the nullspace of nested magic squares · when we want to obtain pandiagonal magic squares 6 by 6,...
TRANSCRIPT
IJRRAS 8 (3) ● September 2011 www.arpapress.com/Volumes/Vol8Issue3/IJRRAS_8_3_07.pdf
311
THE NULLSPACE OF NESTED MAGIC SQUARES
Ayoub ElShokry1 & Saleem Al-Ashhab
2
University of Khartoum ,Faculty of Education ,Department of Mathematics , Omdurman ,Sudan
Depatement of Mathematics, Al Al-Bayt University, Mafrq, Jordan
Email: ayou1975@ maktoob.com; [email protected]
ABSTRACT
In this paper we formulate the general structure of the nullspace and subspaces of the nullspace for nested magic
squares, where we consider two different types of centre. Further, we study the properties of these spaces.
AMS classification number: 15A15.
Key Words: Nullspace, Magic Squares, Mathematical induction.
1. INTRODUCTION
We consider magic squares here as matrices and study the algebraic properties for them. Hence, a semi magic square
is a n by n matrix such that the sum of the entries in each row and columns is the same. The common value is called
the magic constant. If, in addition, the sum of all entries in each left-broken diagonal and each right-broken diagonal
is the magic constant, then we call the matrix a pandiagonal magic square. It is well-known that the following
structure
2s–A–B–C C B A
B+C – E A+E – C 2s– A–B–E E
s – B s – A A+B+C – s
s – C
A+B+E–s
s – E s–B–C+E
s–A–E+C
is a general structure of the pandiagonal magic square 4 by 4. Here, the magic constant is 2s.
When we want to obtain pandiagonal magic squares 6 by 6, having a similar structure to that of the pandiagonal
magic squares 4 by 4, we consider the matrix
a6 a5 a4 a3 a2 a1
a12 a11 a10 a9 a8 a7
a18 a17 a16 a15 a14 a13
s – a3 s – a2 s – a1 s – a6 s – a5 s – a4
s – a9 s – a8 s – a7 s – a12 s – a11 s – a10
s – a15 s – a14 s – a13 s – a18 s – a17 s – a16
We then require the following conditions:
0
0
0
3
3
3
181512963
171411852
161310741
181716151413
121110987
654321
aaaaaa
aaaaaa
aaaaaa
saaaaaa
saaaaaa
saaaaaa
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
312
These conditions suffice to ensure that the resulting square is pandiagonal. We have as a possible form of these
squares:
J–L –K+J+I+H+E+D+C
– 3/2 s
L–J+G–
E+B
K–I+F–
D+A
–L–K–H–G–F–C–B–A+
9/2 s
L K
–J–I–H–G–F+3s J I H G F
–E–D–C–B–A+3s E D C B A
L+K+H+G+F+C+B+A–7/2
s
s–L s–K L+K–J–I–H–E–D–C+
5/2s
s–L+J–
G+E–B
s–K+I–
F+D–A
s–H s–G s–F J+I+H+G+F–2s s–J s–I
s– C s–B s–A E+D+C+B+A –2s s–E s–D
We call this kind of squares the pandiagonal magic square 6 by 6 of the special structure. Note that the magic
constant is now 3s.
The nullspace of a matrix A is the solution set of the homogenous system Ax = 0. The pandiagonal magic square 4
by 4 possesses a nontrivial null space. Using the previous notation for a pandiagonal magic square 4 by 4 we can
easily prove (see [1]) that this nullspace can be written in the following form
}:22
22
{ Rz
CA
CBAs
AC
sCBA
z
Note that the sum of the entries is zero. Furthermore, the multiplication of the square as a matrix with the vector
Rzqs
q
CA
CBAs
AC
sCBA
z
,,
1
1
1
1
222
22
yields the vector
q
q
q
q
This is due to definition of the magic square. In [1] we find that the pandiagonal magic square 6 by 6 of the special
structure possesses also a nontrivial nullspace, which can be written in the following form:
}:
AJ)+DG-DJ+2JC+2BJ+2EG-BF+2CG-EF-AG-3Es+3Gs+3Js-3Bs-2BH+2EH-EI-BI(
DG)+AJ-AG-2CF+2DF+DJ+BF+EF+3Fs-3Ds-3As+2DH+2AH-EI-2IC-2AI-BI-3Is(
BF)-AJ-EF+EI-BI+AG+DG-DJ(
AJ)+DG-DJ+2JC+2BJ+2EG-BF+2CG-EF-AG-3Es+3Gs+3Js-3Bs-2BH+2EH-EI-BI(
DG)+AJ-AG-2CF+2DF+DJ+BF+EF+3Fs-3Ds-3As+2DH+2AH-EI-2IC-2AI-BI-3Is(
BF)-AJ-EF+EI-BI+AG+DG-DJ(
{ Rzz
We will therefore denote the vectors in this nullspace as follows:
3
2
1
3
2
1
x
x
x
x
x
x
In this case the multiplication of the square with the vector
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
313
Rzqs
q
x
x
x
x
x
x
z
, ,
1
1
1
1
1
1
3
3
2
1
3
2
1
yields the vector
q
q
q
q
q
q
Nested magic squares
By nested squares we mean matrices having certain properties, which make them and some of their submatrices
semi magic squares or pandiagonal magic squares. By a nested semi magic square 6 by 6 with a pandiagonal magic
square 4 by 4 we mean a matrix of the form
d11 b14 b13 b12 b11 r11
s – p11 2s–A–B–C C B A p11
s – p21 B +C – E A+E–C 2s–A–B–E E P21
s – p31 s – B s–A A+B+C–s s– C P31
s – p41 A+B+E–s s–E S–B–C+E s–A–E+C P41
f11 s – b14 s – b13 s – b12 s – b11 e11
where we require
sebbbbf
pppprse
bbbbrsd
111413121111
413121111111
141312111111
)1(..........3
3
This ensures that the matrix as a whole is a semi magic square with 3s as a magic constant. In the centre we have
a pandiagonal magic square. We introduce also the concept of the multi-nested semi magic square 2n by 2n with a
pandiagonal magic square 4 by 4. This will be the following matrix
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
314
rmm rm(m-1) . . . rm1 bm1 bm2 bm3 bm4 dm1 . . . dm(m-1) dmm
r(m-1)m . . . . . . . . s –r(m-
1)m
. . . . . . . . . .
. . . . . . . . . .
r2m . . . r22 r21 B21 b22 b23 b24 d21 d22 . . . s –r2m
r1m . . . r12 r11 B11 b12 b13 b14 d11 s –
r12
. . . s –r1m
p1m . . . p12 p11 A B C 2s–A–B–C s –
p11
s –
p12
. . . s –p1m
P2m . . . P22 P21 E 2s–A–
B–E
A+
E –
C
B +C – E s –
P21
s –
P22
. . . s –
P2m
P3m . . . P32 P31
s – C
–s + A
+B+C
s –
A
s – B
s –
P31
s –
P32
. . . s –
P3m
P4m . . . P42 P41 s–A–
E+C
s–B–
C+E
s –
E
A+B+E–s s –
P41
s –
P42
. . . s –
P4m
e1m . . . e12 e11 s – b11 s – b12 s –
b13
s – b14 f11 s –
e12
. . . s –e1m
e2m . . . e22 s –
r21
s – b21 s – b22 s –
b23
s – b24 s –
d21
f22 . . . s –e2m
. . . . . . . . . .
. . . . . . . . . .
e(m-1)m . . . . . . . . s –
e(m-1)m
emm s –
rm(m-1)
. . . s –
rm1
s –bm1 s – bm2 s –
bm3
s – bm4 s –
dm1
. . . s –
dm(m-1)
fmm
where m= n – 2 and
)1(143211)1( ...... mmmmmmmmmmmmmm ddbbbbrrrnsd
mmmmmmmmmmmmmm eepppprrrnse )1(143211)1( ......
sneddbbbbrrf mmmmmmmmmmmmmm )2(...... )1(143211)1(
Note that we obtain semi magic square each time we remove the outer frame of the square.
By a multi-nested semi magic square 2n by 2n with a pandiagonal magic square 6 by 6 we mean the following
matrix:
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
315
rmm rm(m-1) ... rm1 Vm6 Vm5 Vm4 Vm3 Vm2 Vm1 Pm1 … pm(m-1) pmm
s –
p(m-1)m
. . . . . . . . . . P(m-
1)m
. . . . . . . . . . .
. . . . . . . . . .
. r22 r21 v26 v25 v24 v23 v22 v21 p21
p22
.
s –p1m … s –
p12
r11 v16 v15 v14 v13 v12 v11 p11 p12 … P1m
s –
u1m
… s –
u12
s –
u11
K L
–L –K
–H–G
–F–C –
B–A+
9/2s
K–
I+F–
D+A
L – J
+ G –
E + B
–L –
K+J+I
+H+E+
D+C–
3/2s
u11 u12 … u1m
s –
u2m
… s –
u22
s –
u21
F G H I J –J–I –
H–G–
F + 3s
u21 u22 … u2m
s –u3m … s –
u32
s –
u31
A B C D E –E–D –
C–B -
A+3s
u31 u32 … u3m
s –u4m … s –
u42
s –
u41
s– K
+ I –F
+D –
A
s–L
–J–
G+E
–B
L+K– J
– I –H
–E– D–
C+
5/2s
s –K s – L L+K+
H+G+
F+C+B
+A–
7/2 s
u41 u42 … u4m
s –u5m … s –
u52
s –
u51
s – I s – J J + I
+H+G
+F–2s
s – F s – G s – H u51 u52 … u5m
s –u6m … s –
u62
s –
u61
s – D s –
E
E+D+
C+B+
A –2s
s – A s – B s – C u61 u62 … u6m
s –
w1m
… s –
w12
t11 s –
v16
s –
v15
s – v14 s –
v13
s –
v12
s – v11 w11
w12 ... w1m
. t22 s – r21 s –
v26
s –
v25
s – v24 s –
v23
s –
v22
s – v21 s –
p21
w22 .
. . . . . . . . . .
. . . . . . . . . .
s –
w(m-
1)m
. . . . . . . . . . w(m-
1)m
tmm s –rm(m-
1)
. s –rm1 s –vm6 s –
vm5
s –vm4 s –vm3 s –vm2 s –vm1 s –
pm1
. s –pm(m-1) wmm
where m = n – 2,
,...... )1(16543211)1( mmmmmmmmmmmmmmmm rrvvvvvvpppnsr
,...... )1(16543211)1( mmmmmmmmmmmmmmmm wwuuuuuupppnsw
.)2(...... )1(16543211)1( snwrrvvvvvvppt mmmmmmmmmmmmmmmm
For example, we define the nested semi magic square 8 by 8 with a 6 by 6 pandiagonal square as the following
square:
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
316
r11 v16 v15 v14 v13 v12 v11 p11
s – u11 K L –L –K –H
–G –F –C
–B –A +
9/2 s
K–I–
D+F+A
L – J + G
– E + B
– L –K + J +I
+H + E + D +
C
– 3/2 s
u11
s – u21
F
G
H
I
J
–J –I –H –G–
F+ 3s
u21
s – u31
A
B
C
D
E
–E –D –C –B–
A+ 3s
u31
s – u41 s – K –
F +D –
A +I
s – L
+J –
G+ E
–B
L + K – J
– I –H –E
–D – C +
5/2s
s –K s – L L + K + H + G
+ F + C + B +
A – 7/2 s
u41
s – u51
s – I
s – J
J + I + H
+ G +F –
2s
s – F
s – G
s – H
u51
s – u61
s – D
s – E E+D+C +
B+A –2s
s – A
s – B
s –C u61
t11 s – v16 s–v15 s–v14 s – v13 s – v12 s – v11 w11
where
swvvvvvvt
uuuuuupsw
vvvvvvpsr
2
4
4
1116151413121111
6151413121111111
1615141312111111
2. MAIN RESULTS
We will prove several statements about the nullspace of the mentioned nested magic squares. By doing this we will
use mathematical induction. Hence, some results will be used for proving others. We start with
Proposition 1: The nested semi magic square 6 by 6 with a pandiagonal magic square 4 by 4 has the following
space
as a subspace of the nullspace of this square.
Proof: We search for real numbers 61 ,,... aa such that
0... 6611 CaCa
where the capital letters denote the columns of the matrix. We set
61 aa and 1saq . We set
RRz
CA
CBAs
AC
sCBA
z
1
1
1
1
1
1
1
},:
2
222
2
222
{
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
317
)2(.......................
1
1
1
1
222
22
5
4
3
2
s
q
CA
CBAs
AC
sCBA
z
a
a
a
a
where z is a unknown real number. Now, the expression
6611 ... CaCa
is the result of matrix multiplication of the square with the vector
6
5
4
3
2
1
a
a
a
a
a
a
The result of this operation will be according to our choice a vector which has the following middle entries
1411
1311
1211
1111
1
1
1
1
141
131
121
111
apsa
apsa
apsa
apsa
sa
sa
sa
sa
ap
ap
ap
ap
Hence, the middle entries are zero and we have to equate the first and last entry of the vector to zero. We will use
this requirement to determine the values of 1a and z. Thus, we solve the following two equations:
0111514413312211111 adababababar
0)()()()( 111514413312211111 afabsabsabsabsae
This system can be simplified to
)3(...............0)( 51441331221111111 ababababadr
0)()( 514413312211543211111 ababababaaaasafe
According to the definition of the variables (see (1) and (2)) we have
141312111111 3 bbbbsdr
sbbbbfe 141312111111
15432 22
as
qaaaa
Note that we used the sign properties of the vectors belonging to the nullspace of the 4 by 4 square. Upon using the
last relations we convert the system (3) into
0)3( 514413312211114131211 abababababbbbs
0)3( 514413312211114131211 ababababasbbbb
We recognize that one equation is redundant. We substitute the values of 2a , 3a , 4a and 5a . Then, we obtain
from the last equation
0 ))()22()()22(()2(2
314131211114131211 zCAbCBAsbACbsCBAbabbbbs
If the coefficient of 1a is not zero, then we conclude that
za 11
Upon substituting this value for 1a we are done with the proof. If the coefficient of 1a is zero, then we conclude
that 1a can be any number, while z might be zero. In this case we can say
Ra u ,1 ,u 111
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
318
and the proof is done with u instead of z.
Remark: 1) We note that the sum of all entries of any vector in the subspace is zero.
2) When we consider the following numerical example for a nested semi magic square 6 by 6
–26 8 1 4 7 9
– 4 –3 0 3 2 5
–1 –3 8 –9 6 2
1 –2 –1 4 1 0
–2 10 –5 4 –7 3
35 –7 0 –3 –6 –16
We obtain the following reduced row echelon form
00000011
810000
11
4601000
11
1900100
11
3500010
100001
Hence, the nullity of the square is one. Therefore, the described subspace in proposition 1 is actually the nullspace of
the square.
Proposition 2: The multi-nested semi magic square 2n by 2n with a pandiagonal magic square 4 by 4 has the
following space
}:
2
1
23
1
2
2
34
.
.
.
1
2...4
3
3
21
1
2...2
1
1
2...2
122
1
2...2
1
1
2...2
122
1
2...4
3
3
21
.
.
.
1
2
2
34
1
23
2
{ Rz
n
n
nn
n
n
n
nn
n
n
n
nCA
n
nCBAs
n
nAC
n
nsCBA
n
n
n
n
n
nn
n
nn
n
z
for all 3n as a subspace of the nullspace of this square.
Proof: We will use mathematical induction in our proof. We start with
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
319
Basis step: when 3 n we obtain the semi magic square 6 by 6, which has according to proposition 1 a subspace
as claimed.
We continue now with
Induction step: we suppose that the given form of the nullspace is true for kn , i. e. there exists a subspace of
the nullspace of any multi-nested square kk 2by 2 with a pandiagonal magic square 4 by 4, which has the
following structure
Rz
k
k
kk
k
k
k
kk
k
k
k
kCA
k
kCBAs
k
kAC
k
ksCBA
k
k
k
k
k
kk
k
kk
k
z
kx
kx
kx
x
x
,
2
1
23
1
2
2
34
.
.
.
1
2...4
3
3
21
1
2...2
1
1
2...2
122
1
2...2
1
1
2...2
122
1
2...4
3
3
21
.
.
.
1
2
2
34
1
23
2
2
.
.
.
1
.
.
.
2
1
We construct now a subspace of the nullspace of the multi-nested square 22by 22 kk . We search for
221 ,,... kaa such that the following relation holds
0... 222211 kk CaCa
where the capital letters denote the columns of the square. We choose 221 kaa , 1saq . We choose further
)4(.......
1
1
1
.
.
.
1
1
1
2
.
.
.
1
.
.
.
2
1
12
.
.
.
4
3
2
ks
q
kx
kx
kx
x
x
z
ka
a
a
a
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
320
Since the statement of the proposition holds for kn , we conclude that the multiplication of the multi-nested
kk 2by 2 square inside the multi-nested square 22by 22 kk by this vector yields a vector, where each
entry is equal to q . Hence, the multiplication of the multi-nested square 22by 22 kk by
22
.
.
.
3
2
1
ka
a
a
a
yields
Therefore, all the entries are zero except the first and last entry. We will use this requirement to determine z and 1a
. We obtain therefore the following equations
0...
...
1)1)(1(41)1(34)1(
23)1(12)1(1)1(11)1(2)1(1)1)(1(
adadab
abababararar
mmkmkm
kmkmkmkmmmmm
0...)()()(
)()()(...)(
1)1)(1(41)1(34)1(23)1(
12)1(1)1(11)1(2)1(1)1)(1(
afadsabsabs
absabsarsarsae
mmkmkmkm
kmkmkmmmmm
This linear system can be simplified to
0...
...)(
12)1(41)1(34)1(
23)1(12)1(1)1(11)1(2)1(1)1)(1()1)(1(
kmmkmkm
kmkmkmkmmmmmmm
adadab
abababararadr
1)1(1
.
.
.
1)1(21
1)1(11
1)1(41
1)1(31
1)1(21
1)1(11
1)1(11
1)1(21
.
.
.
1)1)(1(1
1)1(1
1
.
.
.
1
1
1
1
1
1
1
1
.
.
.
1
1
1)1(
.
.
.
1)1(2
1)1(1
1)1(4
1)1(3
1)1(2
1)1(1
1)1(1
1)1(2
.
.
.
1)1)(1(
1)1(
amm
esa
am
esa
am
esa
am
psa
am
psa
am
psa
am
psa
am
rsa
am
rsa
amm
rsa
amm
rsa
sa
sa
sa
sa
sa
sa
sa
sa
sa
sa
sa
amm
e
am
e
am
e
am
p
am
p
am
p
am
p
am
r
am
r
amm
r
amm
r
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
321
0...
...)...()(
12)1(41)1(34)1(23)1(12)1(
1)1(11)1(2)1(12321)1)(1()1)(1(
kmmkmkmkmkm
kmkmmmkmmmm
adadababab
abararaaasafe
According to the definition of our variables (see (4)) we have
mmmmmm
mmmmmmmmmm
ddbbb
brrrskdr
)1(1)1(4)1(3)1(2)1(
1)1(1)1()1)(1()1()1)(1()1)(1(
...
...)1(
skdd
bbbbrrfe
mmm
mmmmmmmmmmm
)1(...
...
)1(1)1(
4)1(3)1(2)1(1)1(1)1()1()1)(1()1)(1(
11232 22
... as
qaaa k
According to the definition of )1)(1( mmf and )1)(1( mmd we obtain the following linear system
0...
...
)......)1((
12)1(41)1(
34)1(23)1(12)1(1)1(11)1(2)1(
1)1(1)1(4)1(3)1(2)1(1)1(1)1()1(
kmmkm
kmkmkmkmkmmm
mmmmmmmmmm
adad
ababababarar
addbbbbrrsk
0...
...
))1(......(
12)1(41)1(
34)1(23)1(12)1(1)1(11)1(2)1(
1)1(1)1(4)1(3)1(2)1(1)1(1)1()1(
kmmkm
kmkmkmkmkmmm
mmmmmmmmmm
adad
ababababarar
askddbbbbrr
We recognize that one equation is redundant. We substitute the values of 2a , …, 12 ka and obtain the following
equation:
0)...
...(
)......(1
2)1(31)1(
24)1(13)1(2)1(11)1(21)1(1)1(
1)1(1)1(4)1(3)1(2)1(1)1(1)1()1(
zxdxd
xbxbxbxbxrxr
addbbbbrrksk
k
kmmkm
kmkmkmkmkmmm
mmmmmmmmmm
We can say using a similar argument to the one used in the proof of proposition 1
za k 11
In analogy to that proof we are done.
Remark: We note that the sum of all entries of any vector in the subspace is zero.
We turn our attention now to the squares, which have a pandiagonal magic square 6 by 6 of the special structure
in the centre. We can prove very similar results for this type of squares like the results for multi-nested squares with
4 by 4 square in the centre.
Proposition 3: The nested semi magic square 8 by 8 with a 6 by 6 pandiagonal square has the following space
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
322
} :
1
3
13
3
12
3
11
3
13
3
12
3
11
1
{ Rz
x
x
x
x
x
x
z
as a subspace of the nullspace of the square.
Proof: We search for real numbers 81 ,,... aa such that
0... 8811 CaCa
where the capital letters denote the columns of the matrix. We set 81 aa
and 1saq . We set
Rzs
q
x
x
x
x
x
x
z
a
a
a
a
a
a
,
1
1
1
1
1
1
3
3
2
1
3
2
1
7
6
5
4
3
2
where z is an unknown real number. Now, the expression
8811 ... CaCa
is the result of matrix multiplication of the square with the vector
8
7
6
5
4
3
2
1
a
a
a
a
a
a
a
a
The result of this operation will be according to our choice a vector which has the following middle entries
1611
1511
1411
1311
1211
1111
1
1
1
1
1
1
161
151
141
131
121
111
ausa
ausa
ausa
ausa
ausa
ausa
sa
sa
sa
sa
sa
sa
au
au
au
au
au
au
Hence, the middle entries are zero and we have to equate the first and last entry of the vector to zero. We will use
this requirement to determine the values of 1a and z. Thus, we solve the following two equations:
0111716615514413312211111 aravavavavavavap
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
323
01117
)16
(6
)15
(
5)
14(
4)
13(
3)
12(
2)
11(
111
atavsavs
avsavsavsavsaw
This system can be simplified to
0)( 71661551441331221111111 avavavavavavarp
0716615514413
312211)
765432(
1)
1111(
avavavav
avavaaaaaasatw
According to the definition of the variables we have
1615141312111111 4 vvvvvvsrp
svvvvvvtw 21615141312111111
12
2
765432a
s
qaaaaaa
Note that we used the sign properties of the vectors belonging to the nullspace of the 6 by 6 square. Upon using the
last relations we convert the system into
0)4( 7166155144133122111161514131211 avavavavavavavvvvvvs
0)4( 7166155144133122111161514131211 avavavavavavasvvvvvv
We recognize that one equation is redundant. We substitute the values of 2a , 3a , 4a , 5a , 6a and 7a . Then, we
obtain from the last equation
0)()3(3
43162151143132121111161514131211 zxvxvxvxvxvxvavvvvvvs
If the coefficient of 1a is not zero, then we conclude that
za 11
Upon substituting this value for 1a we are done with the proof. If the coefficient of 1a is zero, then we conclude
that 1a can be any number, while z might be zero. In this case we can say
Ra u ,1 ,u 111
and we have the proof done with u instead of z.
Remark: 1) We note that the sum of all entries of any vector in the subspace is zero.
2) When we consider the following numerical example for a nested semi magic square 8 by 8
–17 7 3 1 0 4 2 8
– 4 –2 1 4 0 2 1 6
0 4 0 1 1 0 0 2
–3 0 4 –3 1 3 1 5
–1 2 0 1 4 1 –2 3
2 1 2 2 –2 2 1 0
1 1 –1 1 2 –2 5 1
30 –5 –1 1 2 –2 0 –17
We obtain the following reduced row echelon form
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
324
0000000063
231000000
63
5050100000
7
170010000
63
650001000
63
4630000100
21
370000010
10000001
Hence, the nullity of the square is one. Therefore, the described subspace in proposition 3 is actually the nullspace of
the square.
Proposition 4: The nested semi magic square 2n by 2n with a 6 by 6 pandiagonal square has the following space
}:
3
1
34
1
3
2
45
.
.
.
1
3...5
3
4
21
1
3...4
2
3
13
.
1
3...4
2
3
12
1
3...4
2
3
11
1
3...4
2
3
13
1
3...4
2
3
12
1
3...4
2
3
11
1
3...5
3
4
21
.
.
.
1
3
2
45
1
34
3
{ Rz
n
n
nn
n
n
n
nn
n
n
n
nx
n
nx
n
nx
n
nx
n
nx
n
nx
n
n
n
n
n
nn
n
nn
n
z
as a subspace of the nullspace of the square for all .4n
Proof: We will use mathematical induction in our proof. We start with
Basis step: when 4n we obtain the semi magic square 8 by 8, which has according to proposition 3 a subspace
as claimed.
We continue now with
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
325
Induction step: we suppose that the given form of the nullspace is true for kn , i.e. there exists a subspace of
the nullspace of the multi-nested square kk 2by 2 , which has the following structure
Rz
k
k
kk
k
k
k
kk
k
k
k
kx
k
kx
k
kx
k
kx
k
kx
k
kx
k
k
k
k
k
kk
k
kk
k
z
ky
ky
ky
y
y
,
3
1
34
1
3
2
45
.
.
.
1
3...5
3
4
21
1
3...4
2
3
13
.
1
3...4
2
3
12
1
3...4
2
3
11
1
3...4
2
3
13
1
3...4
2
3
12
1
3...4
2
3
11
1
3...5
3
4
21
.
.
.
1
3
2
45
1
34
3
2
.
.
.
1
.
.
.
2
1
We construct now a subspace of the nullspace of the multi-nested square 22by 22 kk . We search for
221 ,,... kaa such that the following relation holds
0... 222211 kk CaCa
where the capital letters denote the columns of the square. We choose 221 kaa , 1saq . We choose further
1
1
1
.
.
.
1
1
1
2
.
.
.
1
.
.
.
2
1
12
.
.
.
4
3
2
ks
q
ky
ky
ky
y
y
z
ka
a
a
a
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
326
Since the statement of the proposition holds for kn we conclude that the multiplication of the multi-
nested kk 2by 2 square inside the multi-nested square 22by 22 kk by this vector yields a vector, where
each entry is equal to q . Hence, the multiplication of the multi-nested square 22by 22 kk by
)
22
.
.
.
3
2
1
ka
a
a
a
yields
1)1(1
1)1)(1(1
1)1(11
1)1(61
1)1(51
1)1(41
1)1(31
1)1(21
1)1(11
1)1(11
1)1)(1(1
1)1(1
1
1
1
1
1
1
1
1
1
1
1
1
1)1(
1)1)(1(
1)1(1
1)1(6
1)1(5
1)1(4
1)1(3
1)1(2
1)1(1
1)1(1
1)1)(1(
1)1(
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
awsa
awsa
awsa
ausa
ausa
ausa
ausa
ausa
ausa
apsa
apsa
apsa
sa
sa
sa
sa
sa
sa
sa
sa
sa
sa
sa
sa
aw
aw
aw
au
au
au
au
au
au
ap
ap
ap
mm
mm
m
m
m
m
m
m
m
m
mm
mm
mm
mm
m
m
m
m
m
m
m
m
mm
mm
All the entries are zero except the first and last entry. We will use this requirement to determine z and 1a . We
obtain therefore the following equations
0...
...
1)1)(1(51)1(46)1(35)1(24)1(
13)1(2)1(11)1(21)1(2)1(1)1)(1(
araravavav
avavavapapap
mmkmkmkmkm
kmkmkmkmmmmm
0...)(
)()()()(
)()()(...)(
1)1)(1(51)1(
46)1(35)1(24)1(13)1(
2)1(11)1(21)1(2)1(1)1)(1(
atars
avsavsavsavs
avsavsapsapsaw
mmkm
kmkmkmkm
kmkmkmmmmm
This linear system can be simplified to
0...
...)(
12)1(51)1(46)1(35)1(24)1(
13)1(2)1(11)1(21)1(2)1(1)1)(1()1)(1(
kmmkmkmkmkm
kmkmkmkmmmmmmm
araravavav
avavavapaparp
IJRRAS 8 (3) ● September 2011 ElShokry & Al-Ashhab ● The Nullspace of Nested Magic Squares
327
0...
...)...()(
12)1(
51)1(46)1(35)1(24)1(13)1(2)1(
11)1(21)1(2)1(12321)1)(1()1)(1(
kmm
kmkmkmkmkmkm
kmkmmmkmmmm
ar
aravavavavav
avapapaaasatw
According to the definition of our variables we have
mmmmmmmm
mmmmmmmmmm
rrvvvvv
vpppskrp
)1(1)1(6)1(5)1(4)1(3)1(2)1(
1)1(1)1()1)(1()1()1)(1()1)(1(
...
...)1(
skrrvv
vvvvpptw
mmmmm
mmmmmmmmmmm
)1(...
...
)1(1)1(6)1(5)1(
4)1(3)1(2)1(1)1(1)1()1()1)(1()1)(1(
11232 22
... as
qaaa k
According to the definition of )1)(1( mmr and )1)(1( mmt we obtain the following linear system
0...
...)...
...)1((
12)1(51)1(46)1(35)1(24)1(
13)1(2)1(11)1(21)1(2)1(1)1(1)1(
6)1(5)1(4)1(3)1(2)1(1)1(1)1()1(
kmmkmkmkmkm
kmkmkmkmmmmmm
mmmmmmmmm
araravavav
avavavapaparr
vvvvvvppsk
0...
...))1(...
...(
12)1(51)1(46)1(35)1(24)1(13)1(
2)1(11)1(21)1(2)1(1)1(
1)1(6)1(5)1(4)1(3)1(2)1(1)1(1)1()1(
kmmkmkmkmkmkm
kmkmkmmmmm
mmmmmmmmmm
araravavavav
avavapapaskr
rvvvvvvpp
We recognize that one equation is redundant. We substitute the values of 2a , …, 12 ka and obtain the following
equation
0)...
...()...
...(1
2)1(41)1(36)1(25)1(14)1(
3)1(12)1(21)1(31)1(1)1(1)1(1)1(
6)1(5)1(4)1(3)1(2)1(1)1(1)1()1(
zyryryvyvyv
yvyvyvypyparr
vvvvvvppksk
k
kmmkmkmkmkm
kmkmkmkmmmmmm
mmmmmmmmm
We can say using a similar argument to the one in the proof of proposition 3
za k 11
In analogy to that proof we are done.
ACKNOWLEDGMENT
This paper is extracted from a Msc. Thesis by Ayoub ElShokry (supervised by Saleem Al-Ashhab) at Al-Albayt
University in 2007
REFERENCES
[1]. Al-Ashhab, S., Computer Solutions of Magic 44 Squares and Semi – Magic 44 Squares Problem , Dirasat,
Natural and Engineering Sciences, Vol. , 25 , No , 3 , 1998 , pp . (445-450).
[2]. Al-Ashhab, S., Theory of Magic Squares / Programming and Mathematics, Royal Scientific Society, Jordan,
2000 (in Arabic).
[3]. Anton, H., Elementary Linear Algebra, 7th
edition, Jon Wiley, New York, 1994.
[4]. Rosser, B. & Walker, R., J., On the Transformation Group for Diabolic Magic Squares of Order Four, Amer.
Math. Soc., Bult XLIV, 1938, pp. 416-420.
[5]. Trenkler, D. & Trenkler, G., Magic Squares, Melancholy and The Moore-Penrose Inverse, The Bulletin of
The International Linear Algebra Society, No.27 , 2001, pp (3-10).