UNIT 6:

EUCLIDEAN GEOMETRY

Gijón Polytechnic School of Engineering

University of Oviedo
Contents

1 Euclidean vector space.

2 Matrix associated to an inner product.

3 Metric properties.

4 Orthogonality.

5 Orthogonal diagonalization.
Contents

1 Euclidean vector space.

2 Matrix associated to an inner product.

3 Metric properties.

4 Orthogonality.

5 Orthogonal diagonalization.
Euclidean vector space

Definition 6.1
Let V be a vector space over R. The map
·: V ×V → R
(~v , ~u ) 7 → ~v · ~u
is an inner product (or scalar product, or dot product) if
a) (linearity in each argument)
(α~v1 + β~v2 ) · ~u = α~v1 · ~u + β~v2 · ~u
∀~v1 , ~v2 , ~u ∈ V , ∀α, β ∈ R;
~u · (α~v1 + β~v2 ) = α~u · ~v1 + β~u · ~v2
b) (symmetry)
~v · ~u = ~u · ~v ∀~v , ~u ∈ V ;
c) (positive-definiteness)
~v · ~v ≥ 0 ∀~v ∈ V , and ~v · ~v = 0 ⇒ ~v = ~0.
In that case, (V , ·) is an Euclidean space.
Example

Example 6.1
Show that

(x1 , x2 ) · (y1 , y2 ) = 2x1 y1 + x2 y1 + x1 y2 + x2 y2

defines an inner product on the vector space R2 .

Scalar square

Definition 6.2
Given an Euclidean space (V , ·), the map

V −→ R
~v 7→ ~v 2 = ~v · ~v

is the associated scalar square.

Example 6.2
Find the scalar square associated to the scalar product

(x1 , x2 ) · (y1 , y2 ) = 2x1 y1 + x2 y1 + x1 y2 + x2 y2 ,

and compute (1, 0)2 .

Contents

1 Euclidean vector space.

2 Matrix associated to an inner product.

3 Metric properties.

4 Orthogonality.

5 Orthogonal diagonalization.
Matrix form

Given (V , ·) an Euclidean space and B = {~v1 , ~v2 , . . . , ~vn } a basis of V , we

rewrite the inner product in matrix form as

~ = X t GY ,
~u · w

~ = y1~v1 + y2~v2 + · · · + yn~vn ,

where ~u = x1~v1 + x2~v2 + · · · + xn~vn , w and

~v12 ~v1 .~v2 ... ~v1 .~vn

     
x1 y1
 ~v1 .~v2 ~v 2 ... ~v2 .~vn   x2   y2
    
2
G = . ..  ; X =  ..  ; Y =  .. .
 
. ..
 . . .   .   . 
~v1 .~vn ~v2 .~vn ... ~vn2 xn yn

The matrix G is known as the Gram matrix (or Grammian matrix, or

Gramian) associated to the inner product · : V × V → R.
Gram matrix

Properties:
Since the inner product is symmetric, gij = ~vi · ~vj = ~vj · ~vi = gji and,
therefore, G is a symmetric matrix.
The inner product is positive-definite, so that gii = ~vi · ~vi > 0.
n X
X n
Moreover, (x1 , x2 , . . . , xn ) · (y1 , y2 , . . . , yn ) = gij xi yj ; in other
i=1 j=1
words, gij is the coefficient associated to the term xi yj in the
expression of the inner product.
In terms of the Gram matrix, the scalar square is
n
X n−1 X
X n−1
(x1 , x2 , . . . , xn )2 = gii xi2 + 2gij xi xj .
i=1 i=1 j=i+1
Examples

Example 6.3
Given the inner product on R3

(x1 , x2 , x3 ) · (y1 , y2 , y3 ) =

= x1 y1 + 6x2 y2 + 7x3 y3 + 2x1 y2 + 2x2 y1 − 2x2 y3 − 2x3 y2 ,

find its Gram matrix for the standard basis.

Example 6.4
Consider a scalar square on R2 defined by

(x, y )2 = 3x 2 − xy + 2y 2 .

Compute the Gram matrix associated to its corresponding inner product

for the standard basis.
Characterization of a Gram matrix

Theorem 6.1 (Sylvester’s criterion)

A ∈ Mn (R) is the Gram matrix associated to some inner product if and
only if A is symmetric and
 
a11 a12 ... a1i
 a12 a22 ... a2i 
det  . >0 ∀i = 1, 2, . . . n.
 
.. ..
 .. . . 
a1i a2i ... aii
Example

Example 6.5
Decide which of the following matrices are associated to an inner product:

5 1 2 2 0 1
   

A =  −1 1 −1  ; B= 0 8 2 ;
   

−2 1 3 1 2 −2

1 1 −1 1 3 1
   

C = 1 2 0 ; D =  3 0 1 .
   

−1 0 3 1 1 1
Change of basis

Let (V , ·) be an Euclidean space, and consider two bases B and B 0 of V .

If G is the Gram matrix of the inner product w.r.t. B, whereas H is the
one w.r.t. B 0 , then
H = P t GP,
where P is the change-of-basis matrix corresponding to changing from B
to B 0 .

Example 6.6
Consider the following inner product:

(x1 , x2 ) · (y1 , y2 ) = 2x1 y1 + x2 y1 + x1 y2 + x2 y2 for (x1 , x2 ), (y1 , y2 ) ∈ R2 .

Find its Gram matrices first for the standard basis and then for the basis
B = {(1, 1), (−1, 1)}.
Contents

1 Euclidean vector space.

2 Matrix associated to an inner product.

3 Metric properties.

4 Orthogonality.

5 Orthogonal diagonalization.
Length

Definition 6.3
In an Euclidean space (V , ·), the length (or norm) of a vector ~v ∈ V is
the real number √ √
k~v k = ~v · ~v = ~v 2 .
Any vector ~v ∈ V whose length is 1 (i.e., k~v k = 1) is a unit vector.
For each vector ~v 6= ~0, its normalized vector (or associated unit vector,
or versor) is the unit vector codirectional with ~v :
~v
.
k~v k
Metric properties

Theorem 6.2 (Cauchy-Schwarz inequality)

~ in an Euclidean space (V , ·), it holds that
For all vectors ~v , w

|~v · w
~ | ≤ ||~v || ||~
w ||.

Corollary
In the above inequality, the two sides are equal if and only if ~v and w
~ are
linearly dependent.
Angle between two vectors
As an immediate consequence of Cauchy-Schwarz inequality,
~v · w
~
−1 ≤ ≤1 ~ ∈ V \ {~0}.
∀~v , w
||~v || ||~
w ||
This property gives sense to the following definition.
Definition 6.4
~ ∈ V \ {~0} is the unique real
The angle between two vectors ~v , w
number θ ∈ [0, π] such that

~v · w
~
cos(θ) = .
||~v || ||~
w ||

Remark 6.1
Multiplying both sides in the above equation by the lengths of the vectors,
~v · w
~ = ||~v || ||~
w || cos(θ) ∀~v , w
~ ∈ V.
The metric depends on the inner product

Example 6.7
Consider the inner product on R3 whose associated Gram matrix w.r.t. the
natural basis is  
1 1 −1
G = 1 2 −1  .
−1 −1 2
Compute the angle between (1, 0, 0) and (0, 1, 0).
Contents

1 Euclidean vector space.

2 Matrix associated to an inner product.

3 Metric properties.

4 Orthogonality.

5 Orthogonal diagonalization.
Orthogonal vectors

Definition 6.5
In an Euclidean space (V , ·), two vectors ~u , ~v are orthogonal, ~u ⊥ ~v , if
~u · ~v = 0.

Theorem 6.3
In an Euclidean space (V , ·), consider an indexed set of vectors
S = {~v1 , ~v2 , . . . , ~vn } ⊂ V \ {~0}. If ~vi ⊥ ~vj for all i 6= j, then S is linearly
independent.
Orthogonal projection

Theorem 6.4
Consider an Euclidean space (V , ·) and a vector ~v ∈ V \ {~0}. Then, each
vector ~u ∈ V can be written uniquely in the form ~u = α~v + w~ with α ∈ R
~ ∈ V such that w
and w ~ ⊥ ~v .

Example 6.8
Let us consider R3 endowed with the usual inner product. Denoting
~v = (2, −3, 1), write the vector ~u = (1, 3, 5) as the sum of a vector
proportional to ~v and another orthogonal to it.
Orthogonal and orthonormal bases

Motivation:
Let us consider an Euclidean space (V , ·).
We look for a basis of V for which the Gram matrix associated to the
inner product is as simple as possible: a diagonal matrix or, even better,
the identity.
Thus, we search for a basis of V ,

B = {~v1 , ~v2 , . . . , ~vn } ,

such that 
 1 if i = j,
~vi · ~vj =
0 if i 6= j.

Orthogonal and orthonormal bases

Definition 6.6
In a finite-dimensional Euclidean space, a basis

B = {~v1 , ~v2 , . . . , ~vn }

is an orthogonal basis if ~vi ⊥ ~vj = 0 for all i 6= j. If, in addition, each

vector ~vi is a unit vector, B is an orthonormal basis.

Notice that the Gram matrix w.r.t. an orthogonal basis is diagonal,

whereas w.r.t. an orthonormal basis it is the identity.
Gram-Schmidt process
Let (V , ·) be an Euclidean space and B = {~e1 ,~e2 , . . . ,~en } a basis of V .

We can obtain an orthogonal basis B 0 = {~v1 , ~v2 , . . . , ~vn } as follows:

For p = 1, 2, . . . , n, we write ~ep as a linear combination of ~v1 , ~v2 , . . . , ~vp ,

~ep = α1~v1 + α2~v2 + · · · + αp−1~vp−1 + ~vp ,

and impose that ~vp ⊥ ~vi for all i ∈ {1, 2, . . . , p − 1}; or, equivalently,

~ep · ~vi
αi = ∀i = 1, 2, . . . , p − 1 .
~vi · ~vi

Moreover, from the orthogonal basis B 0 = {~v1 , ~v2 , . . . , ~vn }, we can obtain
the orthonormal basis
 
~v1 ~v2 ~vn
, , ..., .
k~v1 k k~v2 k k~vn k
Example

Example 6.9
Consider the vector space R3 endowed with the inner product whose Gram
matrix w.r.t. the natural basis is
 
1 1 −1
G = 1 2 −1  .
−1 −1 2

Apply the Gram-Schmidt process to deduce an orthonormal basis of R3 .

Orthogonal subspaces

Definition 6.7
Two subspaces U, W of an Euclidean space are orthogonal subspaces,
U ⊥ W , if ~u ⊥ w~ for all ~u ∈ U and w~ ∈ W.
In particular, a vector ~u is orthogonal to W if ~u ⊥ w ~ ∈ W.
~ for all w

Theorem 6.5
Let (V , ·) be an Euclidean space and U a subspace of V . A vector ~b ∈ V
is orthogonal to U if and only if it is orthogonal to each vector of a basis
of U.
The orthogonal complement

Definition 6.8
Let U be a subspace of an Euclidean space (V , ·). The orthogonal
complement of U is the set of all vectors that are orthogonal to those of
U,
U ⊥ = {~w ∈V ; w ~ ⊥ ~u ∀~u ∈ U} .

Theorem 6.6
If U is a subspace of an Euclidean space (V , ·), then also U ⊥ is a subspace
of V and it satisfies
n o
U ∩ U ⊥ = ~0 .
Orthogonal projection

Let us consider (V , ·) an Euclidean space (not necessarily

finite-dimensional), U a vector subspace of V with dim(U) = m, and
~v ∈ V a given vector.
We look for a vector ~u ∈ U such that ~v − ~u is orthogonal to U:

~v − ~u ∈ U ⊥ .

In this case, ~u is called the orthogonal projection of ~v onto U, and it is

denoted by ~u = projU (~v ) or ~u = πU (~v ).

Theorem 6.7
Under the above conditions and when ~v ∈
/ U,

k~v − w
~ k > k~v − projU (~v )k ∀~
w ∈ U with w
~ 6= projU (~v ).
Orthogonal projection

How to find projU (~v):

Let B = {~e1 ,~e2 , · · · ,~em } be a basis of U.
By definition, ~u = projU (~v ) ∈ U, so that it must be of the form

~u = c1~e1 + c2~e2 + ... + cm~em .

Moreover, ~v − ~u should be orthogonal to U:

~v − (c1~e1 + c2~e2 + ... + cm~em ) ∈ U ⊥ ;

equivalently

(~v − (c1~e1 + c2~e2 + ... + cm~em )) ⊥ ~ej ∀j = 1, 2, . . . , m.

Imposing these m conditions leads to a linear system that characterizes

c1 , c2 , . . . , cm .
Example

Example 6.10
Consider the vector space C([0, 1]) endowed with the inner product
Z 1
p·q = p(x) q(x) dx ∀p, q ∈ C([0, 1]).
0

Find the orthogonal projection of the function f (x) = sin(πx) ∈ C([0, 1])
onto the subspace R2 [x].
Example

Orthogonal projection of sin(πx) on R2[x]

Contents

1 Euclidean vector space.

2 Matrix associated to an inner product.

3 Metric properties.

4 Orthogonality.

5 Orthogonal diagonalization.
Symmetric operators

Definition 6.9
Let (V , ·) be an Euclidean space, and T : V → V an endomorphism of V .
The operator T is symmetric if

~u · T (~v ) = T (~u ) · ~v ∀~u , ~v ∈ V .

If T is a symmetric endomorphism of V , and G and A denote the matrices

associated to · and T w.r.t. a basis B, then

GA = At G = (GA)t .

In particular, if B is orthonormal, then G = I and, therefore, A is

symmetric.
Example

Example 6.11
Consider the vector space R3 endowed with the inner product whose Gram
matrix for the natural basis is
 
2 2 2
G =  2 4 0 .
2 0 3

Let T be the endomorphism of V whose matrix for the natural the

canonical basis is  
1 2 a
A=  2 1 b .
2 2 3
Find the values of a, b ∈ R for which T is symmetric.
Basic properties of symmetric operators

Theorem 6.8
The characteristic polynomial of a symmetric endomorphism has the form

p(λ) = (λ1 − λ)n1 (λ2 − λ)n2 · · · (λp − λ)np with λi ∈ R.

Consequence: every symmetric endomorphism is diagonalizable over R.

Theorem 6.9
Let (V , ·) be an Euclidean space, and T : V → V a symmetric
endomorphism. For any pair of different eigenvalues of T , the associated
corresponding eigenspaces are orthogonal: S(α) ⊥ S(β) for all α 6= β.

Corollary
Every symmetric endomorphism is diagonalizable in an orthonormal basis.
Such diagonalization is called orthonormal diagonalization.
Example

Example 6.12
Consider the inner product · : R3 × R3 −→ R whose Gram matrix for the
standard basis is  
2 −3 3
G =  −3 5 −4  ,
3 −4 6
and let T denote the endomorphism of R3 whose associated matrix for the
standard basis is  
3 α 0
M(T ) =  0 −1 0  .
−2 4 −1
Find the values of α ∈ R for which T is symmetric. For such values,
diagonalize T in an orthonormal basis.