186 5. Further Determinant Theory
=−
[
2 n− 1
∑
i=1
(−1)
i+1
Pf
(n)
i ai,^2 n
]
2 n− 1
∑
j=1
(−1)
j+1
Pf
(n)
j φj
=−PfnHnKn− 1. (5.2.27)
Part (c) now follows from Theorem 5.1 and (d) is proved in a similar
manner.
LetR(φ) denote the row vector defined as
R(φ)=
[
φ 1 φ 2 φ 3 ···φ 2 n− 1 •
]
and letB 2 n(φ, ψ) denote the determinant of order 2nwhich is obtained
from|aij| 2 nby replacing the last row by−R(φ) and replacing the last
column byR
T
(ψ).
Theorem 5.7.
B 2 n(φ, ψ)=Hn− 1 HnKn− 1 Kn.
Proof.
B 2 n(φ, ψ)=
2 n− 1
∑
i=1
2 n− 1
∑
j=1
ψiφjA
(2n−1)
ij
.
The theorem now follows (5.2.13), (5.2.24), and (5.2.25).
Theorem 5.8.
B 2 n(φ, ψ)=A
(2n)
2 n− 1 , 2 n.
Proof. Applying the Jacobi identity (Section 3.6),
∣
∣
∣
∣
∣
A
(2n)
2 n− 1 , 2 n− 1 A
(2n)
2 n− 1 , 2 n
A
(2n)
2 n, 2 n− 1
A
(2n)
2 n, 2 n
∣
∣
∣
∣
∣
=A 2 nA
(2n)
2 n− 1 , 2 n;2n− 1 , 2 n
. (5.2.28)
But,A
(2n)
ii ,i =2n−1, 2n, are skew-symmetric of odd order and are
therefore zero. The other two first cofactors are equal in magnitude but
opposite in sign. Hence,
(
A
(2n)
2 n− 1 , 2 n
) 2
=A 2 nA 2 n− 2 ,
A
(2n)
2 n− 1 , 2 n
=PfnPfn− 1. (5.2.29)
Theorem 5.8 now follows from Theorems 5.1 and 5.7.
Ifψr =φr, thenKn =Hn and Theorems 5.1, 5.6a and c, and 5.7
degenerate into identities published in a different notation by Cusick,
namely,
A 2 n=H
4
n
,
B 2 n− 1 (φ)=H
3
n− 1
Hn,