Finite geometry, designs, codes, and Hamada's conjecture
Transcript of Finite geometry, designs, codes, and Hamada's conjecture
![Page 1: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/1.jpg)
Finite geometry, designs, codes, and Hamada’sconjecture
Vladimir D. Tonchev
Michigan Technological University
ASI, Opatija, May 31 - June 11, 2010
1 / 67
![Page 2: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/2.jpg)
Outline1 Designs2 Finite Geometries3 Geometric Designs4 Linear Codes5 Majority Logic Decodable Codes6 Codes that Support t-Designs7 The p-Ranks of Geometric Designs8 Hamada’s Conjecture9 The Proven Cases
10 A Revision of Hamada’s Conjecture11 The Uniqueness Question12 Non-Geometric Designs with the Same p-Rank as Geometric Ones13 Designs from Polarities in PG(n,q)
14 The p-Rank of Polarity Designs15 A Generalization to the Affine Case16 Exponential Bounds17 Open Problems
2 / 67
![Page 3: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/3.jpg)
Designs
A t-(v , k , λ) design D=(X ,B) is a set X of points and a collection B ofsubsets of X called blocks such that:
|X | = v ,|B| = k for each B ∈ B, andEvery t-subset of X s contained in exactly λ blocks.
A 2-designs (t = 2) a called a Balanced Incomplete Block Design orBIBD.
3 / 67
![Page 4: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/4.jpg)
Designs
A t-(v , k , λ) design D=(X ,B) is a set X of points and a collection B ofsubsets of X called blocks such that:
|X | = v ,|B| = k for each B ∈ B, andEvery t-subset of X s contained in exactly λ blocks.
A 2-designs (t = 2) a called a Balanced Incomplete Block Design orBIBD.
3 / 67
![Page 5: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/5.jpg)
Designs
A t-(v , k , λ) design D=(X ,B) is a set X of points and a collection B ofsubsets of X called blocks such that:
|X | = v ,|B| = k for each B ∈ B, andEvery t-subset of X s contained in exactly λ blocks.
A 2-designs (t = 2) a called a Balanced Incomplete Block Design orBIBD.
3 / 67
![Page 6: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/6.jpg)
Designs
A t-(v , k , λ) design D=(X ,B) is a set X of points and a collection B ofsubsets of X called blocks such that:
|X | = v ,|B| = k for each B ∈ B, andEvery t-subset of X s contained in exactly λ blocks.
A 2-designs (t = 2) a called a Balanced Incomplete Block Design orBIBD.
3 / 67
![Page 7: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/7.jpg)
Designs
A t-(v , k , λ) design D=(X ,B) is a set X of points and a collection B ofsubsets of X called blocks such that:
|X | = v ,|B| = k for each B ∈ B, andEvery t-subset of X s contained in exactly λ blocks.
A 2-designs (t = 2) a called a Balanced Incomplete Block Design orBIBD.
3 / 67
![Page 8: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/8.jpg)
Designs
A t-(v , k , λ) design D=(X ,B) is a set X of points and a collection B ofsubsets of X called blocks such that:
|X | = v ,|B| = k for each B ∈ B, andEvery t-subset of X s contained in exactly λ blocks.
A 2-designs (t = 2) a called a Balanced Incomplete Block Design orBIBD.
3 / 67
![Page 9: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/9.jpg)
A small example
A
B
C
D
G
E F
A 2-(7,3,1) design
4 / 67
![Page 10: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/10.jpg)
Properties
The t-designs are highly regular:
If 0 ≤ i ≤ t , any i-subset appears in λi = λ(v−i
t−i
)/(k−i
t−i
)blocks.
i = 0 : Total number of blocks is b = λ(v
t
)/(k
t
)i = 1: Any point x appears in r blocks, where
r = λ1 = λ
(v − 1t − 1
)/
(k − 1t − 1
).
5 / 67
![Page 11: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/11.jpg)
Properties
The t-designs are highly regular:
If 0 ≤ i ≤ t , any i-subset appears in λi = λ(v−i
t−i
)/(k−i
t−i
)blocks.
i = 0 : Total number of blocks is b = λ(v
t
)/(k
t
)i = 1: Any point x appears in r blocks, where
r = λ1 = λ
(v − 1t − 1
)/
(k − 1t − 1
).
5 / 67
![Page 12: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/12.jpg)
Properties
The t-designs are highly regular:
If 0 ≤ i ≤ t , any i-subset appears in λi = λ(v−i
t−i
)/(k−i
t−i
)blocks.
i = 0 : Total number of blocks is b = λ(v
t
)/(k
t
)i = 1: Any point x appears in r blocks, where
r = λ1 = λ
(v − 1t − 1
)/
(k − 1t − 1
).
5 / 67
![Page 13: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/13.jpg)
Properties
The t-designs are highly regular:
If 0 ≤ i ≤ t , any i-subset appears in λi = λ(v−i
t−i
)/(k−i
t−i
)blocks.
i = 0 : Total number of blocks is b = λ(v
t
)/(k
t
)i = 1: Any point x appears in r blocks, where
r = λ1 = λ
(v − 1t − 1
)/
(k − 1t − 1
).
5 / 67
![Page 14: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/14.jpg)
Incidence Matrices
The incidence matrix of a t-(v , k , λ) design is a b × v (0,1) matrixwhose (i , j) entry is 1 if block i contains point j , and 0 otherwise.
The 2-(7,3,1) Design:
A B C D E F GB1 1 1 1 0 0 0 0B2 1 0 0 1 1 0 0B3 1 0 0 0 0 1 1B4 0 1 0 1 0 1 0B5 0 1 0 0 1 0 1B6 0 0 1 1 0 0 1B7 0 0 1 0 1 1 0
6 / 67
![Page 15: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/15.jpg)
Incidence Matrices
The incidence matrix of a t-(v , k , λ) design is a b × v (0,1) matrixwhose (i , j) entry is 1 if block i contains point j , and 0 otherwise.
The 2-(7,3,1) Design:
A B C D E F GB1 1 1 1 0 0 0 0B2 1 0 0 1 1 0 0B3 1 0 0 0 0 1 1B4 0 1 0 1 0 1 0B5 0 1 0 0 1 0 1B6 0 0 1 1 0 0 1B7 0 0 1 0 1 1 0
6 / 67
![Page 16: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/16.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 17: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/17.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 18: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/18.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 19: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/19.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 20: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/20.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 21: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/21.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 22: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/22.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 23: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/23.jpg)
Finite Geometries
Projective Geometry PG(n,q)
points are the 1-dimensional subspaces of Fn+1q .
lines are the 2-dimensional subspaces of Fn+1q
k -dimensional subspaces are the (k + 1)-dimensionalsubspaces of Fn+1
q .
Affine Geometry AG(n,q)
points are the vectors of Fnq
lines are the 1-dimensional subspaces of Fnq and their cosets
k -dimensional subspaces are the k -dimensional subspaces ofFn
q and their cosets (called k -flats).
7 / 67
![Page 24: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/24.jpg)
Geometric Designs
A geometric design is formed from the points and d-subspaces ofPG(n,q) or AG(n,q).
The projective geometry design PGd (n,q):
2−(
qn+1 − 1q − 1
,qd+1 − 1
q − 1,
(qn+1 − q2)(qn+1 − q3) · · · (qn+1 − qd )
(qd+1 − q2)(qd+1 − q3) · · · (qd+1 − qd )
)
The affine geometry design AGd (n,q):
2−(
qn,qd ,(qn − q)(qn − q2) · · · (qn − qd−1)
(qd − q)(qd − q2) · · · (qd − qd−1)
)If q = 2, AGd (n,2) is also a 3-(2n,2d , (2n−22)···(2n−2d−1)
(2d−22)···(2d−2d−1)design.
8 / 67
![Page 25: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/25.jpg)
Geometric Designs
A geometric design is formed from the points and d-subspaces ofPG(n,q) or AG(n,q).
The projective geometry design PGd (n,q):
2−(
qn+1 − 1q − 1
,qd+1 − 1
q − 1,
(qn+1 − q2)(qn+1 − q3) · · · (qn+1 − qd )
(qd+1 − q2)(qd+1 − q3) · · · (qd+1 − qd )
)
The affine geometry design AGd (n,q):
2−(
qn,qd ,(qn − q)(qn − q2) · · · (qn − qd−1)
(qd − q)(qd − q2) · · · (qd − qd−1)
)If q = 2, AGd (n,2) is also a 3-(2n,2d , (2n−22)···(2n−2d−1)
(2d−22)···(2d−2d−1)design.
8 / 67
![Page 26: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/26.jpg)
Geometric Designs
A geometric design is formed from the points and d-subspaces ofPG(n,q) or AG(n,q).
The projective geometry design PGd (n,q):
2−(
qn+1 − 1q − 1
,qd+1 − 1
q − 1,
(qn+1 − q2)(qn+1 − q3) · · · (qn+1 − qd )
(qd+1 − q2)(qd+1 − q3) · · · (qd+1 − qd )
)
The affine geometry design AGd (n,q):
2−(
qn,qd ,(qn − q)(qn − q2) · · · (qn − qd−1)
(qd − q)(qd − q2) · · · (qd − qd−1)
)If q = 2, AGd (n,2) is also a 3-(2n,2d , (2n−22)···(2n−2d−1)
(2d−22)···(2d−2d−1)design.
8 / 67
![Page 27: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/27.jpg)
Geometric Designs
A geometric design is formed from the points and d-subspaces ofPG(n,q) or AG(n,q).
The projective geometry design PGd (n,q):
2−(
qn+1 − 1q − 1
,qd+1 − 1
q − 1,
(qn+1 − q2)(qn+1 − q3) · · · (qn+1 − qd )
(qd+1 − q2)(qd+1 − q3) · · · (qd+1 − qd )
)
The affine geometry design AGd (n,q):
2−(
qn,qd ,(qn − q)(qn − q2) · · · (qn − qd−1)
(qd − q)(qd − q2) · · · (qd − qd−1)
)If q = 2, AGd (n,2) is also a 3-(2n,2d , (2n−22)···(2n−2d−1)
(2d−22)···(2d−2d−1)design.
8 / 67
![Page 28: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/28.jpg)
A small examplei: PG1(2, 2)
A
B
C
D
G
E F
PG1(2,2): The projective plane of order 2
9 / 67
![Page 29: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/29.jpg)
Affine Geometry Designs are Resolvable
AG1(2, 3), or a 2-(9, 3, 1)-design
00
01
02
10
11
12
20
21
22
00
01
02
10
11
12
20
21
22
00
01
02
10
11
12
20
21
22
00
01
02
10
11
12
20
21
22
This design is resolvable into parallel classes.
10 / 67
![Page 30: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/30.jpg)
Linear error-correcting codes
Linear codeA linear q-ary [n, k ,d ] code C is a k -dimensional subspace of then-dimensional vector space over the field GF (q) of order q withminimum Hamming distance d .A code with minimum distance d can correct up to e = [(d − 1)/2]errors.
Dual code
The dual code C⊥ of an [n, k ] code C is the [n,n − k ] code defined by
C⊥ = {y ∈ GF (q)n | y · x = 0 for all x ∈ C}
Parity check matrix
A matrix H of q-rank n − k whose rows are vectors from C⊥ is a paritycheck matrix of C.
11 / 67
![Page 31: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/31.jpg)
Linear error-correcting codes
Linear codeA linear q-ary [n, k ,d ] code C is a k -dimensional subspace of then-dimensional vector space over the field GF (q) of order q withminimum Hamming distance d .A code with minimum distance d can correct up to e = [(d − 1)/2]errors.
Dual code
The dual code C⊥ of an [n, k ] code C is the [n,n − k ] code defined by
C⊥ = {y ∈ GF (q)n | y · x = 0 for all x ∈ C}
Parity check matrix
A matrix H of q-rank n − k whose rows are vectors from C⊥ is a paritycheck matrix of C.
11 / 67
![Page 32: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/32.jpg)
Linear error-correcting codes
Linear codeA linear q-ary [n, k ,d ] code C is a k -dimensional subspace of then-dimensional vector space over the field GF (q) of order q withminimum Hamming distance d .A code with minimum distance d can correct up to e = [(d − 1)/2]errors.
Dual code
The dual code C⊥ of an [n, k ] code C is the [n,n − k ] code defined by
C⊥ = {y ∈ GF (q)n | y · x = 0 for all x ∈ C}
Parity check matrix
A matrix H of q-rank n − k whose rows are vectors from C⊥ is a paritycheck matrix of C.
11 / 67
![Page 33: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/33.jpg)
Linear error-correcting codes
Linear codeA linear q-ary [n, k ,d ] code C is a k -dimensional subspace of then-dimensional vector space over the field GF (q) of order q withminimum Hamming distance d .A code with minimum distance d can correct up to e = [(d − 1)/2]errors.
Dual code
The dual code C⊥ of an [n, k ] code C is the [n,n − k ] code defined by
C⊥ = {y ∈ GF (q)n | y · x = 0 for all x ∈ C}
Parity check matrix
A matrix H of q-rank n − k whose rows are vectors from C⊥ is a paritycheck matrix of C.
11 / 67
![Page 34: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/34.jpg)
Majority logic decoding algorithmIf a codeword x = (x1, . . . , xn) ∈ C is sent over a communicationchannel, and a vector y = (y1, . . . , yn) is received, for each coordinatei , 1 ≤ i ≤ n, the values
y (1)i , . . . , y (ri )
i (1)
of ri linear functions are computed, and yi is decoded as the mostfrequent among the values (1).
Theorem. (Rudolph, 1967)
If C is a linear [n, k ] code such that C⊥ contains a set S of vectors ofweight w whose supports are the blocks of a 2-(n,w , λ) design, thecode C can correct up to
e =
[r + λ− 1
2λ
]errors by majority logic decoding, where r = λ1 = λ(n − 1)/(w − 1).
12 / 67
![Page 35: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/35.jpg)
Majority logic decoding algorithmIf a codeword x = (x1, . . . , xn) ∈ C is sent over a communicationchannel, and a vector y = (y1, . . . , yn) is received, for each coordinatei , 1 ≤ i ≤ n, the values
y (1)i , . . . , y (ri )
i (1)
of ri linear functions are computed, and yi is decoded as the mostfrequent among the values (1).
Theorem. (Rudolph, 1967)
If C is a linear [n, k ] code such that C⊥ contains a set S of vectors ofweight w whose supports are the blocks of a 2-(n,w , λ) design, thecode C can correct up to
e =
[r + λ− 1
2λ
]errors by majority logic decoding, where r = λ1 = λ(n − 1)/(w − 1).
12 / 67
![Page 36: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/36.jpg)
Majority logic decoding algorithmIf a codeword x = (x1, . . . , xn) ∈ C is sent over a communicationchannel, and a vector y = (y1, . . . , yn) is received, for each coordinatei , 1 ≤ i ≤ n, the values
y (1)i , . . . , y (ri )
i (1)
of ri linear functions are computed, and yi is decoded as the mostfrequent among the values (1).
Theorem. (Rudolph, 1967)
If C is a linear [n, k ] code such that C⊥ contains a set S of vectors ofweight w whose supports are the blocks of a 2-(n,w , λ) design, thecode C can correct up to
e =
[r + λ− 1
2λ
]errors by majority logic decoding, where r = λ1 = λ(n − 1)/(w − 1).
12 / 67
![Page 37: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/37.jpg)
Majority logic decoding algorithmIf a codeword x = (x1, . . . , xn) ∈ C is sent over a communicationchannel, and a vector y = (y1, . . . , yn) is received, for each coordinatei , 1 ≤ i ≤ n, the values
y (1)i , . . . , y (ri )
i (1)
of ri linear functions are computed, and yi is decoded as the mostfrequent among the values (1).
Theorem. (Rudolph, 1967)
If C is a linear [n, k ] code such that C⊥ contains a set S of vectors ofweight w whose supports are the blocks of a 2-(n,w , λ) design, thecode C can correct up to
e =
[r + λ− 1
2λ
]errors by majority logic decoding, where r = λ1 = λ(n − 1)/(w − 1).
12 / 67
![Page 38: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/38.jpg)
Sketch of proof.
If a = (a1, . . . ,an) ∈ S then
a1x1 + · · ·+ anxn = 0
for every x ∈ C.
NoteDue to possible errors in the received vector y = (y1, . . . , yn),
a1y1 + · · ·+ anyn
may or may not be zero.
13 / 67
![Page 39: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/39.jpg)
Sketch of proof.
If a = (a1, . . . ,an) ∈ S then
a1x1 + · · ·+ anxn = 0
for every x ∈ C.
NoteDue to possible errors in the received vector y = (y1, . . . , yn),
a1y1 + · · ·+ anyn
may or may not be zero.
13 / 67
![Page 40: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/40.jpg)
Sketch of proof.
If a = (a1, . . . ,an) ∈ S then
a1x1 + · · ·+ anxn = 0
for every x ∈ C.
NoteDue to possible errors in the received vector y = (y1, . . . , yn),
a1y1 + · · ·+ anyn
may or may not be zero.
13 / 67
![Page 41: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/41.jpg)
Sketch of proof.
If a = (a1, . . . ,an) ∈ S then
a1x1 + · · ·+ anxn = 0
for every x ∈ C.
NoteDue to possible errors in the received vector y = (y1, . . . , yn),
a1y1 + · · ·+ anyn
may or may not be zero.
13 / 67
![Page 42: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/42.jpg)
Assume thata1y1 + · · ·+ anyn = 0
and ai 6= 0. Thenyi = −a1
aiy1 − · · · −
an
aiyn.
Linear functions fj for decoding yi :
For each i , 1 ≤ i ≤ n, the set S contains r vectors
a(j) = (a(j)1 , . . . ,a(j)
n ), j = 1, . . . , r
such that a(j)i 6= 0.
We define a set of r + λ linear functions fj = fj(y1, . . . , yn),
fj = −a(j)
1
a(j)i
y1 − · · · −a(j)
n
a(j)i
yn, j = 1, . . . , r ,
fr+1 = fr+2 = · · · fr+λ = yi .14 / 67
![Page 43: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/43.jpg)
Assume thata1y1 + · · ·+ anyn = 0
and ai 6= 0. Thenyi = −a1
aiy1 − · · · −
an
aiyn.
Linear functions fj for decoding yi :
For each i , 1 ≤ i ≤ n, the set S contains r vectors
a(j) = (a(j)1 , . . . ,a(j)
n ), j = 1, . . . , r
such that a(j)i 6= 0.
We define a set of r + λ linear functions fj = fj(y1, . . . , yn),
fj = −a(j)
1
a(j)i
y1 − · · · −a(j)
n
a(j)i
yn, j = 1, . . . , r ,
fr+1 = fr+2 = · · · fr+λ = yi .14 / 67
![Page 44: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/44.jpg)
Assume thata1y1 + · · ·+ anyn = 0
and ai 6= 0. Thenyi = −a1
aiy1 − · · · −
an
aiyn.
Linear functions fj for decoding yi :
For each i , 1 ≤ i ≤ n, the set S contains r vectors
a(j) = (a(j)1 , . . . ,a(j)
n ), j = 1, . . . , r
such that a(j)i 6= 0.
We define a set of r + λ linear functions fj = fj(y1, . . . , yn),
fj = −a(j)
1
a(j)i
y1 − · · · −a(j)
n
a(j)i
yn, j = 1, . . . , r ,
fr+1 = fr+2 = · · · fr+λ = yi .14 / 67
![Page 45: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/45.jpg)
Assume thata1y1 + · · ·+ anyn = 0
and ai 6= 0. Thenyi = −a1
aiy1 − · · · −
an
aiyn.
Linear functions fj for decoding yi :
For each i , 1 ≤ i ≤ n, the set S contains r vectors
a(j) = (a(j)1 , . . . ,a(j)
n ), j = 1, . . . , r
such that a(j)i 6= 0.
We define a set of r + λ linear functions fj = fj(y1, . . . , yn),
fj = −a(j)
1
a(j)i
y1 − · · · −a(j)
n
a(j)i
yn, j = 1, . . . , r ,
fr+1 = fr+2 = · · · fr+λ = yi .14 / 67
![Page 46: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/46.jpg)
Assume thata1y1 + · · ·+ anyn = 0
and ai 6= 0. Thenyi = −a1
aiy1 − · · · −
an
aiyn.
Linear functions fj for decoding yi :
For each i , 1 ≤ i ≤ n, the set S contains r vectors
a(j) = (a(j)1 , . . . ,a(j)
n ), j = 1, . . . , r
such that a(j)i 6= 0.
We define a set of r + λ linear functions fj = fj(y1, . . . , yn),
fj = −a(j)
1
a(j)i
y1 − · · · −a(j)
n
a(j)i
yn, j = 1, . . . , r ,
fr+1 = fr+2 = · · · fr+λ = yi .14 / 67
![Page 47: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/47.jpg)
If there are no errors in y = (y1, . . . , yn), then
y1 = x1, . . . yn = xn,
andfj = xi for all j = 1, . . . , r + λ.
Any erroneous component ym appears in at most λ of the functionsf1, . . . , fr+λ.
Thus, if there are e errors in y = (y1, . . . , yn), and
eλ <r + λ
2,
the majority of the values fj(y1, . . . , yn) will be equal to xi .
15 / 67
![Page 48: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/48.jpg)
If there are no errors in y = (y1, . . . , yn), then
y1 = x1, . . . yn = xn,
andfj = xi for all j = 1, . . . , r + λ.
Any erroneous component ym appears in at most λ of the functionsf1, . . . , fr+λ.
Thus, if there are e errors in y = (y1, . . . , yn), and
eλ <r + λ
2,
the majority of the values fj(y1, . . . , yn) will be equal to xi .
15 / 67
![Page 49: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/49.jpg)
If there are no errors in y = (y1, . . . , yn), then
y1 = x1, . . . yn = xn,
andfj = xi for all j = 1, . . . , r + λ.
Any erroneous component ym appears in at most λ of the functionsf1, . . . , fr+λ.
Thus, if there are e errors in y = (y1, . . . , yn), and
eλ <r + λ
2,
the majority of the values fj(y1, . . . , yn) will be equal to xi .
15 / 67
![Page 50: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/50.jpg)
If there are no errors in y = (y1, . . . , yn), then
y1 = x1, . . . yn = xn,
andfj = xi for all j = 1, . . . , r + λ.
Any erroneous component ym appears in at most λ of the functionsf1, . . . , fr+λ.
Thus, if there are e errors in y = (y1, . . . , yn), and
eλ <r + λ
2,
the majority of the values fj(y1, . . . , yn) will be equal to xi .
15 / 67
![Page 51: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/51.jpg)
If there are no errors in y = (y1, . . . , yn), then
y1 = x1, . . . yn = xn,
andfj = xi for all j = 1, . . . , r + λ.
Any erroneous component ym appears in at most λ of the functionsf1, . . . , fr+λ.
Thus, if there are e errors in y = (y1, . . . , yn), and
eλ <r + λ
2,
the majority of the values fj(y1, . . . , yn) will be equal to xi .
15 / 67
![Page 52: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/52.jpg)
If there are no errors in y = (y1, . . . , yn), then
y1 = x1, . . . yn = xn,
andfj = xi for all j = 1, . . . , r + λ.
Any erroneous component ym appears in at most λ of the functionsf1, . . . , fr+λ.
Thus, if there are e errors in y = (y1, . . . , yn), and
eλ <r + λ
2,
the majority of the values fj(y1, . . . , yn) will be equal to xi .
15 / 67
![Page 53: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/53.jpg)
Variations and Generalizations
Rahman and Blake, 1975:Rudolph’s bound can be improved if If C⊥ supports a t-design witht ≥ 2.If t = 1, λ can be replaced with the maximum frequency ofappearance of pairs of points.If t = 0, r can be replaced with the minimum frequency ofappearance of a point in blocks.
Milti-step majority logic decodingRudolph’s algorithm is an example of one-step majority logic decoding.
There is an iterative multistep version of the algorithm consisting of asequence of one-step decoding of linear combinations of received bits,followed by computing the individual bits yi as a solution of a system oflinear equations.
16 / 67
![Page 54: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/54.jpg)
Variations and Generalizations
Rahman and Blake, 1975:Rudolph’s bound can be improved if If C⊥ supports a t-design witht ≥ 2.If t = 1, λ can be replaced with the maximum frequency ofappearance of pairs of points.If t = 0, r can be replaced with the minimum frequency ofappearance of a point in blocks.
Milti-step majority logic decodingRudolph’s algorithm is an example of one-step majority logic decoding.
There is an iterative multistep version of the algorithm consisting of asequence of one-step decoding of linear combinations of received bits,followed by computing the individual bits yi as a solution of a system oflinear equations.
16 / 67
![Page 55: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/55.jpg)
Variations and Generalizations
Rahman and Blake, 1975:Rudolph’s bound can be improved if If C⊥ supports a t-design witht ≥ 2.If t = 1, λ can be replaced with the maximum frequency ofappearance of pairs of points.If t = 0, r can be replaced with the minimum frequency ofappearance of a point in blocks.
Milti-step majority logic decodingRudolph’s algorithm is an example of one-step majority logic decoding.
There is an iterative multistep version of the algorithm consisting of asequence of one-step decoding of linear combinations of received bits,followed by computing the individual bits yi as a solution of a system oflinear equations.
16 / 67
![Page 56: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/56.jpg)
Variations and Generalizations
Rahman and Blake, 1975:Rudolph’s bound can be improved if If C⊥ supports a t-design witht ≥ 2.If t = 1, λ can be replaced with the maximum frequency ofappearance of pairs of points.If t = 0, r can be replaced with the minimum frequency ofappearance of a point in blocks.
Milti-step majority logic decodingRudolph’s algorithm is an example of one-step majority logic decoding.
There is an iterative multistep version of the algorithm consisting of asequence of one-step decoding of linear combinations of received bits,followed by computing the individual bits yi as a solution of a system oflinear equations.
16 / 67
![Page 57: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/57.jpg)
Variations and Generalizations
Rahman and Blake, 1975:Rudolph’s bound can be improved if If C⊥ supports a t-design witht ≥ 2.If t = 1, λ can be replaced with the maximum frequency ofappearance of pairs of points.If t = 0, r can be replaced with the minimum frequency ofappearance of a point in blocks.
Milti-step majority logic decodingRudolph’s algorithm is an example of one-step majority logic decoding.
There is an iterative multistep version of the algorithm consisting of asequence of one-step decoding of linear combinations of received bits,followed by computing the individual bits yi as a solution of a system oflinear equations.
16 / 67
![Page 58: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/58.jpg)
Variations and Generalizations
Rahman and Blake, 1975:Rudolph’s bound can be improved if If C⊥ supports a t-design witht ≥ 2.If t = 1, λ can be replaced with the maximum frequency ofappearance of pairs of points.If t = 0, r can be replaced with the minimum frequency ofappearance of a point in blocks.
Milti-step majority logic decodingRudolph’s algorithm is an example of one-step majority logic decoding.
There is an iterative multistep version of the algorithm consisting of asequence of one-step decoding of linear combinations of received bits,followed by computing the individual bits yi as a solution of a system oflinear equations.
16 / 67
![Page 59: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/59.jpg)
Variations and Generalizations
Rahman and Blake, 1975:Rudolph’s bound can be improved if If C⊥ supports a t-design witht ≥ 2.If t = 1, λ can be replaced with the maximum frequency ofappearance of pairs of points.If t = 0, r can be replaced with the minimum frequency ofappearance of a point in blocks.
Milti-step majority logic decodingRudolph’s algorithm is an example of one-step majority logic decoding.
There is an iterative multistep version of the algorithm consisting of asequence of one-step decoding of linear combinations of received bits,followed by computing the individual bits yi as a solution of a system oflinear equations.
16 / 67
![Page 60: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/60.jpg)
Which codes support t-designs?
Task
Find a linear code C so that C⊥ supports a t-design with t ≥ 2.
The Assmus-Mattson Theorem, 1969If C is a linear [n, k ] code with minimum distance d such that thenumber of distinct nonzero weights in C⊥ not exceeding n − t issmaller than d − t , then both C and C⊥ support t-designs.
NoteThe Assmus-Mattson Theorem gives a sufficient condition for theexistence of designs in a code.It does not specify how one can find such codes.
17 / 67
![Page 61: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/61.jpg)
Which codes support t-designs?
Task
Find a linear code C so that C⊥ supports a t-design with t ≥ 2.
The Assmus-Mattson Theorem, 1969If C is a linear [n, k ] code with minimum distance d such that thenumber of distinct nonzero weights in C⊥ not exceeding n − t issmaller than d − t , then both C and C⊥ support t-designs.
NoteThe Assmus-Mattson Theorem gives a sufficient condition for theexistence of designs in a code.It does not specify how one can find such codes.
17 / 67
![Page 62: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/62.jpg)
Which codes support t-designs?
Task
Find a linear code C so that C⊥ supports a t-design with t ≥ 2.
The Assmus-Mattson Theorem, 1969If C is a linear [n, k ] code with minimum distance d such that thenumber of distinct nonzero weights in C⊥ not exceeding n − t issmaller than d − t , then both C and C⊥ support t-designs.
NoteThe Assmus-Mattson Theorem gives a sufficient condition for theexistence of designs in a code.It does not specify how one can find such codes.
17 / 67
![Page 63: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/63.jpg)
Which codes support t-designs?
Task
Find a linear code C so that C⊥ supports a t-design with t ≥ 2.
The Assmus-Mattson Theorem, 1969If C is a linear [n, k ] code with minimum distance d such that thenumber of distinct nonzero weights in C⊥ not exceeding n − t issmaller than d − t , then both C and C⊥ support t-designs.
NoteThe Assmus-Mattson Theorem gives a sufficient condition for theexistence of designs in a code.It does not specify how one can find such codes.
17 / 67
![Page 64: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/64.jpg)
Codes with mulitransitive automorphism groupsIf C admits an automorphism group of permutations that actst-transitive (or t-homogeneously) on the set of n code coordinates,then the supports of all codewords of any nonzero weight form at-design.
ExampleThe binary Golay [24,12,8] code and the ternary Golay [12,6,6] codesupport 5-designs.
18 / 67
![Page 65: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/65.jpg)
Codes with mulitransitive automorphism groupsIf C admits an automorphism group of permutations that actst-transitive (or t-homogeneously) on the set of n code coordinates,then the supports of all codewords of any nonzero weight form at-design.
ExampleThe binary Golay [24,12,8] code and the ternary Golay [12,6,6] codesupport 5-designs.
18 / 67
![Page 66: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/66.jpg)
A simple construction using incidence matrices
If C is a linear code over GF (q) of length v with a parity check matrixH being the block by point b × v incidence matrix of a t-(v ,w , λ)design D, then C⊥ supports the t-(v ,w , λ) design D.The dimension of C is k = v − rankqH.
A possible drawback:
Fisher inequalityIf D is a t-(v ,w , λ) design with b blocks such that t ≥ 2 and v > w > 0,then
b ≥ v .
Thus, it can happen that rankqH = v and dim(C) = 0.
19 / 67
![Page 67: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/67.jpg)
A simple construction using incidence matrices
If C is a linear code over GF (q) of length v with a parity check matrixH being the block by point b × v incidence matrix of a t-(v ,w , λ)design D, then C⊥ supports the t-(v ,w , λ) design D.The dimension of C is k = v − rankqH.
A possible drawback:
Fisher inequalityIf D is a t-(v ,w , λ) design with b blocks such that t ≥ 2 and v > w > 0,then
b ≥ v .
Thus, it can happen that rankqH = v and dim(C) = 0.
19 / 67
![Page 68: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/68.jpg)
A simple construction using incidence matrices
If C is a linear code over GF (q) of length v with a parity check matrixH being the block by point b × v incidence matrix of a t-(v ,w , λ)design D, then C⊥ supports the t-(v ,w , λ) design D.The dimension of C is k = v − rankqH.
A possible drawback:
Fisher inequalityIf D is a t-(v ,w , λ) design with b blocks such that t ≥ 2 and v > w > 0,then
b ≥ v .
Thus, it can happen that rankqH = v and dim(C) = 0.
19 / 67
![Page 69: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/69.jpg)
A simple construction using incidence matrices
If C is a linear code over GF (q) of length v with a parity check matrixH being the block by point b × v incidence matrix of a t-(v ,w , λ)design D, then C⊥ supports the t-(v ,w , λ) design D.The dimension of C is k = v − rankqH.
A possible drawback:
Fisher inequalityIf D is a t-(v ,w , λ) design with b blocks such that t ≥ 2 and v > w > 0,then
b ≥ v .
Thus, it can happen that rankqH = v and dim(C) = 0.
19 / 67
![Page 70: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/70.jpg)
NoteIf H is the block by point b × v incidence matrix of a t-(v ,w , λ) designand r = λ(v − 1)/(w − 1) then
det(HT H) = rw(r − λ)v−1.
Thus, if p is a prime which does not divide r − λ then
rankpH = v or v − 1,
and dim(C) ≤ 1.
20 / 67
![Page 71: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/71.jpg)
NoteIf H is the block by point b × v incidence matrix of a t-(v ,w , λ) designand r = λ(v − 1)/(w − 1) then
det(HT H) = rw(r − λ)v−1.
Thus, if p is a prime which does not divide r − λ then
rankpH = v or v − 1,
and dim(C) ≤ 1.
20 / 67
![Page 72: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/72.jpg)
NoteIf H is the block by point b × v incidence matrix of a t-(v ,w , λ) designand r = λ(v − 1)/(w − 1) then
det(HT H) = rw(r − λ)v−1.
Thus, if p is a prime which does not divide r − λ then
rankpH = v or v − 1,
and dim(C) ≤ 1.
20 / 67
![Page 73: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/73.jpg)
NoteIf H is the block by point b × v incidence matrix of a t-(v ,w , λ) designand r = λ(v − 1)/(w − 1) then
det(HT H) = rw(r − λ)v−1.
Thus, if p is a prime which does not divide r − λ then
rankpH = v or v − 1,
and dim(C) ≤ 1.
20 / 67
![Page 74: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/74.jpg)
NoteIf H is the block by point b × v incidence matrix of a t-(v ,w , λ) designand r = λ(v − 1)/(w − 1) then
det(HT H) = rw(r − λ)v−1.
Thus, if p is a prime which does not divide r − λ then
rankpH = v or v − 1,
and dim(C) ≤ 1.
20 / 67
![Page 75: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/75.jpg)
Designs with minimum p-rank
TaskGiven v > w > 0, λ > 0, and a prime p such that p|r − λ,find a 2-(v ,w , λ) design of minimum p-rank.
ExampleLet v = 8, w = 4, λ = 3.
Then r = 7, r − λ = 7− 3 = 4, and p = 2|(r − λ).
There exist four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe 2-(8,4,3) design of minimum 2-rank, 4, is isomorphic to thegeometric design AG2(3,2).
21 / 67
![Page 76: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/76.jpg)
Designs with minimum p-rank
TaskGiven v > w > 0, λ > 0, and a prime p such that p|r − λ,find a 2-(v ,w , λ) design of minimum p-rank.
ExampleLet v = 8, w = 4, λ = 3.
Then r = 7, r − λ = 7− 3 = 4, and p = 2|(r − λ).
There exist four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe 2-(8,4,3) design of minimum 2-rank, 4, is isomorphic to thegeometric design AG2(3,2).
21 / 67
![Page 77: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/77.jpg)
Designs with minimum p-rank
TaskGiven v > w > 0, λ > 0, and a prime p such that p|r − λ,find a 2-(v ,w , λ) design of minimum p-rank.
ExampleLet v = 8, w = 4, λ = 3.
Then r = 7, r − λ = 7− 3 = 4, and p = 2|(r − λ).
There exist four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe 2-(8,4,3) design of minimum 2-rank, 4, is isomorphic to thegeometric design AG2(3,2).
21 / 67
![Page 78: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/78.jpg)
Designs with minimum p-rank
TaskGiven v > w > 0, λ > 0, and a prime p such that p|r − λ,find a 2-(v ,w , λ) design of minimum p-rank.
ExampleLet v = 8, w = 4, λ = 3.
Then r = 7, r − λ = 7− 3 = 4, and p = 2|(r − λ).
There exist four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe 2-(8,4,3) design of minimum 2-rank, 4, is isomorphic to thegeometric design AG2(3,2).
21 / 67
![Page 79: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/79.jpg)
Designs with minimum p-rank
TaskGiven v > w > 0, λ > 0, and a prime p such that p|r − λ,find a 2-(v ,w , λ) design of minimum p-rank.
ExampleLet v = 8, w = 4, λ = 3.
Then r = 7, r − λ = 7− 3 = 4, and p = 2|(r − λ).
There exist four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe 2-(8,4,3) design of minimum 2-rank, 4, is isomorphic to thegeometric design AG2(3,2).
21 / 67
![Page 80: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/80.jpg)
Designs with minimum p-rank
TaskGiven v > w > 0, λ > 0, and a prime p such that p|r − λ,find a 2-(v ,w , λ) design of minimum p-rank.
ExampleLet v = 8, w = 4, λ = 3.
Then r = 7, r − λ = 7− 3 = 4, and p = 2|(r − λ).
There exist four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe 2-(8,4,3) design of minimum 2-rank, 4, is isomorphic to thegeometric design AG2(3,2).
21 / 67
![Page 81: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/81.jpg)
Designs with minimum p-rank
TaskGiven v > w > 0, λ > 0, and a prime p such that p|r − λ,find a 2-(v ,w , λ) design of minimum p-rank.
ExampleLet v = 8, w = 4, λ = 3.
Then r = 7, r − λ = 7− 3 = 4, and p = 2|(r − λ).
There exist four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe 2-(8,4,3) design of minimum 2-rank, 4, is isomorphic to thegeometric design AG2(3,2).
21 / 67
![Page 82: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/82.jpg)
Two fundamental questions
Given parameters v > w > 0, λ > 0, such that a 2-(v ,w , λ) designexists,
What is the minimum p-rank of a 2-(v ,w , λ) design?How many nonisomorphic 2-(v ,w , λ) designs of minimum p-rankare there?
22 / 67
![Page 83: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/83.jpg)
Two fundamental questions
Given parameters v > w > 0, λ > 0, such that a 2-(v ,w , λ) designexists,
What is the minimum p-rank of a 2-(v ,w , λ) design?How many nonisomorphic 2-(v ,w , λ) designs of minimum p-rankare there?
22 / 67
![Page 84: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/84.jpg)
Two fundamental questions
Given parameters v > w > 0, λ > 0, such that a 2-(v ,w , λ) designexists,
What is the minimum p-rank of a 2-(v ,w , λ) design?How many nonisomorphic 2-(v ,w , λ) designs of minimum p-rankare there?
22 / 67
![Page 85: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/85.jpg)
p-Ranks of Geometric Designs
The p-ranks of the geometric designs were computed in the 1960’sand 1970’s.
Theorem. (Graham and MacWilliams ’66, Weldon ’67)For any prime p ≥ 2, and any integer s ≥ 1,
rankpPG1(2,ps) =
(p + 1
2
)s
+ 1.
Theorem. (Sachar ’79)
If Π is a projective plane of prime order p (a 2-(p2 + p + 1,p + 1,1)design) then
rankp(Π) = (p2 + p + 2)/2.
23 / 67
![Page 86: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/86.jpg)
p-Ranks of Geometric Designs
The p-ranks of the geometric designs were computed in the 1960’sand 1970’s.
Theorem. (Graham and MacWilliams ’66, Weldon ’67)For any prime p ≥ 2, and any integer s ≥ 1,
rankpPG1(2,ps) =
(p + 1
2
)s
+ 1.
Theorem. (Sachar ’79)
If Π is a projective plane of prime order p (a 2-(p2 + p + 1,p + 1,1)design) then
rankp(Π) = (p2 + p + 2)/2.
23 / 67
![Page 87: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/87.jpg)
p-Ranks of Geometric Designs
The p-ranks of the geometric designs were computed in the 1960’sand 1970’s.
Theorem. (Graham and MacWilliams ’66, Weldon ’67)For any prime p ≥ 2, and any integer s ≥ 1,
rankpPG1(2,ps) =
(p + 1
2
)s
+ 1.
Theorem. (Sachar ’79)
If Π is a projective plane of prime order p (a 2-(p2 + p + 1,p + 1,1)design) then
rankp(Π) = (p2 + p + 2)/2.
23 / 67
![Page 88: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/88.jpg)
p-Ranks of Geometric Designs
The p-ranks of the geometric designs were computed in the 1960’sand 1970’s.
Theorem. (Graham and MacWilliams ’66, Weldon ’67)For any prime p ≥ 2, and any integer s ≥ 1,
rankpPG1(2,ps) =
(p + 1
2
)s
+ 1.
Theorem. (Sachar ’79)
If Π is a projective plane of prime order p (a 2-(p2 + p + 1,p + 1,1)design) then
rankp(Π) = (p2 + p + 2)/2.
23 / 67
![Page 89: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/89.jpg)
p-Ranks of Geometric Designs
The p-ranks of the geometric designs were computed in the 1960’sand 1970’s.
Theorem. (Graham and MacWilliams ’66, Weldon ’67)For any prime p ≥ 2, and any integer s ≥ 1,
rankpPG1(2,ps) =
(p + 1
2
)s
+ 1.
Theorem. (Sachar ’79)
If Π is a projective plane of prime order p (a 2-(p2 + p + 1,p + 1,1)design) then
rankp(Π) = (p2 + p + 2)/2.
23 / 67
![Page 90: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/90.jpg)
p-Ranks of Geometric Designs
The p-ranks of the geometric designs were computed in the 1960’sand 1970’s.
Theorem. (Graham and MacWilliams ’66, Weldon ’67)For any prime p ≥ 2, and any integer s ≥ 1,
rankpPG1(2,ps) =
(p + 1
2
)s
+ 1.
Theorem. (Sachar ’79)
If Π is a projective plane of prime order p (a 2-(p2 + p + 1,p + 1,1)design) then
rankp(Π) = (p2 + p + 2)/2.
23 / 67
![Page 91: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/91.jpg)
Theorem. (Smith ’67, Goethals and Delsarte ’68)
For any prime p ≥ 2, and any integer s ≥ 1,
rankpPGn−1(n,ps) =
(p + n − 1
n
)s
+ 1.
Theorem. (Graham and MacWilliams ’68)
Le D be the design of points and hyperplanes in a finite geometry ofdimension n. For any prime p ≥ 2, and any integer s ≥ 1,
rankp(D) =
(p + n − 1
n
)s
+ ε,
where ε = 1 if D = PGn−1(n,ps), and ε = 0 if D = AGn−1(n,ps).
24 / 67
![Page 92: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/92.jpg)
Theorem. (Smith ’67, Goethals and Delsarte ’68)
For any prime p ≥ 2, and any integer s ≥ 1,
rankpPGn−1(n,ps) =
(p + n − 1
n
)s
+ 1.
Theorem. (Graham and MacWilliams ’68)
Le D be the design of points and hyperplanes in a finite geometry ofdimension n. For any prime p ≥ 2, and any integer s ≥ 1,
rankp(D) =
(p + n − 1
n
)s
+ ε,
where ε = 1 if D = PGn−1(n,ps), and ε = 0 if D = AGn−1(n,ps).
24 / 67
![Page 93: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/93.jpg)
Theorem. (Smith ’67, Goethals and Delsarte ’68)
For any prime p ≥ 2, and any integer s ≥ 1,
rankpPGn−1(n,ps) =
(p + n − 1
n
)s
+ 1.
Theorem. (Graham and MacWilliams ’68)
Le D be the design of points and hyperplanes in a finite geometry ofdimension n. For any prime p ≥ 2, and any integer s ≥ 1,
rankp(D) =
(p + n − 1
n
)s
+ ε,
where ε = 1 if D = PGn−1(n,ps), and ε = 0 if D = AGn−1(n,ps).
24 / 67
![Page 94: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/94.jpg)
Theorem. (Smith ’67, Goethals and Delsarte ’68)
For any prime p ≥ 2, and any integer s ≥ 1,
rankpPGn−1(n,ps) =
(p + n − 1
n
)s
+ 1.
Theorem. (Graham and MacWilliams ’68)
Le D be the design of points and hyperplanes in a finite geometry ofdimension n. For any prime p ≥ 2, and any integer s ≥ 1,
rankp(D) =
(p + n − 1
n
)s
+ ε,
where ε = 1 if D = PGn−1(n,ps), and ε = 0 if D = AGn−1(n,ps).
24 / 67
![Page 95: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/95.jpg)
Theorem. (Smith ’67, Goethals and Delsarte ’68)
For any prime p ≥ 2, and any integer s ≥ 1,
rankpPGn−1(n,ps) =
(p + n − 1
n
)s
+ 1.
Theorem. (Graham and MacWilliams ’68)
Le D be the design of points and hyperplanes in a finite geometry ofdimension n. For any prime p ≥ 2, and any integer s ≥ 1,
rankp(D) =
(p + n − 1
n
)s
+ ε,
where ε = 1 if D = PGn−1(n,ps), and ε = 0 if D = AGn−1(n,ps).
24 / 67
![Page 96: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/96.jpg)
Theorem. (Smith ’67, Goethals and Delsarte ’68)
For any prime p ≥ 2, and any integer s ≥ 1,
rankpPGn−1(n,ps) =
(p + n − 1
n
)s
+ 1.
Theorem. (Graham and MacWilliams ’68)
Le D be the design of points and hyperplanes in a finite geometry ofdimension n. For any prime p ≥ 2, and any integer s ≥ 1,
rankp(D) =
(p + n − 1
n
)s
+ ε,
where ε = 1 if D = PGn−1(n,ps), and ε = 0 if D = AGn−1(n,ps).
24 / 67
![Page 97: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/97.jpg)
Theorem. (Smith ’67, Goethals and Delsarte ’68)
For any prime p ≥ 2, and any integer s ≥ 1,
rankpPGn−1(n,ps) =
(p + n − 1
n
)s
+ 1.
Theorem. (Graham and MacWilliams ’68)
Le D be the design of points and hyperplanes in a finite geometry ofdimension n. For any prime p ≥ 2, and any integer s ≥ 1,
rankp(D) =
(p + n − 1
n
)s
+ ε,
where ε = 1 if D = PGn−1(n,ps), and ε = 0 if D = AGn−1(n,ps).
24 / 67
![Page 98: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/98.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 99: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/99.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 100: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/100.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 101: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/101.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 102: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/102.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 103: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/103.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 104: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/104.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 105: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/105.jpg)
The general case
Theorem. (Hamada ’73)(a) The p-rank of PGd (n,ps) is given by
∑t0,...,ts
s−1∏j=0
[(tj+1p−tj )/p]∑i=0
(−1)i(
n + 1i
)(n + tj+1p − tj − ip
n
),
where (t0, . . . , ts) are integers such thatts = t0, d + 1 ≤ tj ≤ n + 1, 0 ≤ tj+1p − tj ≤ (n + 1)(p − 1),for j = 0,1, . . . , s − 1.
(b)
rankpAGd (n,ps) = rankpPGd (n,ps)− rankpPGd (n − 1,ps).
25 / 67
![Page 106: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/106.jpg)
Finite Geometry Codes
Corollary
rank2AGd (n,2) =n−d∑i=0
(ni
).
NoteThe binary code spanned by the incidence matrix of AGd (n,2) isequivalent to the Reed-Muller code of length 2n and order d .
Finite geometry codesA q-ary linear code spanned by the incidence matrix of PGd (n,q) orAGd (n,q) is a finite geometry code.
NoteThe main tool used in computing the p-ranks of geometric designs isthe theory of cyclic codes: all projective geometry codes are cyclic.
26 / 67
![Page 107: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/107.jpg)
Finite Geometry Codes
Corollary
rank2AGd (n,2) =n−d∑i=0
(ni
).
NoteThe binary code spanned by the incidence matrix of AGd (n,2) isequivalent to the Reed-Muller code of length 2n and order d .
Finite geometry codesA q-ary linear code spanned by the incidence matrix of PGd (n,q) orAGd (n,q) is a finite geometry code.
NoteThe main tool used in computing the p-ranks of geometric designs isthe theory of cyclic codes: all projective geometry codes are cyclic.
26 / 67
![Page 108: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/108.jpg)
Finite Geometry Codes
Corollary
rank2AGd (n,2) =n−d∑i=0
(ni
).
NoteThe binary code spanned by the incidence matrix of AGd (n,2) isequivalent to the Reed-Muller code of length 2n and order d .
Finite geometry codesA q-ary linear code spanned by the incidence matrix of PGd (n,q) orAGd (n,q) is a finite geometry code.
NoteThe main tool used in computing the p-ranks of geometric designs isthe theory of cyclic codes: all projective geometry codes are cyclic.
26 / 67
![Page 109: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/109.jpg)
Finite Geometry Codes
Corollary
rank2AGd (n,2) =n−d∑i=0
(ni
).
NoteThe binary code spanned by the incidence matrix of AGd (n,2) isequivalent to the Reed-Muller code of length 2n and order d .
Finite geometry codesA q-ary linear code spanned by the incidence matrix of PGd (n,q) orAGd (n,q) is a finite geometry code.
NoteThe main tool used in computing the p-ranks of geometric designs isthe theory of cyclic codes: all projective geometry codes are cyclic.
26 / 67
![Page 110: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/110.jpg)
Finite Geometry Codes
Corollary
rank2AGd (n,2) =n−d∑i=0
(ni
).
NoteThe binary code spanned by the incidence matrix of AGd (n,2) isequivalent to the Reed-Muller code of length 2n and order d .
Finite geometry codesA q-ary linear code spanned by the incidence matrix of PGd (n,q) orAGd (n,q) is a finite geometry code.
NoteThe main tool used in computing the p-ranks of geometric designs isthe theory of cyclic codes: all projective geometry codes are cyclic.
26 / 67
![Page 111: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/111.jpg)
Finite Geometry Codes
Corollary
rank2AGd (n,2) =n−d∑i=0
(ni
).
NoteThe binary code spanned by the incidence matrix of AGd (n,2) isequivalent to the Reed-Muller code of length 2n and order d .
Finite geometry codesA q-ary linear code spanned by the incidence matrix of PGd (n,q) orAGd (n,q) is a finite geometry code.
NoteThe main tool used in computing the p-ranks of geometric designs isthe theory of cyclic codes: all projective geometry codes are cyclic.
26 / 67
![Page 112: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/112.jpg)
Hamada’s Conjecture
Conjecture (Hamada, 1973) : A geometric design over Fpm hasminimum p-rank among all designs with the given parameters.
ExampleLet v = 8, w = 4, λ = 3.
There exist exactly four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe only 2-(8,4,3) design of minimum 2-rank is isomorphic to thegeometric design AG2(3,2).
27 / 67
![Page 113: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/113.jpg)
Hamada’s Conjecture
Conjecture (Hamada, 1973) : A geometric design over Fpm hasminimum p-rank among all designs with the given parameters.
ExampleLet v = 8, w = 4, λ = 3.
There exist exactly four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe only 2-(8,4,3) design of minimum 2-rank is isomorphic to thegeometric design AG2(3,2).
27 / 67
![Page 114: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/114.jpg)
Hamada’s Conjecture
Conjecture (Hamada, 1973) : A geometric design over Fpm hasminimum p-rank among all designs with the given parameters.
ExampleLet v = 8, w = 4, λ = 3.
There exist exactly four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe only 2-(8,4,3) design of minimum 2-rank is isomorphic to thegeometric design AG2(3,2).
27 / 67
![Page 115: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/115.jpg)
Hamada’s Conjecture
Conjecture (Hamada, 1973) : A geometric design over Fpm hasminimum p-rank among all designs with the given parameters.
ExampleLet v = 8, w = 4, λ = 3.
There exist exactly four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe only 2-(8,4,3) design of minimum 2-rank is isomorphic to thegeometric design AG2(3,2).
27 / 67
![Page 116: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/116.jpg)
Hamada’s Conjecture
Conjecture (Hamada, 1973) : A geometric design over Fpm hasminimum p-rank among all designs with the given parameters.
ExampleLet v = 8, w = 4, λ = 3.
There exist exactly four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe only 2-(8,4,3) design of minimum 2-rank is isomorphic to thegeometric design AG2(3,2).
27 / 67
![Page 117: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/117.jpg)
Hamada’s Conjecture
Conjecture (Hamada, 1973) : A geometric design over Fpm hasminimum p-rank among all designs with the given parameters.
ExampleLet v = 8, w = 4, λ = 3.
There exist exactly four non-isomorphic 2-(8,4,3) designs,and their 2-ranks are 4, 5, 6, and 7 respectively.
NoteThe only 2-(8,4,3) design of minimum 2-rank is isomorphic to thegeometric design AG2(3,2).
27 / 67
![Page 118: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/118.jpg)
Implications
Majority logic decodable codes: Hamada’s conjecture indicates thatgeometric designs are the best choice for the given design parameters.
NoteThe number of nonisomorphic designs having the same parameters asgeometric designs grows exponentially: Jungnickel ’84, Kantor ’94,Lam, Lam & T ’00, ’02, Jungnickel & T, ’09, Clark, Jungnickel & T, 09.
Uniqueness: The conjecture rovides a simple characterization of thegeometric designs. Finding isomorphisms is exponentially difficult,while calculating p-rank is done in polynomial time.
28 / 67
![Page 119: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/119.jpg)
Implications
Majority logic decodable codes: Hamada’s conjecture indicates thatgeometric designs are the best choice for the given design parameters.
NoteThe number of nonisomorphic designs having the same parameters asgeometric designs grows exponentially: Jungnickel ’84, Kantor ’94,Lam, Lam & T ’00, ’02, Jungnickel & T, ’09, Clark, Jungnickel & T, 09.
Uniqueness: The conjecture rovides a simple characterization of thegeometric designs. Finding isomorphisms is exponentially difficult,while calculating p-rank is done in polynomial time.
28 / 67
![Page 120: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/120.jpg)
Implications
Majority logic decodable codes: Hamada’s conjecture indicates thatgeometric designs are the best choice for the given design parameters.
NoteThe number of nonisomorphic designs having the same parameters asgeometric designs grows exponentially: Jungnickel ’84, Kantor ’94,Lam, Lam & T ’00, ’02, Jungnickel & T, ’09, Clark, Jungnickel & T, 09.
Uniqueness: The conjecture rovides a simple characterization of thegeometric designs. Finding isomorphisms is exponentially difficult,while calculating p-rank is done in polynomial time.
28 / 67
![Page 121: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/121.jpg)
Implications
Majority logic decodable codes: Hamada’s conjecture indicates thatgeometric designs are the best choice for the given design parameters.
NoteThe number of nonisomorphic designs having the same parameters asgeometric designs grows exponentially: Jungnickel ’84, Kantor ’94,Lam, Lam & T ’00, ’02, Jungnickel & T, ’09, Clark, Jungnickel & T, 09.
Uniqueness: The conjecture rovides a simple characterization of thegeometric designs. Finding isomorphisms is exponentially difficult,while calculating p-rank is done in polynomial time.
28 / 67
![Page 122: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/122.jpg)
The Proven Cases
Hamada’s Conjecture has been proved in the following cases:
Hamada and Ohmori (1975): True for PGn−1(n,2) andAGn−1(n,2).Doyen, Hubaut, Vandensavel (1978):True for PG1(n,2) and AG1(n,3).Teirlinck (1980): True for AG2(n,2).Tonchev (1999): A modified version of Hamada’s conjecture istrue for the complementary designs of PGn−1(n,q) andAGn−1(n,q) for every prime power q and every n ≥ 3,for generalized incidence matrices with entries in Fq.
In all of these cases, the geometric designs arethe unique designs of minimum p-rank.
29 / 67
![Page 123: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/123.jpg)
The Proven Cases
Hamada’s Conjecture has been proved in the following cases:
Hamada and Ohmori (1975): True for PGn−1(n,2) andAGn−1(n,2).Doyen, Hubaut, Vandensavel (1978):True for PG1(n,2) and AG1(n,3).Teirlinck (1980): True for AG2(n,2).Tonchev (1999): A modified version of Hamada’s conjecture istrue for the complementary designs of PGn−1(n,q) andAGn−1(n,q) for every prime power q and every n ≥ 3,for generalized incidence matrices with entries in Fq.
In all of these cases, the geometric designs arethe unique designs of minimum p-rank.
29 / 67
![Page 124: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/124.jpg)
The Proven Cases
Hamada’s Conjecture has been proved in the following cases:
Hamada and Ohmori (1975): True for PGn−1(n,2) andAGn−1(n,2).Doyen, Hubaut, Vandensavel (1978):True for PG1(n,2) and AG1(n,3).Teirlinck (1980): True for AG2(n,2).Tonchev (1999): A modified version of Hamada’s conjecture istrue for the complementary designs of PGn−1(n,q) andAGn−1(n,q) for every prime power q and every n ≥ 3,for generalized incidence matrices with entries in Fq.
In all of these cases, the geometric designs arethe unique designs of minimum p-rank.
29 / 67
![Page 125: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/125.jpg)
The Proven Cases
Hamada’s Conjecture has been proved in the following cases:
Hamada and Ohmori (1975): True for PGn−1(n,2) andAGn−1(n,2).Doyen, Hubaut, Vandensavel (1978):True for PG1(n,2) and AG1(n,3).Teirlinck (1980): True for AG2(n,2).Tonchev (1999): A modified version of Hamada’s conjecture istrue for the complementary designs of PGn−1(n,q) andAGn−1(n,q) for every prime power q and every n ≥ 3,for generalized incidence matrices with entries in Fq.
In all of these cases, the geometric designs arethe unique designs of minimum p-rank.
29 / 67
![Page 126: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/126.jpg)
The Proven Cases
Hamada’s Conjecture has been proved in the following cases:
Hamada and Ohmori (1975): True for PGn−1(n,2) andAGn−1(n,2).Doyen, Hubaut, Vandensavel (1978):True for PG1(n,2) and AG1(n,3).Teirlinck (1980): True for AG2(n,2).Tonchev (1999): A modified version of Hamada’s conjecture istrue for the complementary designs of PGn−1(n,q) andAGn−1(n,q) for every prime power q and every n ≥ 3,for generalized incidence matrices with entries in Fq.
In all of these cases, the geometric designs arethe unique designs of minimum p-rank.
29 / 67
![Page 127: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/127.jpg)
The Proven Cases
Hamada’s Conjecture has been proved in the following cases:
Hamada and Ohmori (1975): True for PGn−1(n,2) andAGn−1(n,2).Doyen, Hubaut, Vandensavel (1978):True for PG1(n,2) and AG1(n,3).Teirlinck (1980): True for AG2(n,2).Tonchev (1999): A modified version of Hamada’s conjecture istrue for the complementary designs of PGn−1(n,q) andAGn−1(n,q) for every prime power q and every n ≥ 3,for generalized incidence matrices with entries in Fq.
In all of these cases, the geometric designs arethe unique designs of minimum p-rank.
29 / 67
![Page 128: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/128.jpg)
The Proven Cases
Hamada’s Conjecture has been proved in the following cases:
Hamada and Ohmori (1975): True for PGn−1(n,2) andAGn−1(n,2).Doyen, Hubaut, Vandensavel (1978):True for PG1(n,2) and AG1(n,3).Teirlinck (1980): True for AG2(n,2).Tonchev (1999): A modified version of Hamada’s conjecture istrue for the complementary designs of PGn−1(n,q) andAGn−1(n,q) for every prime power q and every n ≥ 3,for generalized incidence matrices with entries in Fq.
In all of these cases, the geometric designs arethe unique designs of minimum p-rank.
29 / 67
![Page 129: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/129.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 130: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/130.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 131: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/131.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 132: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/132.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 133: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/133.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 134: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/134.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 135: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/135.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 136: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/136.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 137: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/137.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 138: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/138.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 139: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/139.jpg)
Proven cases
Theorem. (Hamada and Ohmori ’75)
(i) If A is the incidence matrix of a 2-(2n+1 − 1,2n,2n−1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to the complementarydesign of PGn−1(n,2).(ii) If A is the incidence matrix of a 2-(2n+1 − 1,2n − 1,2n−1 − 1) designD then
rank2(A) ≥ n + 2,
with equality if and only if D is isomorphic to PGn−1(n,2).
(iii) If A is the incidence matrix of a 2-(2n,2n−1,2n−1 − 1) design D then
rank2(A) ≥ n + 1,
with equality if and only if D is isomorphic to AGn−1(n,2).30 / 67
![Page 140: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/140.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 141: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/141.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 142: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/142.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 143: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/143.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 144: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/144.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 145: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/145.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 146: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/146.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 147: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/147.jpg)
The results of Doyen, Hubaut and Vandensavel
Theorem. (Doyen, Hubaut and Vandensavel ’78)
(i) The 2-rank of the incidence matrix A of any 2-(2n+1 − 1,3,1) designD satisfies
rank2(A) ≥ 2n+1 − n − 2,
with equality if and only if D is isomorphic to PG1(n,2).
(ii) The 3-rank of the incidence matrix A of any 2-(3n,3,1) design Dsatisfies
rank3(A) ≥ 3n − 1− n,
with equality if and only if D is isomorphic to AG1(n,3).
31 / 67
![Page 148: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/148.jpg)
A result of Teirlinck
Theorem. (Teirlinck ’80)The 2-rank of the incidence matrix A of a 3-(2n,4,1) design D satisfies
rank2(A) ≥ 2n − 1− n,
with equality if and only if D is isomorphic to AG2(n,2).
NoteThe result of Teirlinck and the binary case of Doyen, Hubaut andVandelnsavel’s result are "dual" to the result of Hamada and Ohmori.
In terms of related codes, these results are equivalent to theuniqueness of the first order Reed-Muller code.
32 / 67
![Page 149: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/149.jpg)
A result of Teirlinck
Theorem. (Teirlinck ’80)The 2-rank of the incidence matrix A of a 3-(2n,4,1) design D satisfies
rank2(A) ≥ 2n − 1− n,
with equality if and only if D is isomorphic to AG2(n,2).
NoteThe result of Teirlinck and the binary case of Doyen, Hubaut andVandelnsavel’s result are "dual" to the result of Hamada and Ohmori.
In terms of related codes, these results are equivalent to theuniqueness of the first order Reed-Muller code.
32 / 67
![Page 150: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/150.jpg)
A result of Teirlinck
Theorem. (Teirlinck ’80)The 2-rank of the incidence matrix A of a 3-(2n,4,1) design D satisfies
rank2(A) ≥ 2n − 1− n,
with equality if and only if D is isomorphic to AG2(n,2).
NoteThe result of Teirlinck and the binary case of Doyen, Hubaut andVandelnsavel’s result are "dual" to the result of Hamada and Ohmori.
In terms of related codes, these results are equivalent to theuniqueness of the first order Reed-Muller code.
32 / 67
![Page 151: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/151.jpg)
A result of Teirlinck
Theorem. (Teirlinck ’80)The 2-rank of the incidence matrix A of a 3-(2n,4,1) design D satisfies
rank2(A) ≥ 2n − 1− n,
with equality if and only if D is isomorphic to AG2(n,2).
NoteThe result of Teirlinck and the binary case of Doyen, Hubaut andVandelnsavel’s result are "dual" to the result of Hamada and Ohmori.
In terms of related codes, these results are equivalent to theuniqueness of the first order Reed-Muller code.
32 / 67
![Page 152: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/152.jpg)
A result of Teirlinck
Theorem. (Teirlinck ’80)The 2-rank of the incidence matrix A of a 3-(2n,4,1) design D satisfies
rank2(A) ≥ 2n − 1− n,
with equality if and only if D is isomorphic to AG2(n,2).
NoteThe result of Teirlinck and the binary case of Doyen, Hubaut andVandelnsavel’s result are "dual" to the result of Hamada and Ohmori.
In terms of related codes, these results are equivalent to theuniqueness of the first order Reed-Muller code.
32 / 67
![Page 153: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/153.jpg)
A result of Teirlinck
Theorem. (Teirlinck ’80)The 2-rank of the incidence matrix A of a 3-(2n,4,1) design D satisfies
rank2(A) ≥ 2n − 1− n,
with equality if and only if D is isomorphic to AG2(n,2).
NoteThe result of Teirlinck and the binary case of Doyen, Hubaut andVandelnsavel’s result are "dual" to the result of Hamada and Ohmori.
In terms of related codes, these results are equivalent to theuniqueness of the first order Reed-Muller code.
32 / 67
![Page 154: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/154.jpg)
A result of Teirlinck
Theorem. (Teirlinck ’80)The 2-rank of the incidence matrix A of a 3-(2n,4,1) design D satisfies
rank2(A) ≥ 2n − 1− n,
with equality if and only if D is isomorphic to AG2(n,2).
NoteThe result of Teirlinck and the binary case of Doyen, Hubaut andVandelnsavel’s result are "dual" to the result of Hamada and Ohmori.
In terms of related codes, these results are equivalent to theuniqueness of the first order Reed-Muller code.
32 / 67
![Page 155: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/155.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 156: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/156.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 157: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/157.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 158: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/158.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 159: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/159.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 160: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/160.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 161: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/161.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 162: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/162.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 163: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/163.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 164: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/164.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 165: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/165.jpg)
A revised version of Hamada’s Conjecture
Generalized Incidence Matrix (T. ’99)A generalized incidence matrix of a design has entries in GF (q),with nonzero entries designating incidence.
DefinitionThe dimension of a design D over GF (q), (dimq(D)), is defined as theminimum q-rank of all generalized incidence matrices of D over GF (q).
ExampleThe 3-rank of the (0,1)-incidence matrix of the unique 5-(12,6,1)design D12 is 11, while dim3(D12) = 6.
33 / 67
![Page 166: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/166.jpg)
A revised version of Hamada’s Conjecture
ConjectureHamada’s conjecture is true if ordinary incidence matrices arereplaced by generalized incidence matrices over the related finitefield.
34 / 67
![Page 167: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/167.jpg)
A revised version of Hamada’s Conjecture
ConjectureHamada’s conjecture is true if ordinary incidence matrices arereplaced by generalized incidence matrices over the related finitefield.
34 / 67
![Page 168: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/168.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 169: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/169.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 170: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/170.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 171: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/171.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 172: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/172.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 173: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/173.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 174: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/174.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 175: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/175.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 176: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/176.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 177: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/177.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 178: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/178.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 179: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/179.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 180: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/180.jpg)
A q-analogue of Hamada and Ohmori’s theorem
Theorem. (T ’99)Let q be an arbitrary prime power, and let n ≥ 2.(i) Let D be a 2-((qn+1 − 1)/(q − 1),qn,qn − qn−1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of PGn−1(n,q).(ii) Let D be a 2-(qn,qn − qn−1,qn − qn−1 − 1) design. Then
dimq(D) ≥ n + 1.
The equality dimq(D) = n + 1 holds if and only if D is isomorphic to thecomplementary design of AGn−1(n,q).
35 / 67
![Page 181: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/181.jpg)
ExampleLet D be a 2-(121,100,99) design. Then
dim11(D) ≥ 3,
with equality dim11(D) = 3 if and only if D is isomorphic to thecomplementary design of the Desarguesian affine plane of order 11.
36 / 67
![Page 182: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/182.jpg)
ExampleLet D be a 2-(121,100,99) design. Then
dim11(D) ≥ 3,
with equality dim11(D) = 3 if and only if D is isomorphic to thecomplementary design of the Desarguesian affine plane of order 11.
36 / 67
![Page 183: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/183.jpg)
ExampleLet D be a 2-(121,100,99) design. Then
dim11(D) ≥ 3,
with equality dim11(D) = 3 if and only if D is isomorphic to thecomplementary design of the Desarguesian affine plane of order 11.
36 / 67
![Page 184: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/184.jpg)
ExampleLet D be a 2-(121,100,99) design. Then
dim11(D) ≥ 3,
with equality dim11(D) = 3 if and only if D is isomorphic to thecomplementary design of the Desarguesian affine plane of order 11.
36 / 67
![Page 185: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/185.jpg)
ExampleLet D be a 2-(121,100,99) design. Then
dim11(D) ≥ 3,
with equality dim11(D) = 3 if and only if D is isomorphic to thecomplementary design of the Desarguesian affine plane of order 11.
36 / 67
![Page 186: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/186.jpg)
ExampleLet D be a 2-(121,100,99) design. Then
dim11(D) ≥ 3,
with equality dim11(D) = 3 if and only if D is isomorphic to thecomplementary design of the Desarguesian affine plane of order 11.
36 / 67
![Page 187: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/187.jpg)
ExampleLet D be a 2-(121,100,99) design. Then
dim11(D) ≥ 3,
with equality dim11(D) = 3 if and only if D is isomorphic to thecomplementary design of the Desarguesian affine plane of order 11.
36 / 67
![Page 188: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/188.jpg)
Are geometric designs characterized by theirp-rank?
QuestionAre geometric designs characterized as the unique designs with thegiven parameters and p-rank?
AnswerYes, in all proved cases of Hamada’s Conjecture.
37 / 67
![Page 189: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/189.jpg)
Are geometric designs characterized by theirp-rank?
QuestionAre geometric designs characterized as the unique designs with thegiven parameters and p-rank?
AnswerYes, in all proved cases of Hamada’s Conjecture.
37 / 67
![Page 190: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/190.jpg)
Are geometric designs characterized by theirp-rank?
QuestionAre geometric designs characterized as the unique designs with thegiven parameters and p-rank?
AnswerYes, in all proved cases of Hamada’s Conjecture.
37 / 67
![Page 191: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/191.jpg)
Are geometric designs characterized by theirp-rank?
QuestionAre geometric designs characterized as the unique designs with thegiven parameters and p-rank?
AnswerYes, in all proved cases of Hamada’s Conjecture.
37 / 67
![Page 192: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/192.jpg)
Non-geometric designs with the same p-rank asgeometric ones
There are known non-geometric designs having the sameparameters and the same p-rank as certain geometric designs:
Untill recently, all known such designs were
2− (31,7,7),3− (32,8,7), (p = 2); 2− (64,16,5), (q = 22).
In 2008 and 2009, infinitely many designs were found with forarbitrary prime p ≥ 2.
These designs indicate that although geometric designs may haveminimum p-rank, they are not always the unique designs of minimump-rank.
38 / 67
![Page 193: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/193.jpg)
Non-geometric designs with the same p-rank asgeometric ones
There are known non-geometric designs having the sameparameters and the same p-rank as certain geometric designs:
Untill recently, all known such designs were
2− (31,7,7),3− (32,8,7), (p = 2); 2− (64,16,5), (q = 22).
In 2008 and 2009, infinitely many designs were found with forarbitrary prime p ≥ 2.
These designs indicate that although geometric designs may haveminimum p-rank, they are not always the unique designs of minimump-rank.
38 / 67
![Page 194: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/194.jpg)
Non-geometric designs with the same p-rank asgeometric ones
There are known non-geometric designs having the sameparameters and the same p-rank as certain geometric designs:
Untill recently, all known such designs were
2− (31,7,7),3− (32,8,7), (p = 2); 2− (64,16,5), (q = 22).
In 2008 and 2009, infinitely many designs were found with forarbitrary prime p ≥ 2.
These designs indicate that although geometric designs may haveminimum p-rank, they are not always the unique designs of minimump-rank.
38 / 67
![Page 195: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/195.jpg)
Non-geometric designs with the same p-rank asgeometric ones
There are known non-geometric designs having the sameparameters and the same p-rank as certain geometric designs:
Untill recently, all known such designs were
2− (31,7,7),3− (32,8,7), (p = 2); 2− (64,16,5), (q = 22).
In 2008 and 2009, infinitely many designs were found with forarbitrary prime p ≥ 2.
These designs indicate that although geometric designs may haveminimum p-rank, they are not always the unique designs of minimump-rank.
38 / 67
![Page 196: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/196.jpg)
Non-geometric designs with the same p-rank asgeometric ones
There are known non-geometric designs having the sameparameters and the same p-rank as certain geometric designs:
Untill recently, all known such designs were
2− (31,7,7),3− (32,8,7), (p = 2); 2− (64,16,5), (q = 22).
In 2008 and 2009, infinitely many designs were found with forarbitrary prime p ≥ 2.
These designs indicate that although geometric designs may haveminimum p-rank, they are not always the unique designs of minimump-rank.
38 / 67
![Page 197: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/197.jpg)
Non-geometric designs with the same p-rank asgeometric ones
There are known non-geometric designs having the sameparameters and the same p-rank as certain geometric designs:
Untill recently, all known such designs were
2− (31,7,7),3− (32,8,7), (p = 2); 2− (64,16,5), (q = 22).
In 2008 and 2009, infinitely many designs were found with forarbitrary prime p ≥ 2.
These designs indicate that although geometric designs may haveminimum p-rank, they are not always the unique designs of minimump-rank.
38 / 67
![Page 198: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/198.jpg)
Non-geometric designs with the same p-rank asgeometric ones
There are known non-geometric designs having the sameparameters and the same p-rank as certain geometric designs:
Untill recently, all known such designs were
2− (31,7,7),3− (32,8,7), (p = 2); 2− (64,16,5), (q = 22).
In 2008 and 2009, infinitely many designs were found with forarbitrary prime p ≥ 2.
These designs indicate that although geometric designs may haveminimum p-rank, they are not always the unique designs of minimump-rank.
38 / 67
![Page 199: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/199.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 200: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/200.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 201: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/201.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 202: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/202.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 203: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/203.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 204: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/204.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 205: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/205.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 206: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/206.jpg)
Non-geometric designs of minimum p-rank
Deigns from self-dual codesTheorem (T ’86).(i) In addition to PG2(4,2), there are four non-geometric 2-(31,7,7)designs with block intersection numbers {1,3}, all having 2-rank 16.(ii) In addition to AG3(5,2), there are four non-geometric 3-(32,8,7)designs with even block intersection numbers, all of 2-rank 16.
ProofUse Rudolph’s theorem, the Assmus-Mattson theorem, and theclassification of binary self-dual [32,16,8] codes.
Quasi-symmetric designA design with two distinct block intesection numbers.
NoteTwo 2-(31,7,7) designs supported by the projective geometry codeand the QR code were mentioned by Goethals and Delsarte in 1968.39 / 67
![Page 207: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/207.jpg)
Designs from Nets
Symmetric (µ,m)-Nets
A symmetric (µ,m)-net is a 1-(m2µ,mµ,mµ) design Dsuch that both D and its dual design D∗ areuniquely resolvable ito parallel classes of size m, so that anynon-parallel blocks share exactly µ points .
Class-regular netsA symmetric (µ,m)-net is class-regular if it admits an automorphismgroup of order m that acts transitively on each block and point parallelclass.
The classical (q,q)-net
Points and planes of AG(3,q) that do not contain lines from a givenparallel class.
40 / 67
![Page 208: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/208.jpg)
Designs from Nets
Symmetric (µ,m)-Nets
A symmetric (µ,m)-net is a 1-(m2µ,mµ,mµ) design Dsuch that both D and its dual design D∗ areuniquely resolvable ito parallel classes of size m, so that anynon-parallel blocks share exactly µ points .
Class-regular netsA symmetric (µ,m)-net is class-regular if it admits an automorphismgroup of order m that acts transitively on each block and point parallelclass.
The classical (q,q)-net
Points and planes of AG(3,q) that do not contain lines from a givenparallel class.
40 / 67
![Page 209: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/209.jpg)
Designs from Nets
Symmetric (µ,m)-Nets
A symmetric (µ,m)-net is a 1-(m2µ,mµ,mµ) design Dsuch that both D and its dual design D∗ areuniquely resolvable ito parallel classes of size m, so that anynon-parallel blocks share exactly µ points .
Class-regular netsA symmetric (µ,m)-net is class-regular if it admits an automorphismgroup of order m that acts transitively on each block and point parallelclass.
The classical (q,q)-net
Points and planes of AG(3,q) that do not contain lines from a givenparallel class.
40 / 67
![Page 210: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/210.jpg)
Designs from Nets
Symmetric (µ,m)-Nets
A symmetric (µ,m)-net is a 1-(m2µ,mµ,mµ) design Dsuch that both D and its dual design D∗ areuniquely resolvable ito parallel classes of size m, so that anynon-parallel blocks share exactly µ points .
Class-regular netsA symmetric (µ,m)-net is class-regular if it admits an automorphismgroup of order m that acts transitively on each block and point parallelclass.
The classical (q,q)-net
Points and planes of AG(3,q) that do not contain lines from a givenparallel class.
40 / 67
![Page 211: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/211.jpg)
Designs from Nets
Symmetric (µ,m)-Nets
A symmetric (µ,m)-net is a 1-(m2µ,mµ,mµ) design Dsuch that both D and its dual design D∗ areuniquely resolvable ito parallel classes of size m, so that anynon-parallel blocks share exactly µ points .
Class-regular netsA symmetric (µ,m)-net is class-regular if it admits an automorphismgroup of order m that acts transitively on each block and point parallelclass.
The classical (q,q)-net
Points and planes of AG(3,q) that do not contain lines from a givenparallel class.
40 / 67
![Page 212: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/212.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 213: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/213.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 214: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/214.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 215: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/215.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 216: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/216.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 217: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/217.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 218: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/218.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 219: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/219.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 220: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/220.jpg)
The class-regular (4, 4) nets and their codes
The (4,4) nets
A (4,4)-net consists of 64 points and 64 blocks, each block of size 16and each point in 16 blocks, so that the blocks (as well as and points)are partitioned into 16 parallel classes of size 4, and any twonon-parallel blocks share 4 points.
Theorem. (Harada, Lam & T., 2005)(i) Up to isomorphism, there are exactly 239 class-regular (4,4)-nets.(ii) The minimum 2-rank of a (4,4)-net is 16.(iii) The binary codes of three (4,4)-nets support 2-(64,16,5) designsof 2-rank 16:
The code of the classical (4,4)-net supports AG2(3,22).Two other nets support non-geometric 2-(64,16,5) designshaving the same 2-rank as AG2(3,4).
41 / 67
![Page 221: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/221.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 222: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/222.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 223: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/223.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 224: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/224.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 225: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/225.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 226: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/226.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 227: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/227.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 228: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/228.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 229: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/229.jpg)
Non-geometric designs from line spreads
Theorem. (Mavron, McDonough, & T., 2008)One of the non-geometric 2-(64,16,5) designs of 2-rank 16 found byHarada, Lam and T., can be btained from a (very special) line spreadof PG(5,2).
Theorem. (Mateva and Topalova, 2008)There are 131,044 inequivalent line spreads of PG(5,2).Two of these line spreads yield 2-(64,16,5) designs of 2-rank 16:AG2(3,22), and the non-geometric design found by Harada, Lam,T., and Mavron, McDonough and T.
42 / 67
![Page 230: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/230.jpg)
Designs from Polarities in PG(n, q)
The motivating exampleThe geometric design PG2(4,2) and one of the non-geometric2-(31,7,7) designs of 2-rank 16 share the following structure:
2− (15,7,3) 2− (15,3,1)× 4Planes ∈ PG(3,2) Lines ∈ PG(3,2)
3− (16,4,1)∅ Planes ∈ AG(4,2)
43 / 67
![Page 231: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/231.jpg)
Designs from Polarities in PG(n, q)
The motivating exampleThe geometric design PG2(4,2) and one of the non-geometric2-(31,7,7) designs of 2-rank 16 share the following structure:
2− (15,7,3) 2− (15,3,1)× 4Planes ∈ PG(3,2) Lines ∈ PG(3,2)
3− (16,4,1)∅ Planes ∈ AG(4,2)
43 / 67
![Page 232: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/232.jpg)
Designs from Polarities in PG(n, q)
The motivating exampleThe geometric design PG2(4,2) and one of the non-geometric2-(31,7,7) designs of 2-rank 16 share the following structure:
2− (15,7,3) 2− (15,3,1)× 4Planes ∈ PG(3,2) Lines ∈ PG(3,2)
3− (16,4,1)∅ Planes ∈ AG(4,2)
43 / 67
![Page 233: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/233.jpg)
Designs from Polarities in PG(n, q)
The motivating exampleThe geometric design PG2(4,2) and one of the non-geometric2-(31,7,7) designs of 2-rank 16 share the following structure:
2− (15,7,3) 2− (15,3,1)× 4Planes ∈ PG(3,2) Lines ∈ PG(3,2)
3− (16,4,1)∅ Planes ∈ AG(4,2)
43 / 67
![Page 234: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/234.jpg)
Designs from Polarities in PG(n, q)
The motivating exampleThe geometric design PG2(4,2) and one of the non-geometric2-(31,7,7) designs of 2-rank 16 share the following structure:
2− (15,7,3) 2− (15,3,1)× 4Planes ∈ PG(3,2) Lines ∈ PG(3,2)
3− (16,4,1)∅ Planes ∈ AG(4,2)
43 / 67
![Page 235: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/235.jpg)
Polarities in PG(n, q)
A polarity α of PG(n,q) is an involutory isomorphism betweenPG(n,q) and its dual space:
α : point ←→ hyperplane,. . .
i−subspace ←→ (n-1-i)−subspace. . .
ExampleThe null polarity:
point ←→ hyperplane(a0, . . . ,an) ←→ a0x0 + · · ·+ anxn = 0.
44 / 67
![Page 236: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/236.jpg)
Polarities in PG(n, q)
A polarity α of PG(n,q) is an involutory isomorphism betweenPG(n,q) and its dual space:
α : point ←→ hyperplane,. . .
i−subspace ←→ (n-1-i)−subspace. . .
ExampleThe null polarity:
point ←→ hyperplane(a0, . . . ,an) ←→ a0x0 + · · ·+ anxn = 0.
44 / 67
![Page 237: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/237.jpg)
Polarities in PG(n, q)
A polarity α of PG(n,q) is an involutory isomorphism betweenPG(n,q) and its dual space:
α : point ←→ hyperplane,. . .
i−subspace ←→ (n-1-i)−subspace. . .
ExampleThe null polarity:
point ←→ hyperplane(a0, . . . ,an) ←→ a0x0 + · · ·+ anxn = 0.
44 / 67
![Page 238: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/238.jpg)
Polarities in PG(n, q)
A polarity α of PG(n,q) is an involutory isomorphism betweenPG(n,q) and its dual space:
α : point ←→ hyperplane,. . .
i−subspace ←→ (n-1-i)−subspace. . .
ExampleThe null polarity:
point ←→ hyperplane(a0, . . . ,an) ←→ a0x0 + · · ·+ anxn = 0.
44 / 67
![Page 239: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/239.jpg)
Polarities in PG(n, q)
A polarity α of PG(n,q) is an involutory isomorphism betweenPG(n,q) and its dual space:
α : point ←→ hyperplane,. . .
i−subspace ←→ (n-1-i)−subspace. . .
ExampleThe null polarity:
point ←→ hyperplane(a0, . . . ,an) ←→ a0x0 + · · ·+ anxn = 0.
44 / 67
![Page 240: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/240.jpg)
Polarities in PG(n, q)
A polarity α of PG(n,q) is an involutory isomorphism betweenPG(n,q) and its dual space:
α : point ←→ hyperplane,. . .
i−subspace ←→ (n-1-i)−subspace. . .
ExampleThe null polarity:
point ←→ hyperplane(a0, . . . ,an) ←→ a0x0 + · · ·+ anxn = 0.
44 / 67
![Page 241: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/241.jpg)
A generalization from PG(4, 2) to PG(4, q)
Let α be a polarity of PG(3,q):
α : point←→ plane; line←→ line
PG2(4,q)
PG2(3,q) PG1(3,q)Planes Lines
AG2(4,q)∅ Planes
45 / 67
![Page 242: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/242.jpg)
A generalization from PG(4, 2) to PG(4, q)
Let α be a polarity of PG(3,q):
α : point←→ plane; line←→ line
PG2(4,q)
PG2(3,q) PG1(3,q)Planes Lines
AG2(4,q)∅ Planes
45 / 67
![Page 243: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/243.jpg)
A generalization from PG(4, 2) to PG(4, q)
Let α be a polarity of PG(3,q):
α : point←→ plane; line←→ line
PG2(4,q)
PG2(3,q) PG1(3,q)Planes Lines
AG2(4,q)∅ Planes
45 / 67
![Page 244: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/244.jpg)
A generalization from PG(4, 2) to PG(4, q)
Let α be a polarity of PG(3,q):
α : point←→ plane; line←→ line
PG2(4,q)
PG2(3,q) PG1(3,q)Planes Lines
AG2(4,q)∅ Planes
45 / 67
![Page 245: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/245.jpg)
A new class of quasi-symmetric designs frompolaities in PG(4, q)
Theorem. (Jungnickel & T., 2008)Permuting the lines of a hyperplane H = PG(3,q) ⊂ PG(4,q) via apolarity α of H transforms PG2(4,q) into another non-geometricquasi-symmetric design with intersection numbers {1, q + 1}.
NoteLines of PG(4,q) which meet H = PG(3,q) in one point aretransformed by α into "lines" of size 2.
46 / 67
![Page 246: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/246.jpg)
A new class of quasi-symmetric designs frompolaities in PG(4, q)
Theorem. (Jungnickel & T., 2008)Permuting the lines of a hyperplane H = PG(3,q) ⊂ PG(4,q) via apolarity α of H transforms PG2(4,q) into another non-geometricquasi-symmetric design with intersection numbers {1, q + 1}.
NoteLines of PG(4,q) which meet H = PG(3,q) in one point aretransformed by α into "lines" of size 2.
46 / 67
![Page 247: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/247.jpg)
A new class of quasi-symmetric designs frompolaities in PG(4, q)
Theorem. (Jungnickel & T., 2008)Permuting the lines of a hyperplane H = PG(3,q) ⊂ PG(4,q) via apolarity α of H transforms PG2(4,q) into another non-geometricquasi-symmetric design with intersection numbers {1, q + 1}.
NoteLines of PG(4,q) which meet H = PG(3,q) in one point aretransformed by α into "lines" of size 2.
46 / 67
![Page 248: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/248.jpg)
A new class of quasi-symmetric designs frompolaities in PG(4, q)
Theorem. (Jungnickel & T., 2008)Permuting the lines of a hyperplane H = PG(3,q) ⊂ PG(4,q) via apolarity α of H transforms PG2(4,q) into another non-geometricquasi-symmetric design with intersection numbers {1, q + 1}.
NoteLines of PG(4,q) which meet H = PG(3,q) in one point aretransformed by α into "lines" of size 2.
46 / 67
![Page 249: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/249.jpg)
A generalization to PG(2k , q)
PGk (2k ,q)
PGk (2k − 1,q) PGk−1(2k − 1,q)
∅ AGk (2k ,q)
NoteAny polarity α of PG(2k − 1,q) maps any (k − 1)-subspace to a(k − 1)-subspace.
Theorem. (Jungnickel & T., 2008)Permuting the (k − 1)-subspaces of a hyperplaneH = PG(2k − 1,q) ⊂ PG(2k ,q) via a polarity α transformsD = PGk (2k ,q) to a non-geometric design α(D) having the sameparameters and the same block intersection numbers as PGk (2k ,q).
47 / 67
![Page 250: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/250.jpg)
A generalization to PG(2k , q)
PGk (2k ,q)
PGk (2k − 1,q) PGk−1(2k − 1,q)
∅ AGk (2k ,q)
NoteAny polarity α of PG(2k − 1,q) maps any (k − 1)-subspace to a(k − 1)-subspace.
Theorem. (Jungnickel & T., 2008)Permuting the (k − 1)-subspaces of a hyperplaneH = PG(2k − 1,q) ⊂ PG(2k ,q) via a polarity α transformsD = PGk (2k ,q) to a non-geometric design α(D) having the sameparameters and the same block intersection numbers as PGk (2k ,q).
47 / 67
![Page 251: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/251.jpg)
A generalization to PG(2k , q)
PGk (2k ,q)
PGk (2k − 1,q) PGk−1(2k − 1,q)
∅ AGk (2k ,q)
NoteAny polarity α of PG(2k − 1,q) maps any (k − 1)-subspace to a(k − 1)-subspace.
Theorem. (Jungnickel & T., 2008)Permuting the (k − 1)-subspaces of a hyperplaneH = PG(2k − 1,q) ⊂ PG(2k ,q) via a polarity α transformsD = PGk (2k ,q) to a non-geometric design α(D) having the sameparameters and the same block intersection numbers as PGk (2k ,q).
47 / 67
![Page 252: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/252.jpg)
A generalization to PG(2k , q)
PGk (2k ,q)
PGk (2k − 1,q) PGk−1(2k − 1,q)
∅ AGk (2k ,q)
NoteAny polarity α of PG(2k − 1,q) maps any (k − 1)-subspace to a(k − 1)-subspace.
Theorem. (Jungnickel & T., 2008)Permuting the (k − 1)-subspaces of a hyperplaneH = PG(2k − 1,q) ⊂ PG(2k ,q) via a polarity α transformsD = PGk (2k ,q) to a non-geometric design α(D) having the sameparameters and the same block intersection numbers as PGk (2k ,q).
47 / 67
![Page 253: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/253.jpg)
A generalization to PG(2k , q)
PGk (2k ,q)
PGk (2k − 1,q) PGk−1(2k − 1,q)
∅ AGk (2k ,q)
NoteAny polarity α of PG(2k − 1,q) maps any (k − 1)-subspace to a(k − 1)-subspace.
Theorem. (Jungnickel & T., 2008)Permuting the (k − 1)-subspaces of a hyperplaneH = PG(2k − 1,q) ⊂ PG(2k ,q) via a polarity α transformsD = PGk (2k ,q) to a non-geometric design α(D) having the sameparameters and the same block intersection numbers as PGk (2k ,q).
47 / 67
![Page 254: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/254.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 255: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/255.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 256: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/256.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 257: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/257.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 258: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/258.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 259: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/259.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 260: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/260.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 261: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/261.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 262: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/262.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 263: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/263.jpg)
The p-rank of a design obtained via polarity
Theorem. (Jungnickel & T., 2008)Let α be a polarity of PG(2k − 1,q), where q = ps and p is a prime,and let α(D) be the design obtained from PGk (2k ,q). Then
rankpPGk (2k ,q) ≤ rankpα(D) ≤ 12
(q2k+1 − 1q − 1
+ 1).
If q = p is a prime then
rankpPGk (2k ,q) = rankpα(D).
An example of a non-prime q
If q = 4 = 22 and k = 2, we have
rank2PG2(4,4) = 146 < 154 = rank2α(D) <45 − 14− 1
= 171.
48 / 67
![Page 264: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/264.jpg)
The proof
Hamada’s formula for rp = rankpPGk (2k ,p), p prime,as simplified by Hirschfeld and Shaw ’94, is
rp =p2k+1 − 1
p − 1−
k−1∑i=0
(−1)i(
(k − i)(p − 1)− 1i
)(k + (k − i)p
2k − i
). (2)
What we need is
rp =12
(p2k+1 − 1p − 1
+ 1). (3)
49 / 67
![Page 265: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/265.jpg)
The proof
Hamada’s formula for rp = rankpPGk (2k ,p), p prime,as simplified by Hirschfeld and Shaw ’94, is
rp =p2k+1 − 1
p − 1−
k−1∑i=0
(−1)i(
(k − i)(p − 1)− 1i
)(k + (k − i)p
2k − i
). (2)
What we need is
rp =12
(p2k+1 − 1p − 1
+ 1). (3)
49 / 67
![Page 266: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/266.jpg)
The proof
Hamada’s formula for rp = rankpPGk (2k ,p), p prime,as simplified by Hirschfeld and Shaw ’94, is
rp =p2k+1 − 1
p − 1−
k−1∑i=0
(−1)i(
(k − i)(p − 1)− 1i
)(k + (k − i)p
2k − i
). (2)
What we need is
rp =12
(p2k+1 − 1p − 1
+ 1). (3)
49 / 67
![Page 267: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/267.jpg)
The proof
Hamada’s formula for rp = rankpPGk (2k ,p), p prime,as simplified by Hirschfeld and Shaw ’94, is
rp =p2k+1 − 1
p − 1−
k−1∑i=0
(−1)i(
(k − i)(p − 1)− 1i
)(k + (k − i)p
2k − i
). (2)
What we need is
rp =12
(p2k+1 − 1p − 1
+ 1). (3)
49 / 67
![Page 268: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/268.jpg)
The proof
Hamada’s formula for rp = rankpPGk (2k ,p), p prime,as simplified by Hirschfeld and Shaw ’94, is
rp =p2k+1 − 1
p − 1−
k−1∑i=0
(−1)i(
(k − i)(p − 1)− 1i
)(k + (k − i)p
2k − i
). (2)
What we need is
rp =12
(p2k+1 − 1p − 1
+ 1). (3)
49 / 67
![Page 269: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/269.jpg)
The proof
Hamada’s formula for rp = rankpPGk (2k ,p), p prime,as simplified by Hirschfeld and Shaw ’94, is
rp =p2k+1 − 1
p − 1−
k−1∑i=0
(−1)i(
(k − i)(p − 1)− 1i
)(k + (k − i)p
2k − i
). (2)
What we need is
rp =12
(p2k+1 − 1p − 1
+ 1). (3)
49 / 67
![Page 270: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/270.jpg)
Proof
Claim.The expressions (2) and (3) are equal:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
(4)
A proof by inductionThe identity (4) can be proved by induction, using a recursive formulafor the dimension of the geometric code defined by PGk (2k ,p).
50 / 67
![Page 271: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/271.jpg)
Proof
Claim.The expressions (2) and (3) are equal:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
(4)
A proof by inductionThe identity (4) can be proved by induction, using a recursive formulafor the dimension of the geometric code defined by PGk (2k ,p).
50 / 67
![Page 272: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/272.jpg)
Proof
Claim.The expressions (2) and (3) are equal:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
(4)
A proof by inductionThe identity (4) can be proved by induction, using a recursive formulafor the dimension of the geometric code defined by PGk (2k ,p).
50 / 67
![Page 273: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/273.jpg)
Proof
Claim.The expressions (2) and (3) are equal:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
(4)
A proof by inductionThe identity (4) can be proved by induction, using a recursive formulafor the dimension of the geometric code defined by PGk (2k ,p).
50 / 67
![Page 274: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/274.jpg)
A combinatorial proof of the identity (4)
TheoremThe following identity holds for any positive integer p:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
For a proof, see
J. L. W. V. Jensen: Sur une identité d’Abel et sur d’autres formulesanalogues, Acta Math. 26 (1902), 307-318.
or
M. E. Larsen: Summa Summarum, CMS Treatises in Mathematics,Canadian Mathematical Society, Ottawa, ON; A K Peters, Ltd.,Wellesley, MA (2007).
51 / 67
![Page 275: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/275.jpg)
A combinatorial proof of the identity (4)
TheoremThe following identity holds for any positive integer p:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
For a proof, see
J. L. W. V. Jensen: Sur une identité d’Abel et sur d’autres formulesanalogues, Acta Math. 26 (1902), 307-318.
or
M. E. Larsen: Summa Summarum, CMS Treatises in Mathematics,Canadian Mathematical Society, Ottawa, ON; A K Peters, Ltd.,Wellesley, MA (2007).
51 / 67
![Page 276: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/276.jpg)
A combinatorial proof of the identity (4)
TheoremThe following identity holds for any positive integer p:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
For a proof, see
J. L. W. V. Jensen: Sur une identité d’Abel et sur d’autres formulesanalogues, Acta Math. 26 (1902), 307-318.
or
M. E. Larsen: Summa Summarum, CMS Treatises in Mathematics,Canadian Mathematical Society, Ottawa, ON; A K Peters, Ltd.,Wellesley, MA (2007).
51 / 67
![Page 277: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/277.jpg)
A combinatorial proof of the identity (4)
TheoremThe following identity holds for any positive integer p:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
For a proof, see
J. L. W. V. Jensen: Sur une identité d’Abel et sur d’autres formulesanalogues, Acta Math. 26 (1902), 307-318.
or
M. E. Larsen: Summa Summarum, CMS Treatises in Mathematics,Canadian Mathematical Society, Ottawa, ON; A K Peters, Ltd.,Wellesley, MA (2007).
51 / 67
![Page 278: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/278.jpg)
A combinatorial proof of the identity (4)
TheoremThe following identity holds for any positive integer p:
12
(p2k+1−1
p−1 + 1)
=
p2k+1−1p−1 −
∑k−1i=0 (−1)i((k−i)(p−1)−1
i
)(k+(k−i)p2k−i
).
For a proof, see
J. L. W. V. Jensen: Sur une identité d’Abel et sur d’autres formulesanalogues, Acta Math. 26 (1902), 307-318.
or
M. E. Larsen: Summa Summarum, CMS Treatises in Mathematics,Canadian Mathematical Society, Ottawa, ON; A K Peters, Ltd.,Wellesley, MA (2007).
51 / 67
![Page 279: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/279.jpg)
A Generalization to the Affine Case
Let H be a hyperplane of AG(n,q).A d-dimensional subspace L of AG(n,q), d ≤ n − 1, is either
disjoint from H, orcontained in H, orintersects H in a (d − 1)-space.
Cross BlockWe call L a cross block if dim(L ∩ H) = d − 1.
52 / 67
![Page 280: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/280.jpg)
A Generalization to the Affine Case
Let H be a hyperplane of AG(n,q).A d-dimensional subspace L of AG(n,q), d ≤ n − 1, is either
disjoint from H, orcontained in H, orintersects H in a (d − 1)-space.
Cross BlockWe call L a cross block if dim(L ∩ H) = d − 1.
52 / 67
![Page 281: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/281.jpg)
A Generalization to the Affine Case
Let H be a hyperplane of AG(n,q).A d-dimensional subspace L of AG(n,q), d ≤ n − 1, is either
disjoint from H, orcontained in H, orintersects H in a (d − 1)-space.
Cross BlockWe call L a cross block if dim(L ∩ H) = d − 1.
52 / 67
![Page 282: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/282.jpg)
A Generalization to the Affine Case
Let H be a hyperplane of AG(n,q).A d-dimensional subspace L of AG(n,q), d ≤ n − 1, is either
disjoint from H, orcontained in H, orintersects H in a (d − 1)-space.
Cross BlockWe call L a cross block if dim(L ∩ H) = d − 1.
52 / 67
![Page 283: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/283.jpg)
A Generalization to the Affine Case
Let H be a hyperplane of AG(n,q).A d-dimensional subspace L of AG(n,q), d ≤ n − 1, is either
disjoint from H, orcontained in H, orintersects H in a (d − 1)-space.
Cross BlockWe call L a cross block if dim(L ∩ H) = d − 1.
52 / 67
![Page 284: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/284.jpg)
A Generalization to the Affine Case
Let H be a hyperplane of AG(n,q).A d-dimensional subspace L of AG(n,q), d ≤ n − 1, is either
disjoint from H, orcontained in H, orintersects H in a (d − 1)-space.
Cross BlockWe call L a cross block if dim(L ∩ H) = d − 1.
52 / 67
![Page 285: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/285.jpg)
A Generalization to the Affine Case
Let H be a hyperplane of AG(n,q).A d-dimensional subspace L of AG(n,q), d ≤ n − 1, is either
disjoint from H, orcontained in H, orintersects H in a (d − 1)-space.
Cross BlockWe call L a cross block if dim(L ∩ H) = d − 1.
52 / 67
![Page 286: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/286.jpg)
An Affine Space “Distortion” Construction:
Let D = AGd (n,q).Fix a hyperplane H through 0 in AG(n,q).Fix a permutation α of the (d − 1)-spaces through 0 in H.Replace each cross block B = Bout ∪ Bin containing 0 withα(B) = Bout ∪ α(Bin).Replace each coset of B with a carefully chosen coset of α(B).If q = 2, we must similarly “distort” all other blocks B′ such thatB′in = Bin.
53 / 67
![Page 287: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/287.jpg)
An Affine Space “Distortion” Construction:
Let D = AGd (n,q).Fix a hyperplane H through 0 in AG(n,q).Fix a permutation α of the (d − 1)-spaces through 0 in H.Replace each cross block B = Bout ∪ Bin containing 0 withα(B) = Bout ∪ α(Bin).Replace each coset of B with a carefully chosen coset of α(B).If q = 2, we must similarly “distort” all other blocks B′ such thatB′in = Bin.
53 / 67
![Page 288: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/288.jpg)
An Affine Space “Distortion” Construction:
Let D = AGd (n,q).Fix a hyperplane H through 0 in AG(n,q).Fix a permutation α of the (d − 1)-spaces through 0 in H.Replace each cross block B = Bout ∪ Bin containing 0 withα(B) = Bout ∪ α(Bin).Replace each coset of B with a carefully chosen coset of α(B).If q = 2, we must similarly “distort” all other blocks B′ such thatB′in = Bin.
53 / 67
![Page 289: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/289.jpg)
An Affine Space “Distortion” Construction:
Let D = AGd (n,q).Fix a hyperplane H through 0 in AG(n,q).Fix a permutation α of the (d − 1)-spaces through 0 in H.Replace each cross block B = Bout ∪ Bin containing 0 withα(B) = Bout ∪ α(Bin).Replace each coset of B with a carefully chosen coset of α(B).If q = 2, we must similarly “distort” all other blocks B′ such thatB′in = Bin.
53 / 67
![Page 290: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/290.jpg)
An Affine Space “Distortion” Construction:
Let D = AGd (n,q).Fix a hyperplane H through 0 in AG(n,q).Fix a permutation α of the (d − 1)-spaces through 0 in H.Replace each cross block B = Bout ∪ Bin containing 0 withα(B) = Bout ∪ α(Bin).Replace each coset of B with a carefully chosen coset of α(B).If q = 2, we must similarly “distort” all other blocks B′ such thatB′in = Bin.
53 / 67
![Page 291: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/291.jpg)
An Affine Space “Distortion” Construction:
Let D = AGd (n,q).Fix a hyperplane H through 0 in AG(n,q).Fix a permutation α of the (d − 1)-spaces through 0 in H.Replace each cross block B = Bout ∪ Bin containing 0 withα(B) = Bout ∪ α(Bin).Replace each coset of B with a carefully chosen coset of α(B).If q = 2, we must similarly “distort” all other blocks B′ such thatB′in = Bin.
53 / 67
![Page 292: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/292.jpg)
Affine construction
d-1d-1d-1d-1
Hyperplaneand cosets
Remainderand cosets
Apply a permutation to the (d-1)-spaces in the
hyperplane and cosets
TwistedHyperplaneand cosets
Unchanged Remainderand cosets
54 / 67
![Page 293: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/293.jpg)
Affine Construction: Details
What is a “carefully chosen” coset of α(B)?α(B) is not a vector subspace any longer.There are qn−d cosets of B by elements of H.There are also qn−d cosets of α(Bin) by elements of H.Choose qn−d elements of H so that each coset of B and eachcoset of α(Bin) is represented (possible by Hall’s Theorem).
The binary caseIn the binary case, we must do the same thing for all other blocks B′
such that B′in = Bin, to avoid transforming different blocks into identicalones.
55 / 67
![Page 294: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/294.jpg)
Affine Construction: Details
What is a “carefully chosen” coset of α(B)?α(B) is not a vector subspace any longer.There are qn−d cosets of B by elements of H.There are also qn−d cosets of α(Bin) by elements of H.Choose qn−d elements of H so that each coset of B and eachcoset of α(Bin) is represented (possible by Hall’s Theorem).
The binary caseIn the binary case, we must do the same thing for all other blocks B′
such that B′in = Bin, to avoid transforming different blocks into identicalones.
55 / 67
![Page 295: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/295.jpg)
Affine Construction: Details
What is a “carefully chosen” coset of α(B)?α(B) is not a vector subspace any longer.There are qn−d cosets of B by elements of H.There are also qn−d cosets of α(Bin) by elements of H.Choose qn−d elements of H so that each coset of B and eachcoset of α(Bin) is represented (possible by Hall’s Theorem).
The binary caseIn the binary case, we must do the same thing for all other blocks B′
such that B′in = Bin, to avoid transforming different blocks into identicalones.
55 / 67
![Page 296: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/296.jpg)
Affine Construction: Details
What is a “carefully chosen” coset of α(B)?α(B) is not a vector subspace any longer.There are qn−d cosets of B by elements of H.There are also qn−d cosets of α(Bin) by elements of H.Choose qn−d elements of H so that each coset of B and eachcoset of α(Bin) is represented (possible by Hall’s Theorem).
The binary caseIn the binary case, we must do the same thing for all other blocks B′
such that B′in = Bin, to avoid transforming different blocks into identicalones.
55 / 67
![Page 297: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/297.jpg)
Affine Construction: Details
What is a “carefully chosen” coset of α(B)?α(B) is not a vector subspace any longer.There are qn−d cosets of B by elements of H.There are also qn−d cosets of α(Bin) by elements of H.Choose qn−d elements of H so that each coset of B and eachcoset of α(Bin) is represented (possible by Hall’s Theorem).
The binary caseIn the binary case, we must do the same thing for all other blocks B′
such that B′in = Bin, to avoid transforming different blocks into identicalones.
55 / 67
![Page 298: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/298.jpg)
Affine Construction: Details
What is a “carefully chosen” coset of α(B)?α(B) is not a vector subspace any longer.There are qn−d cosets of B by elements of H.There are also qn−d cosets of α(Bin) by elements of H.Choose qn−d elements of H so that each coset of B and eachcoset of α(Bin) is represented (possible by Hall’s Theorem).
The binary caseIn the binary case, we must do the same thing for all other blocks B′
such that B′in = Bin, to avoid transforming different blocks into identicalones.
55 / 67
![Page 299: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/299.jpg)
Affine Construction: Details
What is a “carefully chosen” coset of α(B)?α(B) is not a vector subspace any longer.There are qn−d cosets of B by elements of H.There are also qn−d cosets of α(Bin) by elements of H.Choose qn−d elements of H so that each coset of B and eachcoset of α(Bin) is represented (possible by Hall’s Theorem).
The binary caseIn the binary case, we must do the same thing for all other blocks B′
such that B′in = Bin, to avoid transforming different blocks into identicalones.
55 / 67
![Page 300: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/300.jpg)
Affine results
NoteAny polarity of PG(2d ,q) permutes affine d-spaces containing 0 inAG(2d + 1,q).
Theorem. (Clark, Jungnickel, Tonchev 2009):Let
α be a polarity of PG(2d ,2), extended to affine d-subspaces inAG(2d + 1,2), andD = AGd+1(2d + 1,2), d ≥ 2.
Then α(D) is a design with the same parameters and the same 2-rankas D, but is not isomorphic to D.
56 / 67
![Page 301: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/301.jpg)
Affine results
NoteAny polarity of PG(2d ,q) permutes affine d-spaces containing 0 inAG(2d + 1,q).
Theorem. (Clark, Jungnickel, Tonchev 2009):Let
α be a polarity of PG(2d ,2), extended to affine d-subspaces inAG(2d + 1,2), andD = AGd+1(2d + 1,2), d ≥ 2.
Then α(D) is a design with the same parameters and the same 2-rankas D, but is not isomorphic to D.
56 / 67
![Page 302: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/302.jpg)
Affine results
NoteAny polarity of PG(2d ,q) permutes affine d-spaces containing 0 inAG(2d + 1,q).
Theorem. (Clark, Jungnickel, Tonchev 2009):Let
α be a polarity of PG(2d ,2), extended to affine d-subspaces inAG(2d + 1,2), andD = AGd+1(2d + 1,2), d ≥ 2.
Then α(D) is a design with the same parameters and the same 2-rankas D, but is not isomorphic to D.
56 / 67
![Page 303: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/303.jpg)
Affine results
NoteAny polarity of PG(2d ,q) permutes affine d-spaces containing 0 inAG(2d + 1,q).
Theorem. (Clark, Jungnickel, Tonchev 2009):Let
α be a polarity of PG(2d ,2), extended to affine d-subspaces inAG(2d + 1,2), andD = AGd+1(2d + 1,2), d ≥ 2.
Then α(D) is a design with the same parameters and the same 2-rankas D, but is not isomorphic to D.
56 / 67
![Page 304: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/304.jpg)
Affine results
NoteAny polarity of PG(2d ,q) permutes affine d-spaces containing 0 inAG(2d + 1,q).
Theorem. (Clark, Jungnickel, Tonchev 2009):Let
α be a polarity of PG(2d ,2), extended to affine d-subspaces inAG(2d + 1,2), andD = AGd+1(2d + 1,2), d ≥ 2.
Then α(D) is a design with the same parameters and the same 2-rankas D, but is not isomorphic to D.
56 / 67
![Page 305: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/305.jpg)
Sketch of Proof
The block code of AGd+1(2d + 1,2) is a self-dual Reed-Mullercode R(d ,2d + 1) of dimension 22d .The block intersection numbers of D and α(D) are 0 and 2i for1 ≤ i < 2d , and are all even.The block code of α(D) is self-orthogonal, andrk2(α(D)) ≤ 22d = rk2(D). Thus
22d = rk2(PGd (2d ,2)) ≤ rk2(α(D)) ≤ rk2(D) = 22d
NoteThe subdesign induced on H is isomorphic to α(PGd (2d ,2)).By the projective construction, its 2-rank is equal to rk2(PGd (2d ,2)),but is not isomorphic to PGd (2d ,2).
57 / 67
![Page 306: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/306.jpg)
Sketch of Proof
The block code of AGd+1(2d + 1,2) is a self-dual Reed-Mullercode R(d ,2d + 1) of dimension 22d .The block intersection numbers of D and α(D) are 0 and 2i for1 ≤ i < 2d , and are all even.The block code of α(D) is self-orthogonal, andrk2(α(D)) ≤ 22d = rk2(D). Thus
22d = rk2(PGd (2d ,2)) ≤ rk2(α(D)) ≤ rk2(D) = 22d
NoteThe subdesign induced on H is isomorphic to α(PGd (2d ,2)).By the projective construction, its 2-rank is equal to rk2(PGd (2d ,2)),but is not isomorphic to PGd (2d ,2).
57 / 67
![Page 307: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/307.jpg)
Sketch of Proof
The block code of AGd+1(2d + 1,2) is a self-dual Reed-Mullercode R(d ,2d + 1) of dimension 22d .The block intersection numbers of D and α(D) are 0 and 2i for1 ≤ i < 2d , and are all even.The block code of α(D) is self-orthogonal, andrk2(α(D)) ≤ 22d = rk2(D). Thus
22d = rk2(PGd (2d ,2)) ≤ rk2(α(D)) ≤ rk2(D) = 22d
NoteThe subdesign induced on H is isomorphic to α(PGd (2d ,2)).By the projective construction, its 2-rank is equal to rk2(PGd (2d ,2)),but is not isomorphic to PGd (2d ,2).
57 / 67
![Page 308: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/308.jpg)
Sketch of Proof
The block code of AGd+1(2d + 1,2) is a self-dual Reed-Mullercode R(d ,2d + 1) of dimension 22d .The block intersection numbers of D and α(D) are 0 and 2i for1 ≤ i < 2d , and are all even.The block code of α(D) is self-orthogonal, andrk2(α(D)) ≤ 22d = rk2(D). Thus
22d = rk2(PGd (2d ,2)) ≤ rk2(α(D)) ≤ rk2(D) = 22d
NoteThe subdesign induced on H is isomorphic to α(PGd (2d ,2)).By the projective construction, its 2-rank is equal to rk2(PGd (2d ,2)),but is not isomorphic to PGd (2d ,2).
57 / 67
![Page 309: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/309.jpg)
Sketch of Proof
The block code of AGd+1(2d + 1,2) is a self-dual Reed-Mullercode R(d ,2d + 1) of dimension 22d .The block intersection numbers of D and α(D) are 0 and 2i for1 ≤ i < 2d , and are all even.The block code of α(D) is self-orthogonal, andrk2(α(D)) ≤ 22d = rk2(D). Thus
22d = rk2(PGd (2d ,2)) ≤ rk2(α(D)) ≤ rk2(D) = 22d
NoteThe subdesign induced on H is isomorphic to α(PGd (2d ,2)).By the projective construction, its 2-rank is equal to rk2(PGd (2d ,2)),but is not isomorphic to PGd (2d ,2).
57 / 67
![Page 310: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/310.jpg)
Nonisomorphic designs with geometricparamatars
The number of nonisomorphic designs with the same parametersas AGn−1(n,q) or PGn−1(n,q), n ≥ 3, grows linearly with n:Bhat and Shrikhande (1970), Griffiths and Mavron (1972).
Jungnickel (1984): The number of nonisomorphic designs withthe same parameters as AGn−1(n,q) or PGn−1(n,q), n ≥ 3,grows exponentially.
Jungnickel’s bounds were later improved:Kantor ’94, C. Lam, S. Lam, & T., 2000,2003.
58 / 67
![Page 311: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/311.jpg)
Nonisomorphic designs with geometricparamatars
The number of nonisomorphic designs with the same parametersas AGn−1(n,q) or PGn−1(n,q), n ≥ 3, grows linearly with n:Bhat and Shrikhande (1970), Griffiths and Mavron (1972).
Jungnickel (1984): The number of nonisomorphic designs withthe same parameters as AGn−1(n,q) or PGn−1(n,q), n ≥ 3,grows exponentially.
Jungnickel’s bounds were later improved:Kantor ’94, C. Lam, S. Lam, & T., 2000,2003.
58 / 67
![Page 312: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/312.jpg)
Nonisomorphic designs with geometricparamatars
The number of nonisomorphic designs with the same parametersas AGn−1(n,q) or PGn−1(n,q), n ≥ 3, grows linearly with n:Bhat and Shrikhande (1970), Griffiths and Mavron (1972).
Jungnickel (1984): The number of nonisomorphic designs withthe same parameters as AGn−1(n,q) or PGn−1(n,q), n ≥ 3,grows exponentially.
Jungnickel’s bounds were later improved:Kantor ’94, C. Lam, S. Lam, & T., 2000,2003.
58 / 67
![Page 313: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/313.jpg)
New bounds for any 3 ≤ d ≤ n − 1
Jungnickel & T., 2009The number of nonisomorphic designs with parameters of PGd (n,q),3 ≤ d ≤ n − 1, grows exponentially.
Theorem. (Clark, Jungnickel, Tonchev, 2009)The number of nonisomprphic 2-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of nonisomprphic 3-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of resolvable 3-designs with the parameters ofAGd (n,2) grows exponentially.
59 / 67
![Page 314: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/314.jpg)
New bounds for any 3 ≤ d ≤ n − 1
Jungnickel & T., 2009The number of nonisomorphic designs with parameters of PGd (n,q),3 ≤ d ≤ n − 1, grows exponentially.
Theorem. (Clark, Jungnickel, Tonchev, 2009)The number of nonisomprphic 2-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of nonisomprphic 3-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of resolvable 3-designs with the parameters ofAGd (n,2) grows exponentially.
59 / 67
![Page 315: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/315.jpg)
New bounds for any 3 ≤ d ≤ n − 1
Jungnickel & T., 2009The number of nonisomorphic designs with parameters of PGd (n,q),3 ≤ d ≤ n − 1, grows exponentially.
Theorem. (Clark, Jungnickel, Tonchev, 2009)The number of nonisomprphic 2-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of nonisomprphic 3-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of resolvable 3-designs with the parameters ofAGd (n,2) grows exponentially.
59 / 67
![Page 316: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/316.jpg)
New bounds for any 3 ≤ d ≤ n − 1
Jungnickel & T., 2009The number of nonisomorphic designs with parameters of PGd (n,q),3 ≤ d ≤ n − 1, grows exponentially.
Theorem. (Clark, Jungnickel, Tonchev, 2009)The number of nonisomprphic 2-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of nonisomprphic 3-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of resolvable 3-designs with the parameters ofAGd (n,2) grows exponentially.
59 / 67
![Page 317: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/317.jpg)
New bounds for any 3 ≤ d ≤ n − 1
Jungnickel & T., 2009The number of nonisomorphic designs with parameters of PGd (n,q),3 ≤ d ≤ n − 1, grows exponentially.
Theorem. (Clark, Jungnickel, Tonchev, 2009)The number of nonisomprphic 2-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of nonisomprphic 3-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of resolvable 3-designs with the parameters ofAGd (n,2) grows exponentially.
59 / 67
![Page 318: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/318.jpg)
New bounds for any 3 ≤ d ≤ n − 1
Jungnickel & T., 2009The number of nonisomorphic designs with parameters of PGd (n,q),3 ≤ d ≤ n − 1, grows exponentially.
Theorem. (Clark, Jungnickel, Tonchev, 2009)The number of nonisomprphic 2-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of nonisomprphic 3-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of resolvable 3-designs with the parameters ofAGd (n,2) grows exponentially.
59 / 67
![Page 319: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/319.jpg)
New bounds for any 3 ≤ d ≤ n − 1
Jungnickel & T., 2009The number of nonisomorphic designs with parameters of PGd (n,q),3 ≤ d ≤ n − 1, grows exponentially.
Theorem. (Clark, Jungnickel, Tonchev, 2009)The number of nonisomprphic 2-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of nonisomprphic 3-designs with parameters ofAGd (n,2), d ≥ 3, grows exponentially.The number of resolvable 3-designs with the parameters ofAGd (n,2) grows exponentially.
59 / 67
![Page 320: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/320.jpg)
ExamplesThere exist at least
10228 non-isomorphic 2-(32,8,35) designs,1075 resolvable 2-(32,8,35) designs,1027 resolvable 3-(32,8,7) designs,
all having the parameters of AG3(5,2).
60 / 67
![Page 321: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/321.jpg)
ExamplesThere exist at least
10228 non-isomorphic 2-(32,8,35) designs,1075 resolvable 2-(32,8,35) designs,1027 resolvable 3-(32,8,7) designs,
all having the parameters of AG3(5,2).
60 / 67
![Page 322: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/322.jpg)
ExamplesThere exist at least
10228 non-isomorphic 2-(32,8,35) designs,1075 resolvable 2-(32,8,35) designs,1027 resolvable 3-(32,8,7) designs,
all having the parameters of AG3(5,2).
60 / 67
![Page 323: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/323.jpg)
ExamplesThere exist at least
10228 non-isomorphic 2-(32,8,35) designs,1075 resolvable 2-(32,8,35) designs,1027 resolvable 3-(32,8,7) designs,
all having the parameters of AG3(5,2).
60 / 67
![Page 324: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/324.jpg)
ExamplesThere exist at least
10228 non-isomorphic 2-(32,8,35) designs,1075 resolvable 2-(32,8,35) designs,1027 resolvable 3-(32,8,7) designs,
all having the parameters of AG3(5,2).
60 / 67
![Page 325: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/325.jpg)
Open Problems
Hamada’s conjecture (strong form)If D is a design having the same parameters as a geometric design G,G = PGd (n,q) or G = AGd (n,q), then
rankqD ≥ rankqG,
with equality rankqD = rankqG if and only if D is isiomorphic to G.
NoteThe strong form (the "only if" part) of Hamada’s conjecture is not truein general.
Open ProblemDetermine the spectrum of parameters n,q,d for which the strongform of Hamada’s conjecture holds true.
61 / 67
![Page 326: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/326.jpg)
Open Problems
Hamada’s conjecture (strong form)If D is a design having the same parameters as a geometric design G,G = PGd (n,q) or G = AGd (n,q), then
rankqD ≥ rankqG,
with equality rankqD = rankqG if and only if D is isiomorphic to G.
NoteThe strong form (the "only if" part) of Hamada’s conjecture is not truein general.
Open ProblemDetermine the spectrum of parameters n,q,d for which the strongform of Hamada’s conjecture holds true.
61 / 67
![Page 327: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/327.jpg)
Open Problems
Hamada’s conjecture (strong form)If D is a design having the same parameters as a geometric design G,G = PGd (n,q) or G = AGd (n,q), then
rankqD ≥ rankqG,
with equality rankqD = rankqG if and only if D is isiomorphic to G.
NoteThe strong form (the "only if" part) of Hamada’s conjecture is not truein general.
Open ProblemDetermine the spectrum of parameters n,q,d for which the strongform of Hamada’s conjecture holds true.
61 / 67
![Page 328: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/328.jpg)
Open Problems
Hamada’s conjecture (strong form)If D is a design having the same parameters as a geometric design G,G = PGd (n,q) or G = AGd (n,q), then
rankqD ≥ rankqG,
with equality rankqD = rankqG if and only if D is isiomorphic to G.
NoteThe strong form (the "only if" part) of Hamada’s conjecture is not truein general.
Open ProblemDetermine the spectrum of parameters n,q,d for which the strongform of Hamada’s conjecture holds true.
61 / 67
![Page 329: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/329.jpg)
Open Problems
Hamada’s conjecture (strong form)If D is a design having the same parameters as a geometric design G,G = PGd (n,q) or G = AGd (n,q), then
rankqD ≥ rankqG,
with equality rankqD = rankqG if and only if D is isiomorphic to G.
NoteThe strong form (the "only if" part) of Hamada’s conjecture is not truein general.
Open ProblemDetermine the spectrum of parameters n,q,d for which the strongform of Hamada’s conjecture holds true.
61 / 67
![Page 330: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/330.jpg)
Open Problems
Hamada’s conjecture (strong form)If D is a design having the same parameters as a geometric design G,G = PGd (n,q) or G = AGd (n,q), then
rankqD ≥ rankqG,
with equality rankqD = rankqG if and only if D is isiomorphic to G.
NoteThe strong form (the "only if" part) of Hamada’s conjecture is not truein general.
Open ProblemDetermine the spectrum of parameters n,q,d for which the strongform of Hamada’s conjecture holds true.
61 / 67
![Page 331: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/331.jpg)
Modified versions of Hamada’s conjecture
Assmus-KeyHamada’s conjecture is true for PGn−1(n,q).
SacharHamada’s conjecture is true for PG1(2,q), that is, for projective planes.
NoteThe Assmus-Key conjecture has been proved for q = 2.Sachar’s conjecture has been verified for q ≤ 9.
62 / 67
![Page 332: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/332.jpg)
Modified versions of Hamada’s conjecture
Assmus-KeyHamada’s conjecture is true for PGn−1(n,q).
SacharHamada’s conjecture is true for PG1(2,q), that is, for projective planes.
NoteThe Assmus-Key conjecture has been proved for q = 2.Sachar’s conjecture has been verified for q ≤ 9.
62 / 67
![Page 333: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/333.jpg)
Modified versions of Hamada’s conjecture
Assmus-KeyHamada’s conjecture is true for PGn−1(n,q).
SacharHamada’s conjecture is true for PG1(2,q), that is, for projective planes.
NoteThe Assmus-Key conjecture has been proved for q = 2.Sachar’s conjecture has been verified for q ≤ 9.
62 / 67
![Page 334: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/334.jpg)
Modified versions of Hamada’s conjecture
Assmus-KeyHamada’s conjecture is true for PGn−1(n,q).
SacharHamada’s conjecture is true for PG1(2,q), that is, for projective planes.
NoteThe Assmus-Key conjecture has been proved for q = 2.Sachar’s conjecture has been verified for q ≤ 9.
62 / 67
![Page 335: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/335.jpg)
A weaker conjecture
Hamada’s conjecture (weaker form)The p-rank of PGd (n,ps) or AGd (n,ps) is an exact lower bound on thep-rank of all designs having the same parameters as PGd (n,ps) orAGd (n,ps).
Open ProblemsProve the weaker form of Hamada’s conjecture.Prove lower bounds on the p-rank of designs with geometricparameters.
63 / 67
![Page 336: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/336.jpg)
A weaker conjecture
Hamada’s conjecture (weaker form)The p-rank of PGd (n,ps) or AGd (n,ps) is an exact lower bound on thep-rank of all designs having the same parameters as PGd (n,ps) orAGd (n,ps).
Open ProblemsProve the weaker form of Hamada’s conjecture.Prove lower bounds on the p-rank of designs with geometricparameters.
63 / 67
![Page 337: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/337.jpg)
A weaker conjecture
Hamada’s conjecture (weaker form)The p-rank of PGd (n,ps) or AGd (n,ps) is an exact lower bound on thep-rank of all designs having the same parameters as PGd (n,ps) orAGd (n,ps).
Open ProblemsProve the weaker form of Hamada’s conjecture.Prove lower bounds on the p-rank of designs with geometricparameters.
63 / 67
![Page 338: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/338.jpg)
A weaker conjecture
Hamada’s conjecture (weaker form)The p-rank of PGd (n,ps) or AGd (n,ps) is an exact lower bound on thep-rank of all designs having the same parameters as PGd (n,ps) orAGd (n,ps).
Open ProblemsProve the weaker form of Hamada’s conjecture.Prove lower bounds on the p-rank of designs with geometricparameters.
63 / 67
![Page 339: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/339.jpg)
A weaker conjecture
Hamada’s conjecture (weaker form)The p-rank of PGd (n,ps) or AGd (n,ps) is an exact lower bound on thep-rank of all designs having the same parameters as PGd (n,ps) orAGd (n,ps).
Open ProblemsProve the weaker form of Hamada’s conjecture.Prove lower bounds on the p-rank of designs with geometricparameters.
63 / 67
![Page 340: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/340.jpg)
A weaker conjecture
Hamada’s conjecture (weaker form)The p-rank of PGd (n,ps) or AGd (n,ps) is an exact lower bound on thep-rank of all designs having the same parameters as PGd (n,ps) orAGd (n,ps).
Open ProblemsProve the weaker form of Hamada’s conjecture.Prove lower bounds on the p-rank of designs with geometricparameters.
63 / 67
![Page 341: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/341.jpg)
More open problems
Affine case, q > 2: Extend the affine construction to fields oforder q > 2.
Study the resulting new codes: compare with the classicalgeometric and Reed-Muller codes.
Develope decoding algorithms for the new codes.
Generalize the remaining sporadic counterexamples havingthe parameters of AG2(3,4).
64 / 67
![Page 342: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/342.jpg)
More open problems
Affine case, q > 2: Extend the affine construction to fields oforder q > 2.
Study the resulting new codes: compare with the classicalgeometric and Reed-Muller codes.
Develope decoding algorithms for the new codes.
Generalize the remaining sporadic counterexamples havingthe parameters of AG2(3,4).
64 / 67
![Page 343: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/343.jpg)
More open problems
Affine case, q > 2: Extend the affine construction to fields oforder q > 2.
Study the resulting new codes: compare with the classicalgeometric and Reed-Muller codes.
Develope decoding algorithms for the new codes.
Generalize the remaining sporadic counterexamples havingthe parameters of AG2(3,4).
64 / 67
![Page 344: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/344.jpg)
More open problems
Affine case, q > 2: Extend the affine construction to fields oforder q > 2.
Study the resulting new codes: compare with the classicalgeometric and Reed-Muller codes.
Develope decoding algorithms for the new codes.
Generalize the remaining sporadic counterexamples havingthe parameters of AG2(3,4).
64 / 67
![Page 345: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/345.jpg)
More open problems
Affine case, q > 2: Extend the affine construction to fields oforder q > 2.
Study the resulting new codes: compare with the classicalgeometric and Reed-Muller codes.
Develope decoding algorithms for the new codes.
Generalize the remaining sporadic counterexamples havingthe parameters of AG2(3,4).
64 / 67
![Page 346: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/346.jpg)
References
N. Hamada,On the p-rank of the incidence matrix of a balanced or partiallybalanced incomplete block design and its applications toerror-correcting codes, Hiroshima Math J. 3 (1973), pp. 153-226.
D. Jungnickel and V.D. Tonchev,Polarities, Quasi-symmetric Designs, and Hamada’s Conjecture,Designs, Codes and Cryptography 51 (2009), 131–140.
D. Jungnickel and V.D. Tonchev,The number of designs with geometric parameters growsexponentially, Designs, Codes and Cryptography, 55 (2010),131-140.
D. Clark, D. Jungnickel, V. D. Tonchev,An Infinite Family of Counterexamples to the Affine Case ofHamada’s Conjecture. submitted.
D. Clark, D. Jungnickel, V.D. Tonchev,Exponential bounds on the number of designs with affineparameters, J. Combinatorial Designs, in press.
65 / 67
![Page 347: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/347.jpg)
Thank You!
Thank You!
66 / 67
![Page 348: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/348.jpg)
Thank You!
Thank You!
66 / 67
![Page 349: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/349.jpg)
Thank You!
Thank You!
66 / 67
![Page 350: Finite geometry, designs, codes, and Hamada's conjecture](https://reader030.fdocuments.in/reader030/viewer/2022012021/6168a423d394e9041f716e27/html5/thumbnails/350.jpg)
Any Questions?
67 / 67