Chapter 5: Euclidean Spaces and Quadratic Forms

Linear Algebra MI1141

Department of Mathematics and Informatics

Hanoi University of Science and Technology
2025

Introduction
In addition to the two basic operations of vector spaces, we encounter other operations
between vectors when studying specific cases. In this chapter, we examine vector spaces
equipped with inner products, alongside the study of bilinear forms and quadratic forms
on vector spaces.

Chapter Contents:

1. Bilinear Forms

2. Quadratic Forms

3. Euclidean Spaces

4. Introduction to Curves and Surfaces of Second Degree

Contents
1 Bilinear Forms 3
1.1 Concept of Bilinear Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Matrix and Expression of a Bilinear Form . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Coordinate Expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.2 Change of Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Quadratic Forms 5
2.1 Concept of Quadratic Forms and Relation to Symmetric Bilinear Forms . . . . . 6
2.2 Matrix and Expression of a Quadratic Form . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Coordinate Expression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Definite Quadratic Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4 Methods for Reducing Quadratic Forms to Canonical Form . . . . . . . . . . . . 7
2.4.1 Lagrange’s Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4.2 Jacobi’s Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5 Sylvester’s Criterion for Definite Quadratic Forms . . . . . . . . . . . . . . . . . 9
2.6 Law of Inertia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

3 Euclidean Spaces 9
3.1 Concept of Inner Product and Definition of Euclidean Space . . . . . . . . . . . . 10
3.2 Length and Angle of Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 Orthogonality Relations in Euclidean Spaces . . . . . . . . . . . . . . . . . . . . 11
3.3.1 Properties of Orthonormal Bases . . . . . . . . . . . . . . . . . . . . . . . 12
3.3.2 Gram-Schmidt Orthogonalization . . . . . . . . . . . . . . . . . . . . . . . 12
3.3.3 Orthogonal Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4 Orthogonal Matrices and Orthogonal Transformations . . . . . . . . . . . . . . . 13
3.5 Orthogonal Diagonalization of Symmetric Matrices . . . . . . . . . . . . . . . . . 14
3.5.1 Orthogonal Diagonalization Problem . . . . . . . . . . . . . . . . . . . . . 14
3.6 Symmetric Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.7 Reducing Quadratic Forms to Canonical Form by Orthogonal Method . . . . . . 16

4 Curves and Surfaces of Second Degree 16

4.1 General Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Transforming Equations and Classifying Curves/Surfaces . . . . . . . . . . . . . 17
4.3 Types of Second-Degree Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.4 Types of Second-Degree Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

1 Bilinear Forms
On an R-vector space V , when considering linear maps from V to R, we have linear forms. When
considering maps from V × V to R that are linear in each variable, we have bilinear forms. More
generally, maps from V n to R that are linear in each variable give us multilinear forms. To better
understand multilinear forms, we begin with bilinear forms—a concept encountered frequently
in familiar vector spaces.

Objectives

Knowledge: Students understand the concept of bilinear forms on a vector space, com-
mon examples, and computational aspects related to bilinear forms.
Skills: Verify the bilinear form concept, determine the matrix and expression of a bilinear
form.

1.1 Concept of Bilinear Forms

Definition 1.1 (Bilinear Form). Let V be a vector space over R. A map

f : V × V → R, (u, v) 7→ f (u, v)

is called a bilinear form on V if it satisfies the following conditions:

1. f (u + v, w) = f (u, w) + f (v, w) for all u, v, w ∈ V

2. f (λu, v) = λf (u, v) for all u, v ∈ V, λ ∈ R

3. f (u, v + w) = f (u, v) + f (u, w) for all u, v, w ∈ V

4. f (u, λv) = λf (u, v) for all u, v ∈ V, λ ∈ R

If f additionally satisfies f (u, v) = f (v, u) for all u, v ∈ V , then f is called a symmetric

bilinear form on V .

1.2 Examples
Example 1.2 (Examples of Bilinear Forms). 1. Let V = R2 . Consider the map:

f (x, y) = x1 y1 + 2x1 y2 − x2 y1 + 3x2 y2

where x = (x1 , x2 ), y = (y1 , y2 ). Then f is a bilinear form on R2 .

2. More generally, let V = Rn and A be a square matrix of order n. Consider the map
f (x, y) = [x]T A[y], where if x = (x1 , x2 , . . . , xn ) then [x] = [x1 x2 · · · xn ], and we identify
the 1 × 1 matrix [a] with the real number a. Then f is a bilinear form on Rn .

3. Let V = Pn [x]. The map f : V × V → R given by:

f (p(x), q(x)) = p(0)q(0) + p(1)q(2)

is a bilinear form.

4. Let V = Pn [x]. The map f : V × V → R given by:

f (p(x), q(x)) = p(0) + q(0)

is NOT a bilinear form.

3
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

5. Let V = C[a, b] be the space of continuous functions on the interval [a, b]. The map
T : V × V → R given by:
Z b
T (f (x), g(x)) = f (x)g(x) dx
a

is a bilinear form.

1.3 Matrix and Expression of a Bilinear Form

Definition 1.3 (Matrix of a Bilinear Form). Let V be a finite-dimensional vector space with
basis B = {u1 , u2 , . . . , un }. Let f be a bilinear form on V . The matrix
 
a11 a12 · · · a1n
 a21 a22 · · · a2n 
A= .
 
. .. . . .. 
 . . . . 
an1 an2 · · · ann

where aij = f (ui , uj ), is called the matrix of the bilinear form f with respect to basis B.
Example 1.4. Consider the bilinear form on V = R2 given by:

f (x, y) = x1 y1 + 2x1 y2 − x2 y1 + 3x2 y2

where x = (x1 , x2 ), y = (y1 , y2 ). The matrix of f with respect to the standard basis of R2 is:
 
1 2
A=
−1 3

1.3.1 Coordinate Expression

Let V be a finite-dimensional vector space with basis B = {u1 , u2 , . . . , un }. Let f be a bilinear
form on V and A be the matrix of f with respect to basis B.
For any two vectors u, v ∈ V , suppose their coordinates with respect to basis B are:
   
x1 y1
 x2   y2 
[u]B =  .  = X, [v]B =  .  = Y
   
.
 .   .. 
xn yn
Then:  
Xn n
X n
X n
X
f (u, v) = f  xi ui , yj uj =
 xi yj f (ui , uj ) = aij xi yj
i=1 j=1 i,j=1 i,j=1
Pn
The formula f (u, v) = i,j=1 aij xi yj is called the coordinate expression of f with respect
to basis B.
In matrix form: f (u, v) = [u]TB · A · [v]B = X T AY .
Remark 1.5. f is a symmetric bilinear form if and only if A is a symmetric matrix.
Example 1.6. Let V = R3 with basis B = {u1 = (1, 1, 2), u2 = (2, −1, 3), u3 = (0, 2, 1)}. Let
the bilinear form f have matrix  
1 4 7
A = 2 5 8
3 6 9
with respect to basis B. Then:

4
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

• f (u2 , u3 ) = 8

• f (u3 , u1 ) = 3
  
1 4 7 1
• f (2u1 + u2 − u3 , u1 − u2 + 3u3 ) = [2 1 − 1] 2 5 8
   −1 = 33
3 6 9 3

1.3.2 Change of Basis

In vector space V , let B1 = {u1 , u2 , . . . , un } and B2 = {v1 , v2 , . . . , vn } be two bases with change-
of-basis matrix from B1 to B2 denoted by C. For any two vectors u, v ∈ V :

[u]B1 = C[u]B2 , [v]B1 = C[v]B2

Let f be a bilinear form on V with matrices A1 , A2 with respect to bases B1 , B2 respectively.

Then for all u, v ∈ V :
f (u, v) = [u]TB1 A1 [v]B1 = [u]TB2 C T A1 C[v]B2
and
f (u, v) = [u]TB2 A2 [v]B2
This gives the change-of-basis formula:

A2 = C T A1 C

Example 1.7. Let V = R2 with bilinear form f ((x1 , x2 ), (y1 , y2 )) = 3x1 y1 +2x1 y2 −x2 y1 +4x2 y2
and basis B = {u1 = (1, 2), u2 = (3, 4)}. Find the matrix of f with respect to basis B.
Solution: The matrix of f with respect to the standard basis is:
 
3 2
A1 =
−1 4

The change-of-basis matrix from E to B is:

 
1 3
C=
2 4

Therefore, the matrix of f with respect to basis B is:

       
T 1 2 3 2 1 3 21 43
A2 = C A1 C = · · =
3 4 −1 4 2 4 49 103

2 Quadratic Forms
A quadratic form is a polynomial in several variables where all terms are monomials of degree
2. In this section, we approach quadratic forms from the perspective of vector spaces and
symmetric bilinear forms. Through this, we examine some properties of quadratic forms in
relation to symmetric matrices.

Objectives

Knowledge: Students understand the concept of quadratic forms and their properties.
Skills: Determine the relationship between quadratic forms and symmetric bilinear
forms, methods for reducing quadratic forms to canonical form, examine positive def-
inite and negative definite quadratic forms.

5
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

2.1 Concept of Quadratic Forms and Relation to Symmetric Bilinear Forms

Definition 2.1 (Quadratic Form). Let f be a symmetric bilinear form on V . The map
φ : V → R, u 7→ φ(u) = f (u, u)
is called the quadratic form on V associated with the symmetric bilinear form f .
Note: The symmetric bilinear form f is completely determined by the quadratic form φ,
since:
1
f (u, v) = [φ(u + v) − φ(u) − φ(v)]
2
The form f is called the polar form of φ.
Example 2.2 (Examples of Quadratic Forms). 1. Consider the symmetric bilinear form f
2
on R given by:
f (x, y) = x1 y1 + 2x1 y2 + 2x2 y1 + 3x2 y2
where x = (x1 , x2 ), y = (y1 , y2 ). The associated quadratic form φ is:
φ(x1 , x2 ) = x21 + 2x1 x2 + 2x2 x1 + 3x22 = x21 + 4x1 x2 + 3x22

2. More generally, let A be a symmetric matrix of order n. Consider the symmetric bilinear
form f on Rn given by f (x, y) = [x]T A[y]. Then φ(x) = [x]T A[x] gives the associated
quadratic form.
3. Let V = Pn [x]. The symmetric bilinear form f : V × V → R defined by:
f (p(x), q(x)) = p(0)q(0) + p(1)q(1)
has associated quadratic form φ(p(x)) = p(0)2 + p(1)2 .
4. Let V = C[a, b] be the space of continuous functions on [a, b]. The symmetric bilinear
form T : V × V → R defined by:
Z b
T (f (x), g(x)) = f (x)g(x) dx
a
Rb 2
has associated quadratic form φ(f (x)) = a [f (x)] dx.

2.2 Matrix and Expression of a Quadratic Form

Definition 2.3 (Matrix of a Quadratic Form). Let V be a finite-dimensional vector space with
basis B = {u1 , u2 , . . . , un }. Let f be a symmetric bilinear form on V and φ be the associated
quadratic form. The symmetric matrix
 
a11 a12 · · · a1n
 a21 a22 · · · a2n 
A= .
 
.. .. .. 
 .. . . . 
an1 an2 · · · ann
where aij = f (ui , uj ), is called the matrix of the quadratic form φ with respect to basis B.
Example 2.4. Consider the quadratic form φ on R2 defined by:
φ(x) = φ(x1 , x2 ) = 3x21 + 2x1 x2 − 2x22
The matrix of φ with respect to the standard basis of R2 is:
 
3 1
A=
1 −2

6
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

2.2.1 Coordinate Expression

Let V be a finite-dimensional vector space with basis B = {u1 , u2 , . . . , un }. Let φ be a quadratic
form on V associated with symmetric bilinear form f , and let A be the matrix of φ with respect
to basis B.
For any vector u ∈ V with coordinates with respect to basis B:
 
x1
 x2 
[u]B =  .  = X
 
 .. 
xn

Then:
 
Xn n
X n
X n
X
φ(u) = f (u, u) = f  xi ui , xj uj =
 xi xj f (ui , uj ) = aij xi xj
i=1 j=1 i,j=1 i,j=1

The formula φ(u) = ni,j=1 aij xi xj , written in matrix form as φ(u) = [u]TB ·A·[u]B = X T AX,
P
is called the (coordinate) expression of the quadratic form φ with respect to basis B.
When A = diag(a1 , a2 , . . . , an ) is a diagonal matrix, we have:

φ(u) = a1 x21 + a2 x22 + · · · + an x2n

This is called the canonical form of the quadratic form.

Especially, if ai ∈ {−1, 0, 1} for all i = 1, . . . , n, we call it the normal form of the quadratic
form.
As with bilinear forms, we have the change-of-basis formula for quadratic forms:
Let A1 and A2 be the matrices of quadratic form φ with respect to bases B1 and B2
respectively, and let C be the change-of-basis matrix from B1 to B2 . Then:

A2 = C T A1 C

2.3 Definite Quadratic Forms

Definition 2.5 (Positive and Negative Definite). A quadratic form φ is called:
• Positive definite if φ(u) > 0 for all u ̸= 0

• Negative definite if φ(u) < 0 for all u ̸= 0

• Indefinite if it can take both positive and negative values

If quadratic form φ is associated with bilinear form f , we also say f is positive definite (or
negative definite) if φ is positive definite (or negative definite).
Remark 2.6. A quadratic form is positive definite (negative definite) if and only if all coeffi-
cients in its canonical form are positive (respectively, all negative).

2.4 Methods for Reducing Quadratic Forms to Canonical Form

2.4.1 Lagrange’s Method
Use appropriate variable substitutions x = Cy to transform the quadratic form φ = xT Ax into
φ = y T (C T AC)y in canonical form. Lagrange’s method is based on two basic manipulations:

b 2 4ac−b2
1. ax2 + bx + c = a x +


2a + 4a

7
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

2. xy = 14 [(x + y)2 − (x − y)2 ]

Example 2.7. Let φ = x21 + 2x22 + 5x23 − 2x1 x2 + 4x1 x3 . Reduce to canonical form.
Solution:

φ = x21 + 2x22 + 5x23 − 2x1 x2 + 4x1 x3

= x21 − 2x1 (x2 − 2x3 ) + 2x22 + 5x23
= x21 − 2x1 (x2 − 2x3 ) + (x2 − 2x3 )2 + x22 + 4x2 x3 + x23
= (x1 − x2 + 2x3 )2 + (x2 + 2x3 )2 − 3x23

Setting y1 = x1 − x2 + 2x3 , y2 = x2 + 2x3 , y3 = x3 , we get:

φ = y12 + y22 − 3y32

in canonical form.
Example 2.8. Let φ = x1 x2 − 2x2 x3 . Reduce to canonical form.
Solution: Setting x1 = y1 + y2 , x2 = y1 − y2 , x3 = y3 :

φ = y12 − y22 − 2y1 y3 + 2y2 y3

= (y12 − 2y1 y3 + y32 ) − y22 + 2y2 y3 − y32
= (y1 − y3 )2 − (y2 − y3 )2

Setting z1 = y1 − y3 , z2 = y2 − y3 , z3 = y3 :

φ = z12 − z22

Remark 2.9. 1. Variable substitutions of the form y = C −1 x or x = Cy where C is invert-

ible correspond to changing from the original basis to a new basis.

2. Every quadratic form can be reduced to canonical form and normal form.

3. For a symmetric matrix A, there always exists an invertible matrix C such that C T AC is
diagonal.

4. For a quadratic form φ associated with symmetric bilinear form f , there always exists a
basis B = {u1 , u2 , . . . , un } of V satisfying f (ui , uj ) = 0 for all i ̸= j.

2.4.2 Jacobi’s Method

Jacobi’s method is based on constructing vectors u1 , u2 , . . . , un in basis B of V satisfying
f (ui , uj ) = 0 for all i ̸= j. Jacobi’s method applies when all principal minors of matrix A
are nonzero.
The principal minor of order r of square matrix A is the determinant:

a11 a12 · · · a1r

a21 a22 · · · a2r
∆r = . .. .. .
.. . . ..
ar1 ar2 · · · arr

Theorem 2.10 (Jacobi’s Theorem). Suppose for basis u1 , u2 , . . . , un of space V we have a

quadratic form with expression:
n
X
φ(u) = aij xi xj , aij = φ(ui , uj )
i=1

8
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

Furthermore, suppose the principal minors ∆1 , ∆2 , . . . , ∆n of matrix A = (aij ) are all

nonzero. Then there exists a basis v1 , v2 , . . . , vn of V such that the quadratic form can be
written as:
φ(u) = k1 y12 + k2 y22 + · · · + kn yn2
∆i−1
where the coefficients ki are determined by ki = ∆i , i = 1, 2, . . . , n, with ∆0 = 1.

2.5 Sylvester’s Criterion for Definite Quadratic Forms

Theorem 2.11 (Sylvester’s Criterion). A quadratic form is positive definite if and only if all
its principal minors are positive. A quadratic form is negative definite if and only if all principal
minors of even order are positive and all principal minors of odd order are negative.
Example 2.12. Consider the quadratic form:

φ(x1 , x2 , x3 ) = x21 + 4x22 + 8x23 + 2x1 x2 − 2x1 x3

The matrix of φ is:  

1 1 −1
A= 1 4 0 
−1 0 8
The principal minors are ∆1 = 1, ∆2 = 3, ∆3 = 20. Therefore, the given quadratic form is
positive definite.

2.6 Law of Inertia

For a quadratic form φ on V , we can reduce φ to many different canonical forms.
Theorem 2.13 (Law of Inertia). The number of positive coefficients, negative coefficients, and
zero coefficients in the canonical form of φ is invariant, depending only on φ, not on the basis.
Thus, suppose φ has canonical form:

a1 x21 + · · · + ar x2r − ar+1 x2r+1 − · · · − ar+s x2r+s

with respect to basis B, where ai > 0 for all i = 1, . . . , r + s.

Suppose φ has canonical form:

b1 y12 + · · · + br′ yr2′ − br′ +1 yr2′ +1 − · · · − br′ +s′ yr2′ +s′

with respect to basis B ′ , where bi > 0 for all i = 1, . . . , r′ + s′ .

Then r = r′ and s = s′ .

3 Euclidean Spaces
In ordinary vector spaces, we are familiar with concepts such as inner product, angle between
vectors, length, orthogonality relation, orthogonal projection, etc. A natural question is: do
these concepts exist in general vector spaces? How are these concepts constructed and applied?

Objectives

Knowledge: Students understand the concept of inner product defining Euclidean

spaces and the issues to be examined in Euclidean spaces.
Skills: Verify inner products, examine orthogonality relations in Euclidean spaces such
as orthogonal systems, orthogonalization, orthogonal spaces, orthogonal projection, or-

9
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

thogonal transformations, orthogonal diagonalization problems.

3.1 Concept of Inner Product and Definition of Euclidean Space

Definition 3.1 (Inner Product). Let V be an R-vector space. An inner product on V is a
symmetric bilinear form that is positive definite.

Example 3.2 (Examples of Inner Products). 1. The bilinear form f on R2 given by:

f (x, y) = x1 y1 + x2 y2

where x = (x1 , x2 ), y = (y1 , y2 ), is symmetric and positive definite, so it is an inner product

on R2 .

2. More generally, let A be a symmetric matrix of order n. Consider the symmetric bilinear
form f on Rn given by:
f (x, y) = [x]T A[y]
This is an inner product if and only if A has all positive principal minors.

3. Let V = Pn [x]. The bilinear form f : V × V → R defined by:

f (p(x), q(x)) = p(0)q(0) + p(1)q(1)

is symmetric but not positive definite (e.g., f (x2 −x, x2 −x) = 0 but x2 −x ̸= 0). Therefore,
this is not an inner product.

4. Let V = C[a, b] be the space of continuous functions on [a, b]. The bilinear form T :
V × V → R defined by:
Z b
T (f (x), g(x)) = f (x)g(x) dx
a
Rb 2
is symmetric and the associated quadratic form φ(f (x)) = a [f (x)] dx is positive definite.
Thus T is an inner product on V .

Definition 3.3 (Euclidean Space). A real vector space of dimension n equipped with an inner
product is called an n-dimensional Euclidean space. The inner product on a Euclidean
space is usually denoted by ⟨·, ·⟩.

Example 3.4. In the space V = Rn , with u = (x1 , . . . , xn ), v = (y1 , . . . , yn ), define:

n
X
⟨u, v⟩ = x i yi
i=1

This is an inner product called the standard inner product of Rn , giving us the Euclidean
space Rn .

3.2 Length and Angle of Vectors

Definition 3.5 (Length/Norm). The length of vector u, denoted ∥u∥, is defined by:
p
∥u∥ = ⟨u, u⟩

Remark 3.6 (Properties of Norm). • ∥u∥ = 0 if and only if u = 0

• ∥ku∥ = |k|∥u∥ for k ∈ R

10
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

• |⟨u, v⟩| ≤ ∥u∥ · ∥v∥ (Cauchy-Schwarz inequality)

• ∥u + v∥ ≤ ∥u∥ + ∥v∥ (Triangle inequality)

From the Cauchy-Schwarz inequality, we have:

|⟨u, v⟩|
≤1
∥u∥ · ∥v∥
⟨u,v⟩
Therefore, the real number r = ∥u∥·∥v∥ can be viewed as the cosine of an angle α, i.e., there
exists an angle α, 0 ≤ α ≤ π, such that:

⟨u, v⟩
cos α =
∥u∥ · ∥v∥

This α is called the angle between vectors u and v.

Example 3.7. Let V = P2 [x] be a Euclidean space with inner product:

⟨p(x), q(x)⟩ = p(0)q(0) + p(1)q(1) + p(2)q(2)

Consider vectors p(x) = 1 + 2x − x2 and q(x) = 3 + x2 . We have:

p √
∥p(x)∥ = p(0)2 + p(1)2 + p(2)2 = 6
p √
∥q(x)∥ = q(0)2 + q(1)2 + q(2)2 = 74
⟨p(x), q(x)⟩ = p(0)q(0) + p(1)q(1) + p(2)q(2) = 1 · 3 + 2 · 4 + 1 · 7 = 18

The angle α between the two vectors satisfies:

18 9
cos α = √ √ =√
6 · 74 111

3.3 Orthogonality Relations in Euclidean Spaces

Let V be a Euclidean space with inner product ⟨·, ·⟩.

Definition 3.8 (Orthogonality Concepts). 1. Two vectors u, v are called orthogonal if ⟨u, v⟩ =
0.

2. A system of vectors u1 , u2 , . . . , us (s ≥ 2) is called orthogonal if all vectors are nonzero

and pairwise orthogonal, i.e., ⟨ui , uj ⟩ = 0 for i ̸= j. An orthogonal system of vectors is
linearly independent.
u u
3. A vector u is called unit or normalized if ∥u∥ = 1. For u ̸= θ, the vectors ∥u∥ and − ∥u∥
are unit vectors, called normalizations of u.
A system u1 , u2 , . . . , us (s ≥ 2) is called orthonormal if it is orthogonal and each vector
is normalized, i.e.: (
0 if i ̸= j
⟨ui , uj ⟩ =
1 if i = j

4. An orthogonal basis is called an orthogonal basis. An orthonormal basis is called an

orthonormal basis.

11
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

3.3.1 Properties of Orthonormal Bases

Let V be a Euclidean space with inner product ⟨·, ·⟩, and B = {u1 , u2 , . . . , un } be an orthonormal
basis of V .

1. For any u ∈ V , the coordinates of u with respect to basis B are [u]TB = (x1 x2 · · · xn )
where xi = ⟨u, ui ⟩, 1 ≤ i ≤ n. In other words:

u = ⟨u, u1 ⟩u1 + ⟨u, u2 ⟩u2 + · · · + ⟨u, un ⟩un

2. Let [u]TB = X T = (x1 x2 · · · xn ) and [v]TB = Y T = (y1 y2 · · · yn ). Then:

⟨u, v⟩ = X T Y = x1 y1 + x2 y2 + · · · + xn yn

3.3.2 Gram-Schmidt Orthogonalization

In a Euclidean vector space, from each basis we can construct an orthonormal basis. This can
be done by the Gram-Schmidt orthogonalization method.
Theorem 3.9 (Gram-Schmidt Process). Let u1 , u2 , . . . , un be a basis of Euclidean space V .
We construct an orthogonal system as follows:
Set e1 = u1 . Find e2 such that ⟨e2 , e1 ⟩ = 0 by setting e2 = u2 + k21 e1 . Since ⟨e1 , e2 ⟩ = 0:
⟨e1 , u2 ⟩
⟨e1 , u2 ⟩ + k21 ⟨e1 , e1 ⟩ = 0 =⇒ k21 = −
⟨e1 , e1 ⟩
Next, find e3 such that e1 , e2 , e3 is orthogonal by setting e3 = u3 + k31 e1 + k32 e2 . From
⟨e1 , e3 ⟩ = ⟨e2 , e3 ⟩ = 0:
⟨e1 , u3 ⟩ ⟨e2 , u3 ⟩
k31 = − , k32 = −
⟨e1 , e1 ⟩ ⟨e2 , e2 ⟩
This process continues to determine er such that ⟨ei , er ⟩ = 0 for i = 1, 2, . . . , r − 1 by setting:

er = ur + kr1 e1 + kr2 e2 + · · · + kr(r−1) er−1

where:
⟨ej , ur ⟩
krj = − , j = 1, 2, . . . , r − 1
⟨ej , ej ⟩
After normalizing this system, we obtain an orthonormal basis of V .

3.3.3 Orthogonal Projection

Let V be a Euclidean space and U be a subspace. Vector w ∈ U is called the orthogonal
projection of vector v onto subspace U if vector v − w is orthogonal to all vectors in U . We
denote the projection of v onto U by ProjU (v).
Remark 3.10. • The orthogonal projection ProjU (v) always exists and is unique.

• If v ∈ U then ProjU (v) = v.

• If U has basis {v1 , v2 , . . . , vp }, we can find ProjU (v) in two ways:

Method 1: Orthonormalize using Gram-Schmidt to get orthonormal basis {u1 , u2 , . . . , up }
of U , then:
ProjU (v) = ⟨v, u1 ⟩u1 + ⟨v, u2 ⟩u2 + · · · + ⟨v, up ⟩up

Method 2: Find projection by definition by setting ProjU (v) = x1 v1 + x2 v2 + · · · + xp vp .

From the condition that v − ProjU (v) is orthogonal to U , we have ⟨ProjU (v), vi ⟩ = ⟨v, vi ⟩,
determining the xi by solving a Cramer system.

12
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

Example 3.11. In space V = R3 with standard inner product, let:

U = span(v1 = (1, 2, 1), v2 = (1, 1, 3)), v = (3, 5, 7)

Find ProjU (v).

Solution: Set ProjU (v) = x1 (1, 2, 1) + x2 (1, 1, 3). Since ⟨ProjU (v), vi ⟩ = ⟨v, vi ⟩ for i = 1, 2:
(
6x1 + 6x2 = 20
6x1 + 11x2 = 29

23
Solving: x1 = 15 , x2 = 59 . Therefore:
 
23 9 10 73 104
ProjU (v) = (1, 2, 1) + (1, 1, 3) = , ,
15 5 3 15 15

3.4 Orthogonal Matrices and Orthogonal Transformations

Definition 3.12 (Orthogonal Matrix). A square matrix A is called orthogonal if AAT = I,
i.e., A−1 = AT .

The following statements are equivalent:

• A is an orthogonal matrix

• The columns of A form an orthonormal basis of Rn with standard inner product

• The rows of A form an orthonormal basis of Rn with standard inner product

Example 3.13. The following matrix is orthogonal:

 1 1
− √13

√ √
6 2
A=
 √2 0 √1 
6 3 
− √16 √1
2
√1
3

Remark 3.14. For orthogonal matrices A, B:

• |A|2 = 1

• A−1 and AB are orthogonal matrices

Proposition 3.15. Let V be a Euclidean space. Let P be the change-of-basis matrix from one
orthonormal basis B1 to another orthonormal basis B2 of V . Then P is an orthogonal matrix.

Definition 3.16 (Orthogonal Transformation). Let V be a Euclidean space with inner product
⟨·, ·⟩. A linear transformation f on V is called an orthogonal transformation if ∥f (v)∥ = ∥v∥
for all v ∈ V .

The following statements are equivalent:

• f is an orthogonal transformation

• f preserves inner products, i.e., ⟨f (u), f (v)⟩ = ⟨u, v⟩ for all u, v ∈ V

• The matrix of f with respect to an orthonormal basis B of V is an orthogonal matrix

13
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

Example 3.17. Consider the linear transformation f on V = R3 with standard inner product
having matrix:  1
√1 − √13

√
6 2
A =  √26 0 √1 

3 
− √16 √12 √1
3

with respect to the standard basis of R3 .

For any u = (x1 , x2 , x3 ) ∈ R3 , writing f (u) = (y1 , y2 , y3 ):
   
y1 x1
y2  = A x2 
y3 x3

Therefore:
   
y1 x1
y12 + y22 + y32 = y1 y2 y2 = x1 x2 x3 A A x2  = x21 + x22 + x23


T
y3   
y3 x3

Thus ∥f (u)∥ = ∥u∥, and f is an orthogonal transformation.

3.5 Orthogonal Diagonalization of Symmetric Matrices

For a (real) symmetric matrix A, we have the following properties:

Theorem 3.18 (Properties of Symmetric Matrices). 1. All eigenvalues of A are real. In

other words, the characteristic polynomial has the form:
s
Y
Pf (λ) = (λi − λ)ri , λi ∈ R
i=1

2. Distinct eigenspaces of A are orthogonal to each other. Moreover, the eigenspace corre-
sponding to eigenvalue λi has dimension ri .

3. Matrix A is diagonalizable.

3.5.1 Orthogonal Diagonalization Problem

Problem: For symmetric matrix A, find orthogonal matrix C such that C T AC is diagonal.
Procedure:
Step 1: Determine the characteristic polynomial of matrix A and find the eigenvalues.
Step 2: For each eigenvalue found, find an orthonormal basis of the corresponding eigenspace.
Step 3: Combine the orthonormal bases of the eigenspaces to obtain an orthonormal basis
B consisting entirely of eigenvectors of A. Let C be the matrix whose columns are the column
vectors of B. Then C is an orthogonal matrix and orthogonally diagonalizes A:

C −1 AC = C T AC = D

is a diagonal matrix. Moreover, the diagonal elements of D are the eigenvalues of A.

 
4 −1 2
Example 3.19. Orthogonally diagonalize matrix A = −1 4 2.
2 2 1

14
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

Step 1: Characteristic polynomial:

4 − λ −1 2
PA (λ) = −1 4 − λ 2 = −λ3 + 9λ2 − 15λ − 25 = −(λ + 1)(λ − 5)2
2 2 1−λ

Eigenvalues: λ = −1, 5.
Step 2:
 √ 
 1/√6 
• For λ = −1: eigenspace VA (−1) has orthonormal basis B1 =  1/ √6 
−2/ 6
 

 √   √ 
 1/ √2 1/√3 
• For λ = 5: eigenspace VA (5) has orthonormal basis B2 = −1/ 2 , 1/√3
0 1/ 3
 

Step 3: Orthonormal basis B = B1 ∪ B2 of space V consists entirely of eigenvectors of A.

Orthogonal matrix:  1 1√ 1  √ √
6 2 3
 √1 − √12 √1 
C=  6 3
− √26 0 √1
3

diagonalizes A:  
−1 0 0
C −1 AC = C T AC =  0 5 0
0 0 5

3.6 Symmetric Transformations

Definition 3.20 (Symmetric Transformation). Let V be a Euclidean space with inner product
⟨·, ·⟩. A linear operator f on V is called a symmetric transformation if:

⟨f (u), v⟩ = ⟨u, f (v)⟩ for all u, v ∈ V

The following are equivalent:

• f is a symmetric transformation

• The matrix of f with respect to any orthonormal basis B of V is a symmetric matrix

• There exists an orthonormal basis of V such that the matrix of f with respect to this
basis is diagonal

Example 3.21. Consider 2

 the linear transformation on V = R with standard inner product,
3 4
having matrix A = with respect to the standard basis E of R2 .
4 5
For any u = (x1 , x2 ) ∈ V and v = (y1 , y2 ) ∈ V :

⟨f (u), v⟩ = [f (u)]TE [v]E = (A[u]E )T [v]E = [u]TE AT [v]E = [u]TE A[v]E = [u]TE [f (v)]E = ⟨u, f (v)⟩

Thus f is a symmetric transformation.

15
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

3.7 Reducing Quadratic Forms to Canonical Form by Orthogonal Method

Let q be a quadratic form on vector space Rn .
Let A be the matrix of q with respect to the standard basis E of Rn .
Then A is a symmetric matrix. We orthogonally diagonalize A, finding orthogonal matrix
P such that:
P −1 AP = P T AP = D = diag[α1 , . . . , αn ]
Let B be the basis of Rn consisting of vectors whose coordinates with respect to the standard
basis are the columns of P . Then the coordinate expression of q with respect to basis B has
canonical form:
q(v) = α1 y12 + · · · + αn yn2 , [v]B = (y1 , . . . , yn )
The change of variables formula is [v]E = P [v]B , i.e., X = P Y ⇔ Y = P T X.

Example 3.22. Reduce the quadratic form ω(x1 , x2 , x3 ) = 2x21 − x22 − x23 + 4x2 x3 to canonical
form by orthogonal method (specify the transformation).
Solution (summary):  
2 0 0
Matrix of ω with respect to standard basis: A = 0 −1 2 
0 2 −1
Characteristic equation:

0 = det(A − λI) = (2 − λ)(λ − 1)(λ + 3) ⇔ λ = 1, 2, −3

   
0 0√
For λ = 1: eigenvector v1 = 1, normalized: u1 = 1/√2
1 1/ 2
 
1
For λ = 2: eigenvector v2 = 0 = u2 (already normalized)
0
   
0 0√
For λ = −3: eigenvector v3 = −1, normalized: u3 = −1/√ 2
1 1/ 2
   
0√ 1 0√ 1 0 0
Set P = 1/√2 0 −1/√ 2. Then P T AP = D = 0 2 0 .
1/ 2 0 1/ 2 0 0 −3
 √
y1 = (x2 + x3 )/ 2
   
x1 y1 
Change of variables: X = x2  = P y2  ⇔ Y = P T X ⇔ y2 = x1
√
x3 y3

y3 = (−x2 + x3 )/ 2

Canonical form: y12 + 2y22 − 3y32 .

4 Curves and Surfaces of Second Degree

Returning to ordinary geometric problems, we have familiar curves and surfaces commonly
encountered in university study and research whose equations are closely related to quadratic
forms and linear forms.

Objectives

Knowledge: Students understand the theory of 9 types of second-degree curves and 17

types of second-degree surfaces and how to classify them.

16
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

Skills: Transform equations of curves and surfaces to determine the types of second-
degree curves and surfaces.

4.1 General Concepts

Let A = [aij ]n×n be a nonzero symmetric real matrix, column a = (a1 , a2 , . . . , an )T , and real
number a0 . The set of tuples x = (x1 , x2 , . . . , xn )T satisfying:

xT Ax + aT x + a0 = 0

is called a hypersurface in space Rn .

The explicit form is:
Xn n
X
aij xi xj + ai xi + a0 = 0
i,j=1 i=1

When n = 2, the hypersurface is called a second-degree curve. When n = 3, the hyper-

surface is called a second-degree surface.

4.2 Transforming Equations and Classifying Curves/Surfaces

Using orthogonal diagonalization, we find orthogonal matrix C such that D = C T AC is diago-
nal. Use the orthogonal transformation x = Cy ⇔ y = C T x. The equation becomes:
n
X n
X
λi x2i + bi xi + b0 = 0
i=1 i=1

Using the formula:

2
b2 − 4ac

2 b
ax + bx + c = a x + −
2a 4a
combined with translation: 

x1 = α1 + y1

x2 = α2 + y2

..


 .

xn = αn + yn


we reduce the equation to canonical form and classify the second-degree curves and surfaces.

4.3 Types of Second-Degree Curves

x2 y2
1. Ellipse: a2
+ b2
=1
x2 y2
2. Imaginary ellipse: a2
+ b2
= −1
x2 y2
3. Two imaginary intersecting lines: a2
+ b2
=0
x2 y2
4. Hyperbola: a2
− b2
=1
x2 y2
5. Two intersecting lines: a2
− b2
=0

6. Parabola: x2 = py

7. Two parallel lines: x2 = a2

17
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

8. Two coincident lines: x2 = 0

9. Two imaginary parallel lines: x2 = −a2

Example 4.1. Determine the shape of the second-degree curve:

x2 − 2xy + y 2 + 4x + 2y − 6 = 0

Solution: φ(x, y) = x2 − 2xy + y 2 . Using orthogonal diagonalization and setting:

   √ √  
x 1/√2 −1/√ 2 x1
=
y 1/ 2 1/ 2 y1

The quadratic form becomes φ√′ (x , y ) = 2x2 .

√ 1 1 1
Final equation: x21 + 2x1 + 3 2y1 − 6 = 0
 √ 2 √  √ 
After completing the square: x1 + 22 + 3 2 y1 − 13 2
12 =0
√
Using translation: x22 + 3 2y2 = 0
This is a parabola.

Example 4.2. Identify the second-degree curve (C): 5x21 + 8x22 − 4x1 x2 = 36 
5 −2
Solution: Quadratic form h = 5x21 + 8x22 − 4x1 x2 has matrix A =
−2 8
(
5 − λ −2 λ1 = 4
Characteristic equation: =0⇔
−2 8 − λ λ2 = 9
 √ √ 
2/√5 −1/√ 5
Orthogonal diagonalization with matrix C =
1/ 5 2/ 5
   
x1 y
Setting =C 1 :
x2 y2
y2 y2
Equation becomes: 4y12 + 9y22 = 36 ⇔ 91 + 42 = 1
Therefore (C) is an ellipse with semi-axes 3 and 2.

4.4 Types of Second-Degree Surfaces

x2 y2 z2
1. Ellipsoid: a2
+ b2
+ c2
=1
x2 y2 z2
2. Imaginary ellipsoid: a2
+ b2
+ c2
= −1
x2 y2 z2
3. Imaginary cone (single point): a2
+ b2
+ c2
=0
x2 y2 z2
4. Hyperboloid of one sheet: a2
+ b2
− c2
=1
x2 y2 z2
5. Cone: a2
+ b2
− c2
=0
x2 y2 z2
6. Hyperboloid of two sheets: a2
+ b2
− c2
= −1
x2 y2
7. Elliptic paraboloid: a2
+ b2
= pz
x2 y2
8. Hyperbolic paraboloid (saddle): a2
− b2
= pz
x2 y2
9. Elliptic cylinder: a2
+ b2
=1
x2 y2
10. Hyperbolic cylinder: a2
− b2
=1

18
Chapter 5: Euclidean Spaces and Quadratic Forms MI1141 - Linear Algebra

11. Parabolic cylinder: x2 = py

x2 y2
12. Imaginary elliptic cylinder: a2
+ b2
= −1
x2 y2
13. Two imaginary intersecting planes: a2
+ b2
=0
x2 y2
14. Two intersecting planes: a2
− b2
=0

15. Two parallel planes: x2 = a2

16. Two imaginary parallel planes: x2 = −a2

17. Two coincident planes: x2 = 0

End of Chapter 5: Euclidean Spaces and Quadratic Forms

19