cohomolgy designs as building blocks in constructions of ... · cohomolgy designs as building...
TRANSCRIPT
Cohomolgy Designs as building blocks in
constructions of Weighing and Hadamard
Matrices
Radel Ben-Av1 Giora Dula4 Assaf Goldberger2 Yossi Strassler3
ArasuFest, Kalamata Aug 1 2019
1Holon Institue of Technology [email protected]
2Tel-Aviv University [email protected]
3 Dan Yishay [email protected]
4 Netanya Accademic College [email protected]
Introduction
Weighing matrices
Definition
A Weighing Matrix W (N,w) is a {0,±1}-matrix W s.t.
WW T = wIN .
Examples (q an odd prime power):
• Hadamard Matrices H(N), (w = N)
• Payley Conference Matrices W (q + 1, q)
• Projective Spaces W(qn+1−1q−1 , qn
)• Smallest unknown: W (35, 25).
1
Monomial Transformations and Automprphisms
Monomial Transformations
Fact: Weighing Matrices are preserved by monomial
transformations:
X = W (n,w) =⇒ MXNT = W (n,w),
where M,N are monomial.
Definition
A monomial pair (M,N) is an Automorphism of X is
MXNT = X .
They form a group Aut(X ).
2
A Motivating Example
This is a W (7, 4):
X =
1 1 1 1 0 0 0
1 −1 0 0 1 1 0
1 0 −1 0 −1 0 1
1 0 0 −1 0 −1 −1
0 1 −1 0 0 1 −1
0 1 0 −1 1 0 1
0 0 1 −1 −1 1 0
• Every (permutation) automorphism of |X | (entrywise abs
value) can be lifted to an automorphism of X .
• How general is this phenomena?
3
A Motivating Example
This is a W (7, 4):
X =
1 1 1 1 0 0 0
1 −1 0 0 1 1 0
1 0 −1 0 −1 0 1
1 0 0 −1 0 −1 −1
0 1 −1 0 0 1 −1
0 1 0 −1 1 0 1
0 0 1 −1 −1 1 0
• Every (permutation) automorphism of |X | (entrywise abs
value) can be lifted to an automorphism of X .
• How general is this phenomena?
3
A Motivating Example
This is a W (7, 4):
X =
1 1 1 1 0 0 0
1 −1 0 0 1 1 0
1 0 −1 0 −1 0 1
1 0 0 −1 0 −1 −1
0 1 −1 0 0 1 −1
0 1 0 −1 1 0 1
0 0 1 −1 −1 1 0
• Every (permutation) automorphism of |X | (entrywise abs
value) can be lifted to an automorphism of X .
• How general is this phenomena?
3
The ALP
The Automprphism Lifting Problem (ALP):
The ALP
Given a {0,1}-matrix |X |, and a subgroup G ⊆ Aut(|X |), find all
{0,−1, 1}-matrices X above |X | and a subgroup G ⊆ Aut(X ),
mapping onto G .
4
Cohomology Developed Matrices
We begin with the following data:
• Let G be a finite group.
• Let X and Y be finite sets with G -action.
• Let O ⊂ X × Y be G -stable.
• Let µ be an Abelian Group, with G action. Let µ+ := µ∪{0}.
Definition
An (µ-valued) X × Y matrix (with support O) is a matrix
F = (f (x , y)) for some function f : X × Y → µ+,
and supp(f ) = O.
• G acts on X × Y matrices:
gF =(gf (g−1x , g−1y)
).
5
Cohomology Developed Matrices
Definition
Two matrices A and B are D-equivalent if
A = D1BD2, Di diagonal.
Write A ∼D B.
Definition
An X × Y F is a Cohomology Developed Matrix=CDM if
∀g ∈ G , gF ∼D F .
• So every g ∈ G induces an automorphism of F .
• But Note: The notion of automorphisms includes twisting
coefficients in µ.
6
Examples
• Group Developed Matrices: Take X = Y = G with actions
(g , x) 7→ xg−1 and (g , y) = gy , and µ with trivial action.
Then F = (f (xy)) satisfies gF = F for all g ∈ G .
• Cocyclic Matrices: Take X ,Y , µ as above, and let
F = (f (xy)ω(x , y)), ω : G × G → µ is a 2-cocycle:
(∗ − cocycle condition) ω(y , z)ω(x , yz) = ω(xy , z)ω(x , y).
Verification
Using the cocycle condition (∗):
ω(xg−1, gy)f (xg−1gy) = ω(g−1, gy)ω(x , y)f (xy)ω(x , g−1)−1
∼D ω(x , y)f (xy).
7
Examples
• Group Developed Matrices: Take X = Y = G with actions
(g , x) 7→ xg−1 and (g , y) = gy , and µ with trivial action.
Then F = (f (xy)) satisfies gF = F for all g ∈ G .
• Cocyclic Matrices: Take X ,Y , µ as above, and let
F = (f (xy)ω(x , y)), ω : G × G → µ is a 2-cocycle:
(∗ − cocycle condition) ω(y , z)ω(x , yz) = ω(xy , z)ω(x , y).
Verification
Using the cocycle condition (∗):
ω(xg−1, gy)f (xg−1gy) = ω(g−1, gy)ω(x , y)f (xy)ω(x , g−1)−1
∼D ω(x , y)f (xy).
7
Examples
• Group Developed Matrices: Take X = Y = G with actions
(g , x) 7→ xg−1 and (g , y) = gy , and µ with trivial action.
Then F = (f (xy)) satisfies gF = F for all g ∈ G .
• Cocyclic Matrices: Take X ,Y , µ as above, and let
F = (f (xy)ω(x , y)), ω : G × G → µ is a 2-cocycle:
(∗ − cocycle condition) ω(y , z)ω(x , yz) = ω(xy , z)ω(x , y).
Verification
Using the cocycle condition (∗):
ω(xg−1, gy)f (xg−1gy) = ω(g−1, gy)ω(x , y)f (xy)ω(x , g−1)−1
∼D ω(x , y)f (xy).
7
Examples-The Fourier Matrix
• The Fourier Matrix: Let G = the Affine Group over Z/n:
G = {x 7→ ax + b | a ∈ (Z/n)×, b ∈ Z/n}.
• Let X = Y = Z/n with the affine G -action.
• Let µ = µn = {n th roots of 1}, with the following twist
action:
(a, b)ζ = ζa2.
• Then F = (exp(2πixy/n)) is a CDM:
(a, b)F = diag(ωb2/2−bx) · F · diag(ωb2/2−by ).
8
Examples-DetFourier
• Let X = Y = (Z/n)2, with the action of
G = {v 7→ Av + a|A ∈ GL2(Z/n), a ∈ (Z/n)2}.
• Again let µ = µn with the twist action
(v 7→ Av + a)ζ = ζdetA.
• Then the n2 × n2 DetFourier Matrix given by
DF (u, v) = exp(2πi det(u, v)/n)
is a CDM. It is also Hermitian and Orthogonal.
9
CDMs and other Mathematics
A list of topics and extensions related to CDM’s:
• Group Cohomology
• The theory of Brauer Groups
• Hecke Algebras and Representation Theory
• Association Schemes and Weights (as of D.G. Higman)
• Constructions of Weighing and Hadamard Matrices
• Simplices and MUB’s in complex projective spaces
• Higher dimensional Design Theory
• (Dual theory) Magic Squares
-Will (not) be slightly discussed.
10
Some Theory
CDM Groups
Definition
Given G ,X ,Y , µ and O ⊂ X × Y as above, let
CDM(G ,O) = {All CDM’s w.r.t. G ,O}.
CDM(G ,O)D = CDM(G ,O)/D-equivalence.
• Both sets are groups w.r.t. the Hadamard Multiplication.
• The group CDM(G ,O)D is best described in terms of a
spectral sequence (not discussed here).
• The elements of CDM(G ,O) have G as a (monomial)
automorphism subgroup, whose permutation data was given
in advance by X ,Y and µ.
11
Cohomology Basics
Let G be a group, acting on a (multiplicative) Abelian group µ.
For any function
f : Gn → µ,
let di f : Gn+1 → µ given by
di f (g0, g1, . . . , gn) =
f (g0, . . . , gi−1gi , . . . , gn) i > 0
g0f (g1, . . . , gn) i = 0.,
and let
df =n∏
i=0
(di f )(−1)i .
Definition
1. f is a cocycle if df = 1,
2. f is a coboundary if f = dh. Coboundaries ⊆ Cocycles.
12
Cohomology Basics
The n-Cohomology Group is
Hn(G , µ) =n − cocycles
n − coboundaries.
If H is is subgroup of G , then there is a homomorphism
res : Hn(G , µ)→ Hn(H, µ),
by restricting cocycles to H.
13
Construction of CDM’s
For the sake of exposition, assume the following:
1) X ,Y are G -transitive, with chosen basepoints x0 ∈ X and
y0 ∈ Y .
2) O is irreducible: If L,R are diagonal and LAR∗ = A for all
O-supported matrices A, then L = R = scalar .
The Data we shall need is:
• G ,X ,Y , µ,O as above
• The stabilizers HX of x0 and HY of y0
• Two 1-Cocycles ψX : HX → µ and ψY : HY → µ
• One 2-Cocycle ω : G × G → µ
It is required that res ω will vanish in H2(HX , µ) and H2(HY , µ).
14
Construction of CDM’s
For the sake of exposition, assume the following:
1) X ,Y are G -transitive, with chosen basepoints x0 ∈ X and
y0 ∈ Y .
2) O is irreducible: If L,R are diagonal and LAR∗ = A for all
O-supported matrices A, then L = R = scalar .
The Data we shall need is:
• G ,X ,Y , µ,O as above
• The stabilizers HX of x0 and HY of y0
• Two 1-Cocycles ψX : HX → µ and ψY : HY → µ
• One 2-Cocycle ω : G × G → µ
It is required that res ω will vanish in H2(HX , µ) and H2(HY , µ).
14
Construction of CDM’s
For the sake of exposition, assume the following:
1) X ,Y are G -transitive, with chosen basepoints x0 ∈ X and
y0 ∈ Y .
2) O is irreducible: If L,R are diagonal and LAR∗ = A for all
O-supported matrices A, then L = R = scalar .
The Data we shall need is:
• G ,X ,Y , µ,O as above
• The stabilizers HX of x0 and HY of y0
• Two 1-Cocycles ψX : HX → µ and ψY : HY → µ
• One 2-Cocycle ω : G × G → µ
It is required that res ω will vanish in H2(HX , µ) and H2(HY , µ).
14
Construction of CDM’s
For the sake of exposition, assume the following:
1) X ,Y are G -transitive, with chosen basepoints x0 ∈ X and
y0 ∈ Y .
2) O is irreducible: If L,R are diagonal and LAR∗ = A for all
O-supported matrices A, then L = R = scalar .
The Data we shall need is:
• G ,X ,Y , µ,O as above
• The stabilizers HX of x0 and HY of y0
• Two 1-Cocycles ψX : HX → µ and ψY : HY → µ
• One 2-Cocycle ω : G × G → µ
It is required that res ω will vanish in H2(HX , µ) and H2(HY , µ).
14
Construction of CDM’s
For the sake of exposition, assume the following:
1) X ,Y are G -transitive, with chosen basepoints x0 ∈ X and
y0 ∈ Y .
2) O is irreducible: If L,R are diagonal and LAR∗ = A for all
O-supported matrices A, then L = R = scalar .
The Data we shall need is:
• G ,X ,Y , µ,O as above
• The stabilizers HX of x0 and HY of y0
• Two 1-Cocycles ψX : HX → µ and ψY : HY → µ
• One 2-Cocycle ω : G × G → µ
It is required that res ω will vanish in H2(HX , µ) and H2(HY , µ).
14
Construction of CDM’s
• Compute trivializations λX and λY s.t. ω|HX×HX= dλX and
ω|HY×HY= dλY .
- For any x ∈ X , y ∈ Y , choose gx , gy ∈ G , s.t.
x = gxx0
y = gyy0.
- Define maps forX : G → HX and forY : G → HY by
forX (g) = g · gg−1x0 ,
forY (g) = g · gg−1y0 .
15
Construction of CDM’s
- For any g ∈ G , x ∈ X , let:
h = forX (g−1x g).
- Let
δg ,x := gx
(ψX (h)λX (h)ω(g−1
x , g)
ω(h, 1)
).
- Finally, let
DX (g) = diag(δg ,x)x
and define DY (g) similarly.
16
Construction of CDM’s
Theorem
- The map A 7→ DX (g)(gA)DY (g)−1 defines a monomial G -action
on O-matrices.
- An invariant matrix A is a CDM.
- All CDM’s arise this way.
Theorem
There is a filtration CDM0 ⊆ CDM1 ⊆ CDM2 := CDMD such that:
CDM0 = {A|gA = A}
CDM1/CDM0 = A subquotient of H1(HX , µ)⊕ H1(HY , µ),
CDM2/CDM1 = A subgroup of H2(G , µ).
17
Orientability
• Having constructed a monomial G -action, we may try toconstruct a matrix by Spreading:
• Choose basepoints (orbit heads) for each G -orbit in
O ⊆ X × Y .
• Fix a value in µ for orbit head.
• Use the G -action to spread the values along the orbits.
Problem
The G -action may give conflicting values to some entries (orbits).
These orbits are said to be non-orientable.
Otherwise the orbit is orientable.
18
Orientability
• Having constructed a monomial G -action, we may try toconstruct a matrix by Spreading:
• Choose basepoints (orbit heads) for each G -orbit in
O ⊆ X × Y .
• Fix a value in µ for orbit head.
• Use the G -action to spread the values along the orbits.
Problem
The G -action may give conflicting values to some entries (orbits).
These orbits are said to be non-orientable.
Otherwise the orbit is orientable.
18
Example G = B3
• Consider the symmetry group of the cube, B3.
• We think of B3 as the group of 3× 3 monomial matrices.
• Let (ei = the standard vectors)
F = Faces of the cube = {±e1,±e2,±e3}
D = Antipodal Edges = {±ei ± ej , i < j}/{±1}.
• There are two B3-orbits in F × D:
1 2 2 1 1 1
1 1 1 1 2 2
2 1 1 2 1 1
2 1 1 2 1 1
1 1 1 1 2 2
1 2 2 1 1 1
19
Example Cont’d
• There are independent characters χ1, χ2 : B3 → µ = {±1}:
χ1(g) = det(g)
χ2(g) = product of signs in g .
• We may form a G -action by A 7→ χ1(g)χ2(g)(gA).
• Then orbit 1 is orientable and orbit 2 isn’t.
• The resulting CDM is
±
1 0 0 1 −1 −1
−1 1 1 −1 0 0
0 −1 −1 0 1 1
0 −1 −1 0 1 1
−1 1 1 −1 0 0
1 0 0 1 −1 −1
.
20
Application to Constructions of
Weighing and Hadamard matrices
Projective Space Weighing Matrices
• The well known W = W(qn+1−1q−1 , q, µn
)weighing matrices (q
a prime power, n|q − 1) happen to be CDM’s.
• In this setting we have
X = The points of Pn(Fq),
Y = The hyperplanes of Pn(Fq),
G = PGL(n + 1,Fq).
• Two orbits in X × Y , by occurence relations.
• Only the non-occuring orbit is orientable.
• The associated 2-cocycle ω is the one associated with the
extension
1→ F×q → GL(n + 1,Fq)→ G → 1.
21
Projective Space Weighing Matrices
• Q: But why is W orthogonal?
• A1: It is balanced. The products Wi ,jW∗i ,k are equi-distributed
along µn.
• A2: WW ∗ is a CDM with two orbits. Only the diagonal is
orientable!
• So non-orientability can be good!
22
Projective Space Weighing Matrices
• Q: But why is W orthogonal?
• A1: It is balanced. The products Wi ,jW∗i ,k are equi-distributed
along µn.
• A2: WW ∗ is a CDM with two orbits. Only the diagonal is
orientable!
• So non-orientability can be good!
22
Projective Space Weighing Matrices
• Q: But why is W orthogonal?
• A1: It is balanced. The products Wi ,jW∗i ,k are equi-distributed
along µn.
• A2: WW ∗ is a CDM with two orbits. Only the diagonal is
orientable!
• So non-orientability can be good!
22
Grassmannian and Flag Weighing Matrices
• By using the power of CDM theory, we may construct:
Grassmannian Weighing Matrices
Under similar conditions
∃W
([d
k
]q
, qk(d−k), µn
)
([ dk
]q=Gaussian Binomial Coefficients).
Flag Variety Weighing Matrices
Let d =∑r
i=0 ki , ki > 0. Then
∃W
([d
k0, k1, . . . , kr
]q
, q∑
i<j kikj , µn
),
Provided that n ≥ r . 23
Some Constructions
The Single/Double Circulant Core Structure:
A Ba b...
...a b
C Dc d...
...c d
e · · · e f · · · f p q
g · · · g h · · · h r s
G = Z/nZ, X = Y = G ∪ G ∪ {∗} ∪ {∗}
24
Single Negacyclic Core of order 2n (n-odd)
a1 a2 a3 a4 a5 a6 a b
−a6 a1 a2 a3 a4 a5 b −a−a5 −a6 a1 a2 a3 a4 −a −b−a4 −a5 −a6 a1 a2 a3 −b a
−a3 −a4 −a5 −a6 a1 a2 a b
−a2 −a3 −a4 −a5 −a6 a1 b −ac d −c −d c −d e f
−d c d −c −d c −f e
G = Z/nZ, X = Y = Z/nZ ∪ Z/2,
Negacyclic action. Such matrices form an Algebra.
25
This is an OD(12, 6, 6)
a −b −b b b b −b −a −a a a a
−b a b −b b b −a −b a −a a a
−b b a b −b b −a a −b a −a a
b −b b a −b b a −a a −b −a a
b b −b −b a b a a −a −a −b a
b b b b b a a a a a a −b
b a a −a −a −a a −b −b b b b
a b −a a −a −a −b a b −b b b
a −a b −a a −a −b b a b −b b
−a a −a b a −a b −b b a −b b
−a −a a a b −a b b −b −b a b
−a −a −a −a −a b b b b b b a
G ' A5 × Z/2Z, X = Y = G/D5 × {0}
Such matrices generate over Q the field Q(√
5,√−1).
26
Orthogonal Pairs
Definition
1) Two (possibly rectangular) CDM’s A,B are an Orthogonal Pair
(OP) if AB∗ is nowhere orienatble.
2) A tuple of CDM’s (A1, . . . ,Am) is an Orthogonal Tuple if each
(Ai ,Aj) is an OP.
• If (A,B) is an OP, then necessarily AB∗ = 0.
• Orthogonality is a consequence of the group action. So one
obtains symbolic (A,B) which are formally orthogonal.
• One may substitute (rectangular) blocks in place of symbols.
No commutativity/amicability restrictions.
27
Here is an example of an OP:
A =
−a −b a −b −a −b b b −b a b b
−b a b b b −b −b b −a −b a −a−b b −b −a b a −a a −b b −b b
−b b −b a b −a a −a −b b −b b
−a b a b −a b −b −b b a −b −b−b −a b b b −b −b b a −b −a a
B =
c −d −c −d −c −d d d −d c d d
−d c d −d d −d d d c −d −c −cd d −d −c −d −c c c d d −d −d−d −d d c d c −c −c −d −d d d
−c d c d c d −d −d d −c −d −dd −c −d d −d d −d −d −c d c c
.
28
This is a W (16, 8) constructed from OT’s. Each 16× 8 column isan OT .
1 0 1 0 −1 −1 0 0 0 −1 0 −1 0 0 1 1
−1 0 −1 0 0 0 1 −1 0 −1 0 −1 1 1 0 0
0 −1 0 1 1 1 0 0 −1 0 −1 0 0 0 1 1
0 −1 0 1 0 0 1 −1 1 0 1 0 −1 −1 0 0
0 0 0 0 −1 −1 1 −1 −1 1 −1 1 0 0 0 0
−1 1 −1 −1 0 0 0 0 0 0 0 0 −1 −1 1 1
0 −1 0 −1 0 0 −1 −1 1 0 −1 0 −1 1 0 0
0 1 0 1 −1 1 0 0 1 0 −1 0 0 0 1 −1
1 0 −1 0 0 0 1 1 0 1 0 −1 −1 1 0 0
−1 0 1 0 1 −1 0 0 0 1 0 −1 0 0 1 −1
−1 1 1 1 0 0 0 0 0 0 0 0 −1 1 −1 1
0 0 0 0 1 −1 1 1 1 −1 −1 1 0 0 0 0
1 1 0 0 1 0 0 −1 1 1 0 0 1 0 0 1
0 0 1 −1 0 1 1 0 0 0 1 1 0 1 1 0
−1 −1 0 0 −1 0 0 1 1 1 0 0 1 0 0 1
0 0 −1 1 0 −1 −1 0 0 0 1 1 0 1 1 0
G = B3 (Symmetry Group of the Cube)
Sets: X1 = X2 = Edges/{±1}, X3 = X4 = Orientations
Y1 = Y2 =Vertices,
Cohomology Classes: trivial 2-cocycle, nontrivial 1-cocycles.
29
A similar Hadamard 16 example (Same Group, different Sets)
1 −1 −1 1 −1 1 −1 1 1 −1 1 1 −1 −1 1 −1
−1 1 1 −1 −1 1 −1 1 −1 1 −1 −1 −1 −1 1 −1
1 1 −1 −1 −1 1 1 −1 −1 −1 −1 1 1 1 1 −1
1 1 −1 −1 −1 1 1 −1 1 1 1 −1 −1 −1 −1 1
−1 1 1 −1 1 −1 1 −1 1 −1 1 1 −1 −1 1 −1
−1 1 1 −1 −1 1 −1 1 1 −1 1 1 1 1 −1 1
1 −1 1 −1 −1 −1 1 1 1 −1 −1 −1 −1 1 −1 −1
−1 1 −1 1 −1 −1 1 1 −1 1 1 1 −1 1 −1 −1
1 1 1 1 −1 −1 −1 −1 −1 −1 1 −1 1 −1 −1 −1
−1 −1 −1 −1 1 1 1 1 −1 −1 1 −1 1 −1 −1 −1
1 −1 1 −1 −1 −1 1 1 −1 1 1 1 1 −1 1 1
1 −1 1 −1 1 1 −1 −1 −1 1 1 1 −1 1 −1 −1
1 1 1 1 1 1 1 1 1 1 1 −1 1 1 1 −1
−1 −1 1 1 −1 1 1 −1 1 1 −1 1 1 −1 −1 −1
1 1 −1 −1 1 −1 −1 1 1 1 −1 1 1 −1 −1 −1
1 1 1 1 1 1 1 1 −1 −1 −1 1 −1 −1 −1 1
30