Let A be a real symmetric matrix. The eigenvalues of A are real and there exists a
real orthogonal matrix S such thatS −1 AS = S T AS is a diagonal matrix, say D. The
diagonalization is done by applying a series of orthogonal transformations S1 , S2 , ..., Sn ,
... as follows:
• Among the o-diagonal elements let aik be the numerically largest element in mag-
nitude.
• Then we form a 2 × 2 sub matrix A1 with the elements aii , aik , aki , akk
aii aik
A1 = (1)
aki akk
• Using the orthogonal matrix S1∗ , we nd θ such that the matrix A1 is diagonalized
by performing the rotation S1∗ −1 AS1∗ as follows:
∗ −1 ∗ ∗T ∗ cos θ sin θ aii aik cos θ − sin θ
S1 AS1 = S1 AS1 =
− sin θ cos θ aki akk sin θ cos θ
aii cos2 θ + aik sin 2θ + akk sin2 θ (akk − aii ) sin θ cos θ + aik cos 2θ
=
(akk − aii ) sin θ cos θ + aik cos 2θ aii sin2 θ − aik sin 2θ + akk cos2 θ
For the above matrix to be of diagonal form we put
(akk − aii ) sin θ cos θ + aik cos 2θ = 0
i.e.,
2aik
tan 2θ = (3)
aii − akk
or
1 2aik
θ = tan−1
(4)
2 aii − akk
• With this value of θ the o-diagonal elements of S1∗ T AS1∗ vanish. Hence, we perform
the rotation S1T AS1 on the matrix A with the matrix
1 0 · · · · · · · 0
0 1 0 · · · · · · ·
· · · · · · · · · ·
· · · · cos θ · · − sin θ · ·
· · · · · · · · · ·
(5)
S1 = · · · · · · 1 · ·
· · · · · · · · · ·
· · · · sin θ · · cos θ · ·
· · · · · · · · · ·
· · · · · · · 0 1 0
0 · · · · · · · 0 1
where cos θ, − sin θ, sin θ, cos θ are located in the (i, i),(i, k), (k, i), (k, k) positions
respectively.
With this rotation the rst step is completed and we get the new rotated matrix
B1 = S1T AS1 .
• In the next step, again we choose the largest o-diagonal element in magnitude in
the new rotated matrix B1 and the procedure is repeated to get the matrices
B2 = S2−1 B1 S2 = S2−1 S1−1 AS1 S2 ,
B3 = S3−1 B2 S3 = S3−1 S2−1 S1−1 AS1 S2 S3 ,
..
.
Br = Sr−1 Br−1 Sr = Sr−1 Sr−1
−1
· · · S2−1 S1−1 AS1 S2 · · · Sr−1 Sr
= (S1 S2 · · · Sr−1 Sr )−1 A (S1 S2 · · · Sr−1 Sr )
√1 0 − √12
2
• S1 = 0 1 0
√1 0 √12
2
√
√1 √1
1
0 0 − √12
√
2 2 √12 √2 2
B1 = S1−1 AS1 = S1T AS1 = 0 1 0 2 √3 2 0 1 0
− √12 0 √1
2 2 2 1 √1
2
0 √12
•
3 2 0
= 2 3 0
0 0 −1
With this the rst rotation completes and we see that zero is created in the a13 and
a31 positions of B1 .
• Again the largest o-diagonal element in B1 is a12 = a21 = 2.
3 2
• The 2 × 2 submatrix with elements a11 , a12 , a21 , a22 . i.e.,
2 3
• θ = 12 tan−1 3−3
2×2
= 21 tan−1 04 = π/4
1
− √12 0
√
2
• S2 = √1 2
√1
2
0
0 0 1
√1 √1
1
0 3 2 0 − √1 0
√
2 2 2 2
B2 = S2−1 B1 S2 = S2T AS2 = − √1
2
√1
2
0 2 3 0 √12 √12 0
0 0 1 0 0 −1 0 0 1
•
5 0 0
= 0 1 0
0 0 −1
With this the second rotation completes and the matrix is diagonalized.
• We have the matrix of eigenvectors as
1
0 − √12
1
− √12 0
√ √
2 2
S = S1 S2 = 0 1 0 √12 √12 0
√1 0 √12 0 0 1
2
1
2
− 21 − √12
√1 √1 0
= 2 2
1
2
− 21 √12
• Hence the eigenvalues are the diagonal elements of the diagonal matrix B2 i.e.,
5, −1, 1 and the corresponding eigenvectors are the columns of S .
Exericse. Find all the eigenvalues and the corresponding eigenvectors of the following
matrices using Jacobi method.
3 2 2
1. A = 2 5 2
2 2 3
√ √
Answer: Eigenvalues
are
λ = 1,5 − 2, 5 + 2 2,
2
−1 1
√ √1
Eigenvectors are 0 , − 2 , 2
1 1 1
1 −2 4
2. A = −2 5 −2
4 −2 1
√ √
Answer: Eigenvalues
are −3,5 − 2 2, 5 + 2 2
λ =
−1 √1 1
√
Eigenvectors are 0 , 2 , − 2
1 1 1
√
√1 3 √4
3. A = 3 √5 3
4 3 1
√ √
Answer: Eigenvalues
are λ = −3, 5
− 6, 5 + 6
1 √1 √ 1
Eigenvectors are 0 , − 2 , 2
1 1 1