548 Chapter 19The matrix eigenvalue problem
The canonical form
We have seen (equation (19.30)) that a symmetric matrix Ais reduced to diagonal
form by the similarity transformationX
TAXwhere Xis the orthogonal matrix whose
columns are the orthonormal eigenvectors of A. BecauseXX
T1 = 1 I(for orthogonal
matrix X), we can write (19.38) as
Q 1 = 1 x
T(XX
T)A(XX
T)x
= 1 (x
TX) (X
TAX) (X
Tx)
= 1 y
TDy (19.40)
whereD 1 = 1 X
TAXis the diagonal matrix of the eigenvalues of A, and
y 1 = 1 X
Tx (19.41)
is the vector obtained from xby the orthogonal transformationX
T.
The quadratic form Qin the nvariablesx
1, x
2, x
3, =, x
nhas been transformed into
an equivalent form in the nvariablesy
1, y
2, y
3, =, y
nthat contains only pure square
terms:
(19.42)
This is the canonical formof Q, and the variables y
kare the canonical variables.
3EXAMPLE 19.14Transform the following quadratic form into canonical form:
Q 1 = 15 x
121 + 18 x
1x
21 + 15 x
22We have
and the orthonormal eigenvectors of the symmetric matrix Aare
xx
121
2
1
1
1
2
1
1
=
−
=
,
Qxx
x
x
=
=
()
121254
45
xAx
TQyyyyy
knkk nn()y ==++++
=∑
121122223322λλλλ λ
3Cayley and Sylvester developed the theory of forms between 1854 and 1878. Sylvester claimed that he
discovered and developed the reduction of a quadratic form to canonical form at one sitting ‘with a decanter of
port wine to sustain nature’s flagging energies’. A general description of canonical forms was given by Camille
Jordan (1838–1922) in his Traité des substitutions et des équations algébriques(Treatise on substitutions and
algebraic equations) of 1871, in which he presented many of the modern concepts of group theory within the
context of groups of permutations (substitutions).