Natalia Lazzati

Mathematics for Economics

Note 1: Linear Algebra

Note 1 is based on Searle and Willett (2001) and Simon and Blume (1994, Ch. 6, 7, 8, 9, 11, 26
and 27).

Although most of the models economists are interested in are nonlinear in nature, linear
algebra plays a fundamental role in economic theory. The reason is that to study nonlinear
systems we often use the meta-principle of calculus: to study the behavior of the solutions to a
nonlinear system we examine the behavior of a closely related system of linear equations!
Most of the chapters in Simon and Blume (1994), SB, were assigned to read during summer,
so the main purpose of this note is to collect (and connect) the main ideas in a single …le. (The
notation we use is slightly di¤erent from the one in that book.)

Matrix Algebra

De…nition of Matrix

A matrix is a rectangular array of elements arranged in rows and columns. All rows are of equal
length, as well as all columns. It is a devise frequently used in organizing the presentation of
numerical data so that they may be handled with ease mathematically. In this sense we can
think of matrix algebra as a vehicle by which mathematical procedures for many problems, both
large and small, can be described independently of the size of the problem. The use of matrices
is fundamental in many areas of economics, for instance, in econometrics.
In terms of notation aij indicates the element in the ith row and jth column of a matrix. If

1
a matrix A has m rows and n columns then it can be represented as
2 3
6 a11 a12 ::: a1j ::: a1n 7
6 7
6 a21 a22 ::: a2j ::: a2n 7
6 7
6 . .. .. .. .. .. 7
6 .. . . . . . 7
6 7
A=6 7:
6 a ai2 ::: aij ::: ain 7
6 i1 7
6 . .. .. .. 7
6 . .. .. 7
6 . . . . . . 7
4 5
am1 am2 ::: amj ::: amn

An alternative representation is

A = faij g for i = 1; 2; :::; m and j = 1; 2; :::; n:

Basic Operations

Pages 153-157 in SB cover …ve basic matrix operations: addition, substraction, scalar multipli-
cation, matrix multiplication and transposition. These pages also describe the laws of matrix
algebra: associative, distributive and commutative. All these operations appear quite often in
both statistics and econometrics (so you should try to remember them).

Special Matrices

Several matrices with particular properties receive speci…c names. Historically, each of them
probably originated in a mathematical problem or application.
Some simple special forms are: square, diagonal, identity, lower and upper triangular, sym-
metric and idempotent matrices (see SB, pp. 160-161). A matrix which is just a single column is
called a column vector. Similarly, a row vector is a matrix which is just a single row. In terms of
notation we will indicate a column vector by x, and a row vector by xT : A scalar is just a matrix
of dimension 1 1, that is, a single element.
A square matrix is nonsingular if its rank equals the number of its rows (or columns). (Recall
that the rank of a matrix is the number of nonzero rows in its raw echelon form.) When such
a matrix arises as a coe¢ cient matrix in a system of linear equations, then the system has one
and only one solution. So nonsingular matrices are of fundamental importance, and we will o¤er
other characterizations of this property.

2
Determinants

De…nition of Determinant

Determinants are important in the analysis of mathematical models. For example, they help to
de…ne if a system of linear equations has a solution, to compute the solution when it exists, and
to determine whether a given nonlinear system can be well approximated by a linear one. They
are also important to test the su¢ cient conditions of optimization problems.
A determinant is just a scalar which is the sum of selected products of the elements of the
matrix from which it is derived. It is de…ned only for square matrices— the determinant of a
non-square matrix is unde…ned and does not exist. The customary notation for the determinant
of the matrix A is just jAj :
Let A be an n n matrix, and let Aij be the (n 1) (n 1) matrix obtained from A by
deleting the ith row and jth column. The determinant of Aij , Mij = jAij j, receives the name of
(i; j) minor of A. In addition Cij = ( 1)i+j Mij is the (i; j) cofactor of A:
The determinant of a 1 1 matrix is the value of its sole elements. The determinant of an
n n matrix A can be de…ned, inductively, as follows
Pn Pn
jAj = j=1 a1j ( 1)1+j M1j = j=1 a1j C1j : (1)

This method of expanding a determinant is known as expansion by the elements of the …rst row.
The selection of the …rst row is arbitrary, as we can in fact expand the determinant of A by using
any row or any column. (Then you should choose the one with more zeros! Why?)

Theorem 1. If we de…ne the determinant of A as in (1), then for any i or j,

Pn
jAj = j=1 aij ( 1)i+j Mij (2)
Pn
= i=1 aij ( 1)i+j Mij (3)

where (2) expands the determinant through the ith row and (3) uses the jth column.

Proof. See SB p. 743.

Corollary 2. For any n n matrix A, jAj = AT :

Proof. This result follows as a simple corollary of Theorem 1.

3
Properties of the Determinant

Determinants have some useful properties. The next lines assume A is a square matrix.

(a) If we add to any row of A a multiple of another row, then the determinant does not change;

(b) if we form matrix B by multiplying any row of A by a scalar r, then jBj = r jAj ;

(c) if we form matrix B by interchanging two rows of A then jBj = jAj ;

i.e., they are linearly dependent
(d) if two rows of A are the same, then jAj = 0;

(e) if A is lower or upper triangular, then its determinant is just the product of the elements
of its principal diagonal; and

(f) if R is the row echelon form of A, then jRj = jAj :

(Show each of the previous properties. These properties are in terms of rows, explain
why the same results hold for the columns.)
Properties (a)-(f) are quite useful to justify two fundamental theorems that we discuss next.
The …rst one states that the determinant of a matrix de…nes whether or not the matrix is singular.

Theorem 3. A square matrix A is nonsingular if and only if jAj =

6 0:

Proof. By the last property of a determinant, (f), if R is the row echelon form of matrix A then
jRj = jAj : If A is nonsingular then R has all its diagonal elements di¤erent from zero. Since R
is upper diagonal then the theorem follows by the …fth property of a determinant, (e).

The second theorem shows that for two arbitrary n n matrices A and B, the determinant
of the product is the same as the product of their determinants.

Theorem 4. For arbitrary n n matrices A and B we get jABj = jAj jBj.

Proof. See SB p. 733.

4
Inverse Matrix

In matrix algebra there is no division; we never divide a matrix. Instead, under certain circum-
stances we can multiply by what is known as an inverse matrix. This operation between matrices
is the counterpart of division. We next de…ne inverse matrices, the conditions for their existence
and some important applications.
The inverse of a matrix A is a matrix B that satis…es

BA = I = AB: (4)

We will indicate the inverse of A, when it exists, as A 1. From the last de…nition, it follows that
a necessary condition for A 1 to exist is that A be a square matrix. Moreover, A and A 1 will
share the same dimension, i.e. if A is an n n matrix, so is A 1:

Given that A must be square for A 1 to possibly exist, we can use (4) again to derive the
conditions under which A 1 actually exists. First take the determinant of both sides of (4)
1
A A = jIj : (5)

By the product rule

1
A jAj = 1: (6)

The important consequence is that jAj =

6 0 for the inverse of A to exist: Then A 1 can exists if
A is square, and it does exists if in addition jAj =
6 0:

Theorem 5. If A 1 exists, then it is unique.

Proof. (Show this result.)

Let us de…ne adjA = fCij g, where Cij is the (i; j) cofactor of A (as described before): The
inverse of A, when it exists, can be obtained as follows
1
A 1
(adjA)T :
= (7)
jAj
0 1
a x
Example 1. Consider a matrix A = @ A. Its determinant is jAj = ab xy: Then the
y b
inverse exists if and only if ab xy 6= 0: If the latter condition holds, then
2 3
1 b x
A 1= 4 5: N
(ab xy) y a

5
Linearly (In)dependent Vectors

This section and the following one relate to certain mathematical concepts and relationships
concerning individual rows (and/or columns) of a matrix. They provide methods for ascertaining
if a square matrix A has a zero determinant. These methods are easier than calculating the
determinant, and will shed extra light later on the characterization of solutions of systems of
equations.

Linear Combination of Vectors

If x1 ; x2 ; :::; xn denote n vectors of the same dimension, then for scalars a1 ; a2 ; :::; an ;

a1 x1 + a2 x2 + ::: + an xn (8)

is called linear combination of the set of n vectors.

Example 2. Let 2 3 2 3
1 3
x1 = 4 5 and x2 = 4 5:
2 7
A linear combination of the two vectors is given by
2 3 2 3 2 3
1 3 a1 + 3a2
a1 4 5 + a2 4 5 = 4 5
2 7 2a1 + 7a2

for some scalars a1 and a2 : N

Note that the …nal line in the last example can be written as
2 3 2 32 3 2 3
a1 + 3a2 1 3 a1 h i a1
4 5=4 54 5 = x1 x2 4 5 = Xa:
2a1 + 7a2 2 7 a2 a2

This is true in general. If

h i
X= x1 x2 ::: xn and aT = (a1 ; a2 ; :::; an ) (9)

then the linear combination in (8) is a vector that can be written as

Pn
a1 x1 + a2 x2 + ::: + an xn = i=1 ai xi = Xa: (10)

6
As a consequence, given a matrix X and a vector a for which Xa exists, Xa is a column vector,
i.e. a linear combination of the columns of X. Similarly, if we let bT = (b; :::; bm ) then bT X is a
row vector, i.e. a linear combination of the rows of X: Moreover, AB is a matrix with rows that
are linear combinations of the rows of B, and its columns are linear combinations of the columns
of A.

De…nition of Linear (In)dependence

As we explained before, the product

Xa = a1 x1 + a2 x2 + ::: + an xn (11)

is a vector. If there exists a vector a 6= 0, such that a1 x1 + a2 x2 + ::: + an xn = 0, then provided

none of x1 ; x2 ; :::; xn is null, those vectors are said to be linearly dependent vectors. An alternative
statement of the de…nition is: if Xa = 0 for some non-null a, then the columns of X are linearly
dependent vectors, provided none is null.

Example 3. Let 2 3 2 3
1 2
x1 = 4 5 and x2 = 4 5:
2 4
Then 2 3 2 3 2 3
1 2 0
24 5 + ( 1) 4 5=4 5:
2 4 0

Since neither x1 nor x2 are null vectors and aT = (2; 1) 6= 0, it follows that these two vectors
are linearly dependent. N

If on the contrary a = 0 is the only vector for which a1 x1 + a2 x2 + ::: + an xn = 0; then

provided none of x1 ; x2 ; :::; xn is null, those vectors are said to be linearly independent (LIN)
vectors. An alternative statement is: if Xa = 0 only for a = 0, then the columns of X are LIN
vectors. (Show that the two vectors in Example 2 are LIN.)
Assume that x1 ; x2 ; :::; xn are linearly dependent, and suppose that a1 and a2 are non-zero
(this arbitrary selection is without loss of generality). Then

x1 + (a2 =a1 ) x2 + ::: + (an =a1 ) xn = 0 (12)

7
so that
x1 = (a2 =a1 ) x2 ::: (an =a1 ) xn

which means that x1 can be expressed as a linear combination of the other x0 s: Therefore, the
notion of dependent vectors relates to the possibility of expressing some of them as the linear
combinations of some others. It is the existence of such combinations that is important rather
than the speci…c scalars that multiply the vectors in those combinations!
The next theorem captures an important result related to LIN vectors.

Theorem 6. A set of LIN vectors of order n never contains more than n vectors.

Proof. Let u1 ; u2 ; :::; un be n LIN vectors of order n. Let u be other non-null vector of order
n. We show that u and u1 ; u2 ; :::; un are linearly dependent.
Since U = [u1 ; u2 ; :::; un ] has LIN columns, jU j 6= 0 and U 1 exists. Let q = U 1u 6= 0,
because u = U q 6= 0. Then U q + u = 0, which can be written as
2 3 2 3
h i q q
U u 4 5 = 0 with 4 5 6= 0:
1 1

Therefore u1 ; u2 ; :::; un ; u are linearly dependent. Then there is no other vector u that can be
added to the matrix U with the property that the n + 1 vectors be LIN.

Zero Determinants and Inverse

Suppose that n linearly dependent vectors of order n are used as either columns or rows of a
matrix. Then the determinant is zero. (Show this statement using the properties of the
determinants.) Therefore if a square matrix of order n has linearly dependent columns or rows
its inverse does not exist.

Example 4. Let 2 3
3 0 2
h i 6 7
6 7
x1 x2 x3 =6 6 5 1 7:
4 5
9 5 1
Note that x1 = (3=2) x2 + (3=2) x3 . So substracting (3=2) x2 + (3=2) x3 from x1 we get a
new matrix with all 0’s in its …rst column. Since this operation does not change the determinant

8
(Why?), then 2 3
0 0 2
6 7
6 7
x1 x2 x 3 = 6 0 5 1 7 =0
4 5
0 5 1
by expanding the determinant through the …rst column. N

Rank

De…nition of Rank and Inverse

The rank is an extremely important and useful characteristic of every matrix. It concerns its
number of LIN rows (and columns).
In the previous section we stated that a determinant of a matrix is zero when either any of its
rows is a linear combination of other rows or any of its columns is a linear combination of other
columns. Then a square matrix cannot have simultaneously LIN rows and linearly dependent
columns. A natural question arises: What is the relationship between the number of LIN rows
and LIN columns in a matrix? The next theorem provides an answer.

Theorem 7. The number of LIN rows in a matrix is the same as the number of LIN columns.

Proof. See Searle and Willett (2001), pp. 162-163.

The last theorem states that every matrix has the same number of LIN rows and LIN columns.
This result guarantees the consistency of the next de…nition.

De…nition 8. (Rank) The rank of a matrix is the number of LIN rows (and columns) in the
matrix.

In terms of notation we will write rA for the rank of matrix A: An alternative way to de…ne
rA is by the number of non-zero rows in the row echelon form of matrix A: This de…nition o¤ers
a procedure to …nd the rank of any given matrix, but is less transparent with respect to the
meaning of the concept.
Table 1 summarizes the relationships between matrix inverse, rank and linearly independence.

9
 Table 1: Equivalent Statements for a Square Matrix A of order n
Inverse Existing Inverse Non-Existing
A 1 exists A 1 does not exist
jAj =
6 0 jAj = 0
rA = n rA < n
A has n linearly independent rows A has fewer than n linearly independent rows
A has n linearly independent columns A has fewer than n linearly independent columns

Spanning sets, Bases and Subspaces

The space Rn is a special case of a wider concept called vector space. We say a set S Rn is a
subspace of Rn if the next two conditions hold

(a) for every pair xi and xj such that xi 2 S and xj 2 S, we also …nd that xi + xj 2 S; and

(b) for every real scalar a and xi 2 S, we also …nd that axi 2 S:

Suppose that every vector in S can be expressed as a linear combination of the t vectors
x1 ; x2 ; :::; xt : Then this set of t vectors is said to span or generate S, and it is called a spanning
set for S. When the vectors of a spanning set are also LIN then the set is said to be a basis for
S. It is a basis and not the basis because for any S there may be many bases. All bases of a
subspace S have the same number of vectors and that number is called the dimension of S, or
dim(S).

Example 5. Let 2 3 2 3 2 3
1 1 3
6 7 6 7 6 7
6 7 6 7 6 7
x 1 = 6 2 7 , x2 = 6 1 7 and x3 = 6 0 7
4 5 4 5 4 5
0 0 0
be three vectors in R3 . Any two of them form a basis for the vector space in R3 that has 0 as the
last element. The dimension of the space is 2. This space is in fact a subspace of R3 : N

(How many vectors form a basis for S = Rn ? Provide one simple basis.)
For an m n matrix A, the rows of A have n components. We write Row(A), row space of
A, for the subspace of Rn spanned by the m rows of A. If r1 ; r2 ; :::; rm 2 Rn are the m vectors

10
that form the rows of A, then

Row(A) fa1 r1 + a2 r2 + ::: + am rm : a1 ; a2 ; :::; am 2 Rg :

By de…nition, dim[Row(A)] = rA :
Just as we formed the row space of a given matrix A as the set spanned by the rows of A,
we can do the same for the columns of A. If A is an m n matrix, then its columns have m
components. We write Col(A), column space of A, for the subspace of Rm spanned by the n
columns of A. If c1 ; c2 ; :::; cn 2 Rm are the n columns of A, then

Col(A) fa1 c1 + a2 c2 + ::: + an cn : a1 ; a2 ; :::; an 2 Rg :

By Theorem 7 the number of LIN rows of a matrix equals the number of LIN columns, then
dim[Col(A)] = dim[Row(A)] = rA : This is not the same as to say Col(A) is equal to Row(A):
(Provide an example illustrating this claim.)

System of Linear Equations

All previous results are extremely important in the analysis of systems of linear equations.

De…nition of System of Linear Equations and Characterization of the Solu-

tion(s)

Linear equations in several unknowns can be represented as

Ax = (13)

where x is the vector of unknowns, is a vector of known values and A is the matrix of coe¢ cients.
In economics we often refer to x as the endogenous variables and to as the exogenous ones
(or parameters). While the former are determined within the model, the latter are exogenously
given.

Example 6. Let 2 3 2 3 2 3
a11 a12 x1 1
A=4 5, x = 4 5 and =4 5:
a21 a22 x2 2

11
Then (13) takes the form of either
2 32 3 2 3
a11 a12 x1 1 a11 x1 + a12 x2 = 1
4 54 5=4 5 or :N
a21 a22 x2 2 a21 x1 + a22 x2 = 2

System (13) can also be written as

x1 c1 + x2 c2 + ::: + xn cn = (14)

where cj is the jth column of A, for j = 1; :::; n: When the system of linear equations is expressed
in this way, it is apparent that Ax = has a solution if and only if 2 Col(A), i.e. if belongs to
the space generated by the columns of the coe¢ cient matrix. Therefore, the statement [Ax =
has a solution for all 2 Rm ] is equivalent to the statement [Col(A) is all of Rm ]. The next
theorem captures these observations.

Theorem 9. Let A be an m n matrix, then

(a) Ax = has a solution for a particular 2 Rm if and only if 2 Col(A); and

(b) Ax = has a solution for every 2 Rm if and only if rA = m:

When the system of equations Ax = has a solution we say the system is consistent. Consis-
tent equations may have more that one solution. To describe the set of solutions of a consistent
system Ax = ; it is often easier to start studying the corresponding homogenous system Ax = 0:
The next theorem states that the solution set of a homogenous linear system of n variables is
a subspace of Rn :

Theorem 10. Let A be an m n matrix. The set V of solutions to the system of equations
Ax = 0 is a subspace of Rn :

Proof. To this end we need to show that V is closed under vector addition and multiplication.
Let u and v be vectors in V , and let ru + sv be a linear combination of the two vectors. Then

A (ru + sv) = Aru + Asv = rAu + sAv = 0 + 0 = 0

12
since u and v are in V: Thus V is closed under linear combinations and it is a subspace of Rn :

We next de…ne null-space of A.

De…nition 11. (Null-space) The subspace of solutions to the homogeneous system Ax = 0 is

called the null-space of A; and is written as Null(A) :

The solution sets to non-homogeneous equations are not subspaces, i.e. they do not have a
linear structure. They are a¢ ne subspaces, which is to say they are translates of a subspace.

De…nition 12. Let V be a subspace of Rn and let d 2 Rn be a …xed vector. The set

fx 2 Rn : x = d + v for some v 2 V g

is called the set of translates of V by d and is written d + V . Subsets of Rn of the form d + V ,

where V is a subspace of Rn , are called a¢ ne subspaces of Rn :

Theorem 13. Let Ax = be an m n system of linear equations. Let d in Rn be a solution

of this system. Then every other solution x of Ax = can be written as x = d + v where v is a
vector in the null-space of A. Then the solution set of Ax = is the a¢ ne subspace d+ Null(A) :

Proof. Let x0 solve Ax = : Then

A x0 d = Ax0 Ad = =0

so v = x0 d 2 Null(A) :
Conversely, if v 2 Null(A), then A (d + v) = Ad + Av = +0= .

The next theorem is often called Fundamental Theorem of Linear Algebra. It describes the
dimension of Null(A) in terms of rA .

Theorem 14. Let A be an m n matrix. Then

dim Null (A) = n rA :

Proof. See Searle and Willett (2001), pp. 178-179.

13
 Combining this result with Theorem 9 we get that when the system Ax = is consistent,
then the solution set is an a¢ ne subspace whose dimension equals the number of variables minus
the rank of A. (Use this comment to provide necessary and su¢ cient conditions in
terms of n; m and rA for Ax = to have one, and only one, solution for all 2 Rm :)

Example 7. Let the system of linear equations be

0 1
0 1 x1 0 1
1 1 2 B C 4
@ AB C
B x2 C = @ A: (15)
2 2 6 @ A 10
x3

Here m = 2, n = 3, and the rank of the coe¢ cient matrix is 2. (Why?) By Theorem 9(b)
the system (15) has a solution. To characterize the solution set let us study the corresponding
homogenous system. Notice that dimNull(A) = n rA = 1: Then the Null(A) is a line through
the origin in a three-dimensional space. This line can described by (a; a; 0) with a 2 R: In
addition, notice that d = (1; 1; 1) is a particular solution to system (15). By Theorem 13, the
solution set of (15) is given by all vector x that satis…es xT = (1; 1; 1) + (a; a; 0) with a 2 R: N

Crammer’s Rule for Linear Equations

Crammer’s rule is an easy method for solving linear equations. Assume A is a square matrix, and
let us write the system of equations Ax = as in (14)

x1 c1 + ::: + xj cj + ::: + xn cn = :

Crammer’s rule states that xj can be obtained as

h i
c1 ::: cj 1 cj+1 ::: cn
xj =
jAj
where the numerator for xj is just the determinant of a matrix derived from A by substituting
its jth column by .

Example 8. Consider a …rm that produces two outputs, y1 and y2 , by using only capital, K,
and labor, L. The production function is
0 12 3 2 3
a b K y
@ A4 5=4 1 5
c d L y2

14
where a; b; c and d are all strictly positive coe¢ cients. (Provide precise conditions under
which this system has a unique solution for all y1 and y2 :)
By Crammer’s rule
dy1 by2 ay2 cy1
K= and L = :N
ad cb ad cb

Linear Forms

At the beginning of this note, we de…ned a matrix as a rectangular array of elements arranged in
rows and columns. This section states that there is a one-to-one correspondence between linear
functions from Rn to Rm and m n matrices, i.e. each linear function f : Rn ! Rm is an fA for
a unique m n matrix A: So matrices are more than just rectangular arrays of elements, they
are also representations of linear functions.
There is a convenient way to think about functions of the form f : Rn ! Rm that we will
invoke often. We can de…ne mapping f as a vector of component functions fi , each of which is a
real-valued function of n variables

f = (f1 ; f2 ; :::; fm )T with fi : Rn ! R for i = 1; :::; m:

We next formalize the notion of linear function.

De…nition 15. (Linear function) A linear function from Rn to Rm is a function f that

preserves the vector space structure

f (x + y) = f (x) + f (y) and f (rx) = rf (x)

for all x and y in Rn and all r in R: Linear functions are sometimes called linear transformations.

An example of a linear function from Rn to Rm with m = 1 is the function

f (x) = aT x = a1 x1 + a2 x2 + ::: + an xn

for some vector aT = (a1 ; a2 ; :::; an ) in Rn : (Check this function satis…es De…nition 15.)
The next theorem states that every linear real-valued function is of the form f (x) = aT x:

Theorem 16. Let f : Rn ! R be a linear function. Then, there exists a vector a 2 Rn such
that f (x) = aT x for all x 2 Rn :

15
Proof. We develop the proof for n = 3: Let
0 1 0 1 0 1
1 0 0
B C B C B C
B C B C B C
e1 = B 0 C ; e2 = B 1 C ; e3 = B 0 C
@ A @ A @ A
0 0 1

be the canonical basis for R3 : Let ai = f (ei ) for i = 1; 2; 3; let aT = (a1 ; a2 ; a3 ) : Then, for any
vector x 2 R3 ;
0 1 0 1 0 1 0 1
x1 1 0 0
B C B C B C B C
B C B C B C B C
x = B x2 C = x1 B 0 C + x2 B 1 C + x3 B 0 C = x1 e1 + x2 e2 + x3 e3
@ A @ A @ A @ A
x3 0 0 1

and

f (x) = f (x1 e1 + x2 e2 + x3 e3 )

= f (x1 e1 ) + f (x2 e2 ) + f (x3 e3 )

= x1 f (e1 ) + x2 f (e2 ) + x3 f (e3 )

= aT x

which completes the proof. (Generalize the proof for an arbitrary n:)

Theorem 16 implies that every real-valued linear function on Rn can be associated with a
unique vector a 2 Rn so that
f (x) = aT x:

Conversely, every such a induces a linear map.

The same correspondence between linear functions and matrices carries over to functions from
Rn to Rm : If A is an m n matrix, then fA (x) = Ax is a linear function. The following theorem
states the converse.

Theorem 17. Let f : Rn ! Rm be a linear function. Then, there exists an m n matrix A such
that f (x) = Ax for all x 2 Rn :

Proof. See SB, p. 288.

16