (1998).

11 / 22

slide-19
SLIDE 19

Are there others?

Brilliant idea due to DING and YUAN (2006): Try other planar functions! Exactly one gives new example: f(x) = x10 + x6 − x2 in F3n COULTER, MATTHEWS (1998). ... still no theoretical proof that it is “new” in general

11 / 22

slide-20
SLIDE 20

... rekindled interest in planar functions...

DING and YUAN also proved: f(x) = x10 − x6 − x2 is planar and also gives skew Hadamard difference set.

12 / 22

slide-21
SLIDE 21

Another look at Ding-Yuan

composition of a permutation polynomial and x2: (x5 ± x3 − x) ◦ x2 DICKSON of order 5.

13 / 22

slide-22
SLIDE 22

DING, WANG, XIANG (2007)

q = 32h+1, α = 3h+1, u ∈ Fq Use permutation polynomial f(x) = x2α+3 + (ux)α − u2x (which is not planar):

14 / 22

slide-23
SLIDE 23

DING, WANG, XIANG (2007)

q = 32h+1, α = 3h+1, u ∈ Fq Use permutation polynomial f(x) = x2α+3 + (ux)α − u2x (which is not planar): Image set of f ◦ x2 is skew Hadamard. Inequivalence only in small cases proved.

14 / 22

slide-24
SLIDE 24

DING, P., WANG (2013)

q = 3m, m ≡ 0 mod 3, u ∈ Fq Use DICKSON of order 7: f(x) = x7 − ux5 − u2x3 − u3x. (which is not planar). Inequivalence only in small cases proved.

15 / 22

slide-25
SLIDE 25

Proof I

Proof resembles Ding, Wang, Xiang. Have to show |Ψ(D)|2 = 3m+1

4

for additive characters Ψ. Thanks to CHEN, SEHGAL, XIANG (1994), it is sufficient to show: Ψ(D) ≡ 3(m−1)/2 − 1 2 mod 3(m−1)/2.

16 / 22

slide-26
SLIDE 26

Proof II

Show Sβ =

  • z∈F∗

q

Ψβ(f(z))χ(z) ≡ 0 mod 3(m−1)/2 where χ is the quadratic character and Ψβ(z) = ζTrace(βz)

3

. This reduces to

  • z∈F∗

q

ζTrace(z7+ηz5+γz)

3

χ(z) for some η and γ.

17 / 22

slide-27
SLIDE 27

Proof III

  • z∈F∗

q

ζTrace(z7+ηz5+γz)

3

χ(z) Use ζTrace(z)

3

= 1 q − 1

q−2

  • b=0

g(ω−b)ωb(z) where g(ω−b) is Gauss sum with respect to multiplicative character ω−b, where ω has order q − 1.

18 / 22

slide-28
SLIDE 28

Proof IV

If γ = 0, we obtain Sβ = ± 1 q − 1

q−2

  • b=0

g(ω−b)g(ω− q−1

2 +5−17b) × root of unity

Then use STICKELBERGER and combinatorial arguments. Case γ = 0 is similar.

19 / 22

slide-29
SLIDE 29

... use polynomials ...

◮ to construct more Hadamard difference sets; ◮ to construct Sidelnikov sequences x2 − 1; ◮ to construct more skew Hadamard difference sets.

Problem: Show inequivalence!

20 / 22

slide-30
SLIDE 30

MUZYCHUK (2010)

Mikhail Muzychuk has another construction in Fq3 using orbits

  • f vectors in F 3

q under the action of GL(3, q).

21 / 22

slide-31
SLIDE 31

MUZYCHUK (2010)

Mikhail Muzychuk has another construction in Fq3 using orbits

  • f vectors in F 3

q under the action of GL(3, q).

He can show inequivalence.

21 / 22

slide-32
SLIDE 32

MUZYCHUK (2010)

Mikhail Muzychuk has another construction in Fq3 using orbits

  • f vectors in F 3

q under the action of GL(3, q).

He can show inequivalence. Inequivalence of some cyclotomic examples and squares has been shown by KOJI MOMIHARA.

21 / 22

slide-33
SLIDE 33

Inequivalence

Difference set corresponds to a design!

◮ triple intersection numbers; ◮ rank of incidence matrix; ◮ automorphism groups.

22 / 22

slide-34
SLIDE 34

Inequivalence

Difference set corresponds to a design!

◮ triple intersection numbers; MOMIHARA, computer ◮ rank of incidence matrix; ◮ automorphism groups.

22 / 22

slide-35
SLIDE 35

Inequivalence

Difference set corresponds to a design!

◮ triple intersection numbers; MOMIHARA, computer ◮ rank of incidence matrix; always the same for skew H.d.s ◮ automorphism groups.

22 / 22

slide-36
SLIDE 36

Inequivalence

Difference set corresponds to a design!

◮ triple intersection numbers; MOMIHARA, computer ◮ rank of incidence matrix; always the same for skew H.d.s ◮ automorphism groups. MUZYCHUK

22 / 22