Lecture 1.

5: Dual vector spaces

Matthew Macauley

School of Mathematical & Statistical Sciences

Clemson University
http://www.math.clemson.edu/~macaule/

Math 8530, Advanced Linear Algebra

M. Macauley (Clemson) Lecture 1.5: Dual vector spaces Math 8530, Advanced Linear Algebra 1/7
Scalar functions
Let X be a vector space over a field K . A scalar function is any function from X to K .

A scalar function ` : X → K is linear if

`(x + y ) = `(x) + `(y ), for all x, y ∈ X ;

`(cx) = c`(x), for all x ∈ X , c ∈ K .

Or equivalently, if

`(c1 x1 + · · · + cn xn ) = c1 `(x1 ) + · · · + cn `(xn ), for all ci ∈ K , xi ∈ X .

Definition
The set of linear scalar functions ` : X → K is a vector space called the dual of X , and
denoted X 0 .
Addition and scalar multiplication is defined naturally:
Addition: (` + m)(x) := `(x) + m(x),
Scalar multiplication: (c`)(x) := c`(x).

M. Macauley (Clemson) Lecture 1.5: Dual vector spaces Math 8530, Advanced Linear Algebra 2/7
Examples of scalar functions
Example 1
Let X = C([0, 1], R), the continuous functions [0, 1] → R, and fix t1 , . . . , tn ∈ [0, 1]. The
following are linear scalar functions:
`(f ) = f (t1 );
n
X
`(f ) = ai f (ti ), ai ∈ R;
i=1
Z 1
`(f ) = f (t) dt.
0

Example 2
Let X = C ∞ (R) be the set of smooth functions R → R. For a fixed t0 ∈ R,
n n
X di X di f
` := ai , ` : f 7−→ ai
i=1
dt i t=t0
i=1
dt i t=t0

is a linear scalar function (i.e., an element of X 0 ).

M. Macauley (Clemson) Lecture 1.5: Dual vector spaces Math 8530, Advanced Linear Algebra 3/7
The dual space
If dim X = n, then X ∼ = K n . Thus, we can associate a vector x ∈ X with an n-tuple
x = (c1 , . . . , cn ) of scalars.

For any fixed a1 , . . . , an ∈ K , the function

` : X −→ K , `(x) = a1 c1 + · · · + an cn (1)

is linear, i.e., ` ∈ X 0 .

Theorem 1.8
If dim X = n < ∞, then every ` ∈ X 0 can be written as in Eq. (1).

Proof

M. Macauley (Clemson) Lecture 1.5: Dual vector spaces Math 8530, Advanced Linear Algebra 4/7
The dual space
Corollary 1.9
If dim X < ∞, then X ∼
= X ’.

One way to think of this is to:

1. associate a vector x ∈ X with a column vector,
2. associate a scalar function ` ∈ X 0 with a row vector.

Notation
A linear function ` ∈ X 0 applied to a vector x ∈ X depends on the n-tuples (c1 , . . . , cn ) for x
and (a1 , . . . , an ) for `. We can use scalar product notation

(`, x) := `(x).

Sometimes, elements ` ∈ X0 are called co-vectors, or dual vectors.

Definition
Let x1 , . . . , xn be a basis for X . The dual basis in X 0 is `1 , . . . , `n , where
(
1 i =j
(`i , xj ) =
0 i 6= j.

Think of `i as the function that “picks off” the coefficient of xi .

M. Macauley (Clemson) Lecture 1.5: Dual vector spaces Math 8530, Advanced Linear Algebra 5/7
Duality in infinite dimensional spaces
Consider the vector space
n ∞ o
X = `1 (R) :=
P
(x1 , x2 , . . . ) | xi ∈ R, |xi | < ∞ .
i=1

Given vectors y = (a1 , a2 , . . . ) and x = (c1 , c2 , . . . ),

∞
X
(y , x) = ai ci < ∞,
i=1

so every y ∈ X defines a co-vector in X 0 .

But there are others! If z = (1, 1, 1, . . . ),

∞
X
(z, x) = ci < ∞,
i=1

but z 6∈ X .

M. Macauley (Clemson) Lecture 1.5: Dual vector spaces Math 8530, Advanced Linear Algebra 6/7
The double dual
The scalar product (`, x) is a bilinear function of ` and x. That is, if we fix one argument, it
is linear in the other. Equivalently,

(a`, x) = a(`, x) = (`, ax) for all x ∈ X , ` ∈ X 0 , a ∈ K .

| {z } | {z }
=a`(x) `(ax)

If dim X = n < ∞, then every linear scalar function X → K is of the form

(`, x), for some fixed ` = (a1 , . . . , an ) ∈ K n .

Since X 0 is a vector space, it has a dual, called the double dual of X , and denoted
X 00 := (X 0 )0 . Every linear scalar function X 0 → K is of the form

(`, x), for some fixed x = (c1 , . . . , cn ) ∈ K n .

Key points
Let x1 , . . . , xn be a basis of X
Think of the dual basis `1 , . . . , `n as “pick-off functions”
Think of elements in the double dual as “evaluation functions”
The bilinear function (`, x) naturally identifies X 00 with X .

M. Macauley (Clemson) Lecture 1.5: Dual vector spaces Math 8530, Advanced Linear Algebra 7/7