8.17 QUADRATIC AND HERMITIAN FORMS
i.e.Qis unchanged by considering only the symmetric part ofM. Hence, with no
loss of generality, we may assumeA=ATin (8.106).
From its definition (8.105),Qis clearly a basis- (i.e. coordinate-) independent
quantity. Let us therefore consider a new basis related to the old one by an
orthogonal transformation matrixS, the components in the two bases of any
vectorxbeing related (as in (8.91)) byx=Sx′or, equivalently, byx′=S−^1 x=
STx. We then have
Q=xTAx=(x′)TSTASx′=(x′)TA′x′,
where (as expected) the matrix describing the linear operatorA in the new
basis is given byA′=STAS(sinceST=S−^1 ). But, from the last section, if we
choose asSthe matrix whose columns are thenormalisedeigenvectors ofAthen
A′=STASis diagonal with the eigenvalues ofAas the diagonal elements. (Since
Ais symmetric, its normalised eigenvectors are orthogonal, or can be made so,
and henceSis orthogonal withS−^1 =ST.)
In the new basis
Q=xTAx=(x′)TΛx′=λ 1 x′ 1
2
+λ 2 x′ 2
2
+···+λNx′N
2
, (8.110)
where Λ = diag(λ 1 ,λ 2 ,...,λN)andtheλiare the eigenvalues ofA. It should be
noted thatQcontains no cross-terms of the formx′ 1 x′ 2.
Find an orthogonal transformation that takes the quadratic form (8.107) into the form
λ 1 x′ 12 +λ 2 x′ 22 +λ 3 x′ 32.
The required transformation matrixShas thenormalisedeigenvectors ofAas its columns.
We have already found these in section 8.14, and so we can write immediately
S=
1
√
6
√
3
√
√^21
3 −
√
2 − 1
0
√
2 − 2
,
which is easily verified as being orthogonal. Since the eigenvalues ofAareλ=2,3,and
−6, the general result already proved shows that the transformationx=Sx′will carry
(8.107) into the form 2x′ 12 +3x′ 22 − 6 x′ 32 .This may be verified most easily by writing out
the inverse transformationx′=S−^1 x=STxand substituting. The inverse equations are
x′ 1 =(x 1 +x 2 )/
√
2 ,
x′ 2 =(x 1 −x 2 +x 3 )/
√
3 , (8.111)
x′ 3 =(x 1 −x 2 − 2 x 3 )/
√
6.
If these are substituted into the formQ=2x′ 12 +3x′ 22 − 6 x′ 32 then the original expression
(8.107) is recovered.
In the definition ofQit was assumed that the componentsx 1 ,x 2 ,x 3 and the
matrixAwere real. It is clear that in this case the quadratic formQ≡xTAxis real