28 3. Intermediate Determinant Theory
Substituting the first line of (3.3.9 and the second line of (3.3.8),
A=
n
∑
i 1 =1
n
∑
i 2 =1
ai
1 j 1
ai
2 j 2
∂
2
A
∂ai
1 j 1
∂ai
2 j 2
=
n
∑
i 1 =1
n
∑
i 2 =1
ai 1 j 1 ai 2 j 2 Ai 1 i 2 ;j 1 j 2 ,i 1 <i 2 andj 1 <j 2. (3.3.10)
Continuing in this way and applying (3.3.7) in reverse,
A=
n
∑
i 1 =1
n
∑
i 2 =1
···
n
∑
ir=1
ai
1 j 1
ai
2 j 2
···ai
rjr
∂
r
A
∂ai
1 j 1
∂ai
2 j 2
···∂ai
rjr
=
n
∑
i 1 =1
n
∑
i 2 =1
···
n
∑
ir=1
ai 1 j 1 ai 2 j 2 ···airjrAi 1 i 2 ...ir;j 1 j 2 ...jr, (3.3.11)
subject to the inequalities associated with (3.3.7) which require that theis
andjsshall be in ascending order of magnitude.
In this multiple sum, thoserth cofactors in which the dummy variables
are not distinct are zero so that the corresponding terms in the sum are
zero. The remaining terms can be divided into a number of groups according
to the relative magnitudes of the dummies. Sincerdistinct dummies can
be arranged in a linear sequence inr! ways, the number of groups isr!.
Hence,
A=
(r! terms)
∑
Gk 1 k 2 ...,kr,
where
Gk 1 k 2 ...kr=
∑
i≤ik 1 <ik 2 <···<ikr≤n
aik
1
jk
1
aik
2
jk
2
···aik
rjkr
Aik
1
ik
2
···ikr;jk
1
jk
2
...jkr. (3.3.12)
In one of theser! terms, the dummiesi 1 ,i 2 ,...,irare in ascending order
of magnitude, that is,is<is+1,1≤s≤r−1. However, the dummies
in the other (r!−1) terms can be interchanged in such a way that the
inequalities are valid for those terms too. Hence, applying those properties
ofrth cofactors which concern changes in sign,
A=
∑
1 ≤i 1 <i 2 <···<ir≤n
[∑
σrai 1 j 1 ai 2 j 2 ···airjr
]
Ai 1 i 2 ...ir;j 1 j 2 ...jr,
where
σr= sgn
{
123 ··· r
i 1 i 2 i 3 ··· ir
}
. (3.3.13)
(Appendix A.2). But,
∑
σrai 1 j 1 ai 2 j 2 ···airjr=Ni 1 i 2 ...ir;j 1 j 2 ...jr.