4.8 Hankelians 1 105
It follows that
aji=aij,
so that Hankel determinants are symmetric, but it also follows that
ai+k,j−k=aij,k=± 1 ,± 2 ,.... (4.8.2)
In view of this additional property, Hankel determinants are described as
persymmetric. They may also be called Hankelians.
A single-suffix notation has an advantage over the usual double-suffix
notation in some applications.
Put
aij=φi+j− 2. (4.8.3)
Then,
An=
∣ ∣ ∣ ∣ ∣ ∣ ∣ ∣ ∣
φ 0 φ 1 φ 2 ··· φn− 1
φ 1 φ 2 φ 3 ··· φn
φ 2 φ 3 φ 4 ··· φn+1
.............................
φn− 1 φn φn+1 ··· φ 2 n− 2
∣ ∣ ∣ ∣ ∣ ∣ ∣ ∣ ∣ n
, (4.8.4)
which may be abbreviated to
An=|φm|n, 0 ≤m≤ 2 n− 2. (4.8.5)
In column vector notation,
An=
∣
∣
C 0 C 1 C 2 ···Cn− 1
∣
∣
n
,
where
Cj=
[
φjφj+1φj+2···φj+n− 1
]T
, 0 ≤j≤n− 1. (4.8.6)
The cofactors satisfyAji=Aij, butAij=F(i+j) in general, that is,
adjAis symmetric but not Hankelian except possibly in special cases.
The elementsφ 2 andφ 2 n− 4 each appear in three positions inAn. Hence,
the cofactor
∣ ∣ ∣ ∣ ∣ ∣ ∣
φ 2 ··· φn− 1
.
.
.
.
.
.
φn− 1 ··· φ 2 n− 4
∣ ∣ ∣ ∣ ∣ ∣ ∣
(4.8.7)
also appears in three positions inAn, which yields the identities
A
(n)
12;n− 1 ,n
=A
(n)
1 n, 1 n
=A
(n)
n− 1 ,n;12
.
Similarly
A
(n)
123;n− 2 ,n− 1 ,n
=A
(n)
12 n;1,n,n− 1
=A
(n)
1 n,n−1;12n
=A
(n)
n− 2 ,n− 1 ,n;123