Free extrema of two

variables functions

Apellidos, Nombre: Alicia Herrero Debón (aherrero@mat.upv.es)

Departamento: Departamento de Matemática Aplicada

Instituto de Matemática Multidisciplnar

Centro: Escuela Técnica Superior de Ingenierı́a del Diseño

Universitat Politècnica de València
1. Abstract and objectives
In this paper we work with functions of two variables and introduce the
concept of relative (free) extrema for this kind of functions. That is, we
calculate the maximum and minumum value of a function of two variables in
its domain of definition. This represents a generalization of the same concept
as in one variable functions.
Once studied this paper the student will be able to obtain the relative
extrema of a function of two variables and to classify them.

2. Introduction
In many situations it is important to know where a function reaches a local
maximum or minimum value. For instance, if we consider the temperature
function, it can be interesting to know in which points the temperature is
higher or lower than at any point near them. This is what we are going to
consider in this work.
It is known that a one variable function has a local maximum or mi-
nimum when the growing behavior changes from increasing to decreasing
(maximum) or from decreasing to increasing (minimum). These situations
can be characterized using the first derivative of the function. In the same
way we generalize the concept of maximum and minimum value of a function
to functions of several variables.

Then, we are going to study what a local (relative) extreme of a two

variables function is, how to calculate the relative extrema and how to classify
them to know if they are maximum or minimum points of the function.

1. Free extreme points

Firstly, we introduce the concept of relative or free maxima and minima
values of a function of two variables pointing out the difference between local

1
and absolute extrema as happens in one variable functions.

Definition 1 A real function f (x, y) has

a relative maximum at the point (x0 , y0 ) if there exists δ > 0 such

that ∀(x, y) ∈ D with |(x, y) − (x0 − y0 )| < δ it is satisfied that

f (x, y) ≤ f (x0 , y0 )

a relative minimum at the point (x0 , y0 ) if there exists δ > 0 such

that ∀(x, y) ∈ D with |(x, y) − (x0 − y0 )| < δ it is satisfied that

f (x, y) ≥ f (x0 , y0 )

an absolute or global maximum at the point (x0 , y0 ) if f (x, y) ≤

f (x0 , y0 ) for any point (x, y) of its domain of definition.

an absolute or global minimum at the point (x0 , y0 ) if f (x, y) ≥

f (x0 , y0 ) for any point (x, y) of its domain of definition.

In this work we are interested in the obtaining of the relative maxima

and minima of a function of two variables.

Remark 1 The relative extreme points of a function are the points where
the function has a relative maximum or minimum. They are also called free
extreme points.

Remark 2 Note that the above definitions are also valid for functions of
several variables in general.

As in one variable functions, derivatives play an important role to study

the relative extrema of a function z = f (x, y). So, we can define the critical
points as follows.

Definition 2 A point (x0 , y0 ) is called a critical point of z = f (x, y) if

the first order partial derivatives of f (x, y) at that point are equal to zero or
don’t exist:
∂f
(x0 , y0 ) don’t exist or equals 0 and
∂x

2
 ∂f
(x0 , y0 ) don’t exist or equals 0.
∂y

This concept can also be generalized to functions of more than two va-
riables. In fact, a point x~0 ∈ Rn is a critical point of a function f (~x), with
~x ∈ Rn , if all the first partial derivatives of the function at that point are
null or one of them doesn’t exist.

Theorem 1 Necessary condition

If the function f (x, y) has a relative maximum or minimum ar the point
(x0 , y0 ) then (x0 , y0 ) is a critical point of f (x, y).

Note that, as in one variable functions, the first order partial derivatives
can vanish for points that are no extrema. That is, not all critical points are
extrema of f (x, y).

Example 1 Obtain the critical points of f (x, y) = x3 − 3x2 + y 2 − 2xy

Solution: Calculate the first partial derivatives of f (x, y) and equal them
to zero.
∂f

2
 ∂ x = 3x − 6x − 2y = 0



∂f
= 2y − 2x = 0



∂y
From second equation, x = y and substituting at the firs equation

3x2 − 6x − 2x = 3x2 − 8x = x(3x − 8) = 0

So, x = 0 or x = 8/3. Then, the points (0, 0) and (8/3, 8/3) are critical
points of f (x, y).

In one variable functions, critical points can be a relative maximum, a

relative minimum or an inflection point. Something similar occurs for fun-
ctions of two variables. Look at the following graphs, where (0, 0) represents
a relative minimum, a relative maximum and a saddle point, respectively.

3
 In order to classify the critical points of a function of two variables we can
use a geometrical method (plot the function) or the second derivative test to
determine what kind of extrema is each critical point. The difference with
one variable function is that we don’t have a second derivative of the function
but four second partial derivatives. To consider all of them we construct a
square matrix with all of them and calculate its determinant.

Definition 3 The Hessian of the function f (x, y) at the point (x0 , y0 ) is

the determinant
∂2 f ∂2 f
(x ,
0 0y ) (x0 , y0 )
∂ x2 ∂x∂y
Hf (x0 , y0 ) = .
∂2 f ∂2 f
(x0 , y0 ) (x0 , y0 )
∂ x ∂y ∂ y2

Using the second partial derivatives and this determinant we obtain a way
of classifying the critical points to determine if they are minima, maxima or
saddle points.

Theorem 2 Sufficient condition

Let (x0 , y0 ) be a critical point of a function f (x, y) whose second partial
derivatives are continuous at (x0 , y0 ).

∂2 f
If Hf (x0 , y0 ) > 0 and (x0 , y0 ) > 0 then f (x0 , y0 ) is a relative mini-
∂ x2
mum.
∂2 f
If Hf (x0 , y0 ) > 0 and (x0 , y0 ) < 0 then f (x0 , y0 ) is a relative ma-
∂ x2
ximum.

If H(x0 , y0 ) < 0 then f (x, y) has a saddle point at (x0 , y0 ).

If H(x0 , y0 ) = 0 then this test is inconclusive.

There are other equivalent versions of the previous test. For example,
∂ 2f
some texts may use instead of the corresponding derivative of f (x, y)
∂y 2
with respect to x twice.

4
Example 2 Classify the critical points of f (x, y) = x3 − 3x2 + y 2 − 2xy
Solution: The first partial derivatives of f (x, y) are

∂f ∂f
= 3x2 − 6x − 2y and = 2y − 2x
∂x ∂y

Considering their nullity, the points (0, 0) and (8/3, 8/3) are critical points
of f (x, y).
Now, the hessian of the function at any point (x, y) is

6x − 6 −2
Hf (x, y) = = 12x − 8.
−2 2

For the first critical point one has Hf (0, 0) = −8 < 0, then the function
has a saddle point at (0, 0).
For the second critical point one has

∂ 2f 8 8
   
8 8
Hf , = 24 > 0 and , = 10 > 0,
3 3 ∂x2 3 3
 
8 8
then, the function has a relative minimum at the point , and its value
  3 3
8 8 256
is f , =− .
3 3 27

Example 3 Classify the critical points of f (x, y) = x3 + y 2 − 6xy − 39x +

18y + 20
Solution: Calculate the first partial derivatives of f (x, y) ,

∂f ∂f
= 3x2 − 6y − 39 and = 2y − 6x + 18,
∂x ∂y
and equal them to zero. From the second equation y = 3x − 9 and substituting
at the first equation

3x2 − 6(3x − 9) − 39 = 3x2 − 18x + 15 = 3(x2 − 6x + 5) = 0

So x = 5 or x = 1. Then, the points (5, 6) and (1, −6) are critical points of
f (x, y).

5
 Now, the hessian of the function at any point (x, y) is

6x −6
Hf (x, y) = = 12x − 36.
−6 2

For the first critical point one has

∂ 2f
Hf (5, 6) > 0 and (5, 6) = 30 > 0,
∂x2
then, the function has a relative minimum at the point (5, 6) and its value is
f (5, 6) = −86.
For the second critical point one has Hf (1, −6) < 0 and , then the function
has a saddle point at (1, −6).

3. Closing
We have studied relative extrema of a two variables function, how to
calculate and classify them. Extrema of a function are necessarily critical
points, so we can use the first partial derivatives of the function to find them.
Critical points are those points satisfying that the first partial derivatives of
the function are both null or at least one of them doesn’t exist. Once we have
the critical points we can use the second derivative test to determine if they
are maxima, minima or saddle points of the function. That is, we calculate
the hessian of the function at each critical point. If the hessian is negative
then the point represents a saddle point, if the hessian is positive then the
point is a maximum or a minimum. In some cases (when the hessian equals
zero) this test can not conclude what kind of extrema is the critical point.
In such cases we would have to use other methods (for example, geometrical
methods).
Definitions of relative maxima and minima can be extended to functions
with more than two variables. In this cases extrema are also critical points,
which can be obtained from the nullity of all the first partial derivatives of the
function or if some of them do not exist. Then, the Hessian is a determinant
of a square matrix with all the possible second derivatives (that is, of size n,
with n equal to the number of variables of the function). However the second
derivative test is different and it is not studied in this work.

6
4. Bibliography
There are many books studying several variables functions. Some of them
are:

1. Cálculo en varias variables. Jon Rogawski.

2. 5000 problemas de análisis matemático. B.P. Demidóvich.

3. Cálculo de una y varias variables con geometrı́a analı́tica, tomo 1. Sa-

turnino L. Salas.

4. Cálculo y geometrı́a analı́tica, vol. 2. Howard Anton.

5. Cálculo diferencial e integral, vol. 2. N. Piskunov.

6. Cálculo multivariable. James Stewart.

7. Numerical methods for mathematics, science, and engineering. John H.

Mathews.

8. Calculus. Ron Larson.

9. Calculus. James Stewart.