MAT2002 - ADDE Dr. T.

Phaneendra
Module 2 Professor of Mathematics
Quadratic Forms: A homogeneous polynomial of second degree of the form
n n
  (x1, x2, ..., xn)   i 1  j 1 aij xi x j , … (1)
where aij’s are real constants is known as a real quadratic form or simply a quadratic form
(abbreviated as qf) in n (real) variables x1, x2, …, xn.

Symmetric matrix: corresponding to the qf (1) is A  [bij ]nn , where

 aij a ji if i j
bij   2 for all i and all j. … (2)
 aii if i j
 x1 
 
The qf (1) can be written as   X AX , where X   x 2  is the vector of unknowns x1, x2, …, xn.
T

 
 xn 
Example 1. If   ( x1 , x2 , .x3 )  x1  2x1 x2  3x2  x2 x3  5x32 , we can write
2 2

 1.x1 x1  1.x1 x2  0.x1 x3 1 1 0 x1 

   
  ( x1 , x2 , .x3 )   1.x2 x1  3.x2 x2  12 x2 x3  x1 x2 x3 1  3 12  x2   X T AX .
 0.x3 x1  1 x3 x2  5.x3 x3  0 1 5  
 2  2  x n 
1 1 0
 
So the corresponding symmetric matrix is A  1  3 12  .
 0 1 5
 2 
 0 12 3 
1 2 
Example 2. The symmetric matrix of the qf   x1x2  2x2x3  3x1x3 is A   2 0  1  .
3 1 0 
2 

Canonical form: Sum of squares form of qf (1).

Orthogonal reduction: Suppose we wish to reduce the qf (1) to a sum of squares form. We
use the following procedure:
Step 1. Find the symmetric matrix A of (1).
Step 2. Obtain the eigenvalues 1, 2, …, n of A.
Step 3. Find the corresponding eigenvectors X1, X2, …, Xn.
Step 4. Find an orthonormal set of eigenvectors of A.
Step 5. The modal matrix P  [e1 e2 ... en ] is orthogonal and P 1  PT .
Step 6. The diagonal form of A is D  P 1 AP  PT AP  diag [1 , 2 ,..., n]
 x1   y1 
x   
Step 7. The orthogonal transformation is X  PY , where X    and Y   y2  .
2
   
 xn   yn 
Step 8. With X  PY , we see that
  X T AX  (PY )T A(PY )  Y T (PT AP )Y  Y T DY .
or Q  1 y12  2 y22      n yn2 , … (5)
which is the required canonical form of the qf (1).

Rank, index and signature of a quadratic form: Consider the qf (1).

Rank The number of nonzero terms in the canonical form (5) is the rank (r) of (1).
Index The number of positive terms in the canonical form (5) is the index (p)of (1).
Signature Excess/surplus number of positive terms over the negative terms in the
canonical form (5) is the signature (s) of the qf (1), that is
s = p – (r – p) = 2p – r.

SJT 511-A10 Page 1 [email protected]

MAT2002 - ADDE Dr. T. Phaneendra
Module 2 Professor of Mathematics

Nature of a quadratic form:

Let 1, 2, …, n be eigenvalues of the symmetric matrix A corresponding to the qf (1).

rank index
eigenvalues nature sign. (s)
(r) (p)
i > 0 for all i positive and definite n n n
i  0 for all i with n–k
positive and
at least one i = 0, <n n–k n–k
semi-definite
say k i’s are zero
i < 0 for all i negative and definite n 0 –n
i  0 for all i with n–l
negative and
at least one i = 0, <n 0 – (n – l)
semi-definite
say l i’s are zero
i  0 or  0 indefinite – – –

Example 3. Reduce the quadratic form

  3 x12  2 x1 x2  3x22  2 x2 x3  3x32  2 x1 x3
to a canonical form and find its rank, index, signature and the nature.
3 1 1
Solution. Given that   X T AX , where A  1 3  1 .
1  1 3 
 
Eigenvalues of A are  = 4, 4, 1 (not distinct).
The corresponding eigenvectors are

Eigenvalue Eigenvectors orthogonality

1 1
1 = 2 = 4 X1   1  , X2   0 X1 not  X2
 0 1
   
  1 X3  X1
3 = 1 X3   1 
1 X3  X2
 

A vector which is orthogonal to X1 is given by

X X   1  1  1  1 / 2  1  1 
V2  X 2   1 2 2  X 1   0      1    1 / 2    1
 X   1  2  0  1  2  2 
 1         
We Re-label the pairwise orthogonal vectors of A as
1 1   1
U1   1  , U 2   1 and U3   1 
0 2 1
     
whose normalized forms are
1 / 2   1/ 6   1 / 3 
e1  1 / 2  , e2   1 / 6  and e3   1 / 3  .
   
 0   1/ 6   1/ 3 
     
Modal matrix is P  [e1 e2 ... en ] . The diagonal form of the symmetric
 4 0 0
matrix A is D  P 1 AP  PT AP  diag [4, 4,1]   0 4 0 .
 0 0 1
 
Orthogonal transformation is X  PY . The canonical form of the qf  is
Q  Y T DY  4 y12  4 y22  y32 .
The qf  is positive and definite and rank = index = signature = 3.

SJT 511-A10 Page 2 [email protected]

MAT2002 - ADDE Dr. T. Phaneendra
Module 2 Professor of Mathematics

Example 4. Find an orthogonal matrix C that reduces

  x   2 x12  4 x1 x2  5 x22
to diagonal form.

Solution:
Writing   x   2 x12  2 x1 x2  2 x2 x1  5 x22
2 2
we see that   x   X T AX where A   .
 2 5
Eigenvalues and the corresponding normalized eigenvectors of A are
1 1
1  1 , u1   2, 1 ; 2  6 , u2  1, 2 
5 5
1  2 1
so that C  
5  1 2 
 x1  1  2 1   y1 
The orthogonal transformation is X  CY or     
 x2  5  1 2   y2 
and
1  2 1 2 2  2 1   y1 
  x   y1 , y2      
5  1 2  2 5  1 2   y2 
 1 0   y1 
  y1 , y2       y12  6 y22  Y T DY
 0 6   y2 

Geometric interpretation: The quadratic form Q  x   2 x12  4 x1 x2  5 x22  c

represents the ellipse plotted in the x1x2-plane. The orthogonal operator C represents a rotation
of the coordinate axes to the y1y2-system relative to which Q  y   y12  6 y22  c is diagonal.

Example 5. Identify the conic section whose equation is   2 x12  x22  4 x1 x2  1

Ans. Hyperbola

Example 6. Reduce the quadratic form

  9 x12  9 x22  9 x32  2 x1 x2  2 x2 x3  2 x1 x3
to a canonical form and find its rank, index, signature and the nature.

9 1 1
Solution. Given that   X T AX , where A   1 9 1  .
1 1 9
 
Eigenvalues of A are  = 11, 8, 8 (not distinct).
The corresponding eigenvectors are

Eigenvalue Eigenvectors orthogonality

1 X3  X1
1 = 11 X 1  1
1 X3  X2
 
 1  1
2 =3 = 8 X2   1  , X3   0  X2 not  X3
0 1
   

SJT 511-A10 Page 3 [email protected]

MAT2002 - ADDE Dr. T. Phaneendra
Module 2 Professor of Mathematics

A vector which is orthogonal to X2 is given by

X X   1  1   1  1 / 2  1 1 
V3  X 3   2 2 3  X 2   1      0    1     2 
 X   0   2   1   1 / 2  2  1 
 2       
We Re-label the pairwise orthogonal vectors of A as
1  1 1
U1   1 , U 2   0  and U 3   2 
1 1 1
     
whose normalized forms are
1 / 3   1 / 2   1/ 6 
e1  1 / 3  , e2   0  and e3   2 / 6  .
1 / 3     1/ 6 
   1/ 2   
1 / 3 1 / 2 1 / 6 
Modal matrix is P  [e1 e2 e3 ]  1 / 3 0 2 / 6  .

1 / 3 1 / 2 1 / 6 

The diagonal form of the symmetric matrix A is
11 0 0 
D  P 1 AP  PT AP  diag [11,8,8]   0 8 0  .
 0 0 8
 
Orthogonal transformation is X  PY , that is
 x1  1 / 3 1 / 2 1 / 6   y1  x1  13  22  36
y y y

 x2   1 / 3 0 2 / 6   y2  or x2  13
y
 63
2y

 x    
 3  1 / 3 1 / 2 1 / 6   y3  x1  13  22  36
y y y

The canonical form of the qf  is

Q  Y T DY  11y12  8 y22  8 y32 .
The qf  is positive and definite and rank = index = signature = 3.

Example 7. Reduce the quadratic form

  8 x12  7 x22  3 x32  12 x1 x2  8 x2 x3  4 x1 x3
to a sum of squares form and find its rank, index, signature and the nature.
 8 6 2 
Ans. Given that   X T AX , where A   6 7 4  .
 2 4 3 
 
Eigenvalues of A are  = 15, 3, (distinct).
The corresponding eigenvectors are

Eigenvalue Eigenvectors orthogonality

 2
1 = 15 X 1   2 
1
 
 2 X1  X2
2 = 3 X2   1  X2  X3
 2 
  X1  X3
1
3 = 0 X3   2
 2
 

SJT 511-A10 Page 4 [email protected]

MAT2002 - ADDE Dr. T. Phaneendra
Module 2 Professor of Mathematics

The normalized forms are

 2/3   2/3  1/ 3 
e1   2 / 3  , e 2   1 / 3  and e3   2 / 3  .
 1/ 3   2 / 3   2 / 3
     
Modal matrix is P  [e1 e2 e3 ] . The diagonal form of the symmetric matrix A is
D  P 1 AP  PT AP  diag [15,3,0] .
Orthogonal transformation is X  PY

Example 8. Identify the conic section whose equation is

  6 x12  3 x22  3 x32  4 x1 x2  2 x2 x3  4 x1 x3
Ans. Ellipsoid

Exercises Reduce the following quadratic forms to an appropriate sum of squares form
(a)   2 x1 x2  2 x2 x3  2 x1 x3 (Eigenvalues not Distinct)
(b)   3 x12  5 x22  3 x32  2 x1 x2  2 x2 x3  2 x1 x3 (Distinct Eigenvalues)
(c)   5 x12  6 x22  7 x32  4 x1 x2  4 x2 x3 (Distinct Eigenvalues)

SJT 511-A10 Page 5 [email protected]