RMIT Classification: Trusted

MATH 2394
Engineering Maths
Topic 6: Multivariable Functions

Course Coordinator: Dr. Hai Dao

Lecturer: Hai Dao

 RMIT Classification: Trusted

Today’s Topics
1. Multivariable Functions: Definition and what they look like.

2. Partial Differentiation: Find limits on multivariable surfaces and

derivatives of surfaces.

3. Applications of Multivariable Functions: Small changes approximation,

optimization

2
 RMIT Classification: Trusted

Introduction to Multivariable
Functions
Curvy surfaces
 RMIT Classification: Trusted

Multivariable Functions - Introduction

• In engineering and business applications, a parameter is

often influenced by 2 or more variables.

o Example 1: The temperature 𝑇 at a point on the surface of

the earth at any given time depends on the longitude 𝑥 and
latitude 𝑦 of the location. Hence, T is a function of 𝑥 and 𝑦.

o Example 2: The amount of money 𝑃 a person is willing to

spend on a product is often depend on their age 𝑎 and
income level 𝑖. Hence, P is a function of 𝑎 and 𝑖.
Source: Link
4
 RMIT Classification: Trusted

Multivariable Functions - Introduction

Global Temperature Map

Source: Link
5
 RMIT Classification: Trusted

Multivariable Functions - Introduction

• In order to model phenomenon or problems that involve 2

or more independent variables, we need Multivariable
Functions.

• A function 𝑓 of two variables is a rule that assigns to each

ordered pair of real numbers (𝑥, 𝑦) in a set 𝐷 a unique real
number denoted by 𝑓(𝑥, 𝑦).

• The set 𝐷 is the domain of 𝑓 and its range 𝑅 is the set of

values that 𝑓 takes on.
Source: Link
6
 RMIT Classification: Trusted

Multivariable Functions–Example 1

𝑥+𝑦+1
Example 1: Given the multivariable function 𝑓 𝑥, 𝑦 = , find the
𝑥−1

value 𝑓 3, 2 and the domain D of the function.

(1) Find the value of 𝑓 3, 2 :
3+2+1 6
𝑓 3, 2 = =
3−1 2

(2) To find the domain D of the function, we should

observe when 𝑓 𝑥, 𝑦 is undefined. There are 2
such locations:

a. When 𝑥 + 𝑦 + 1 < 0 → 𝑦 < −1 − 𝑥.

b. When 𝑥 = 1

(3) Hence the domain D includes all points on or

above the line 𝑦 = −𝑥 − 1 and 𝑥 ≠ 1. (Illustration)

7
 RMIT Classification: Trusted

Multivariable Functions–Example 1

𝑥+𝑦+1
Example: Given the multivariable function 𝑓 𝑥, 𝑦 = , find the value
𝑥−1

𝑓 3, 2 and the domain D of the function.

Domain D of the function

8
 RMIT Classification: Trusted

Multivariable Functions–Example 2

Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 9 − 𝑥 2 − 𝑦 2 , find the

Domain and Range of this function.
(1) To find the domain D of the function, we should
observe when 𝑓 𝑥, 𝑦 is undefined. There is 1
region where the function is undefined:
9 − 𝑥2 − 𝑦2 < 0
→ 9 < 𝑥2 + 𝑦2

Hence the domain D includes all points within the

circle with radius 3 defined by 𝑥 2 + 𝑦 2 ≤ 9.

(2) The range R of 𝑓(𝑥, 𝑦) are all possible value of

𝑓(𝑥, 𝑦) corresponding to the domain D. Hence the
range R is:
0, 9 − 0 − 0 → {0, 3}
(Illustration)

9
 RMIT Classification: Trusted

Multivariable Functions - Graphs

• A different way to visualize the behavior of a function of two

variables is to consider its graph.

• The next few slides contains the graph associated with

some Multivariable Functions.

Source: Link
10
 RMIT Classification: Trusted

Multivariable Functions - Graphs

𝑧 = 𝑓 𝑥, 𝑦 = 4𝑥 2 + 𝑦 2
11
 RMIT Classification: Trusted

Multivariable Functions - Graphs

𝑧 = 𝑓 𝑥, 𝑦 = sin 𝑥 + sin(𝑦) sin 𝑥 sin 𝑦

𝑧 = 𝑓 𝑥, 𝑦 =
𝑥𝑦

12
 Multivariable Functions RMIT Classification: Trusted

– Contour Map
• Another way to visualize functions with 2 variables is
Contour map.

• Contour map is a mapmaking techniques which utilizes

contour curves (i.e. level curves) to join points of constant
elevation.

• The level curves of a function 𝑓 of two variables are the

curves with equations 𝒇 𝒙, 𝒚 = 𝒌, where 𝑘 is a constant.

Source: Link
13
Multivariable Functions RMIT Classification: Trusted

– Contour Map

Example of Contour Map Geographical Contour Map

14
 Multivariable Functions– Contour
RMIT Classification: Trusted

Map Example
Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 6 − 3𝑥 − 2𝑦. Find the
contour line equations for 𝑘 = −6, 0, 6, 12.

(1) For the contour line at 𝑘 = −6:

−6 = 6 − 3𝑥 − 2𝑦
→ 2𝑦 = 12 − 3𝑥
𝟑
→𝒚=𝟔− 𝒙
𝟐
(2) For the contour line at 𝑘 = 0:
0 = 6 − 3𝑥 − 2𝑦
𝟑
→ 2𝑦 = 6 − 3𝑥 → 𝒚 = 𝟑 − 𝒙
𝟐
(3) Similarly for the other 𝑘 values:
𝟑
𝑘 = 6, ; 𝒚 = − 𝒙
𝟐
𝟑
𝑘 = 12, ; 𝒚 = −𝟔 − 𝒙
𝟐

15
 Multivariable Functions – Section
RMIT Classification: Trusted

Highlight
• Multivariable functions (MVFs) are functions with multiple
independent variables.

• The Domain of MVFs is the set of all possible 𝑥 and 𝑦

values which will allow the functions to yield real results.

• The Range of MVFs is the set of all real results of 𝑓(𝑥, 𝑦).

• Multivariable Functions can be illustrated with Surface

graphs or Contour maps.
Source: Link
16
 RMIT Classification: Trusted

Partial Differentiation

Derivatives in 3 Dimensions
 RMIT Classification: Trusted

Partial Differentiation - Introduction

• Previously, we learned about differentiation in 2D which is

to get the equations for slopes of tangent lines along a
curve:

′ ′
𝑓 𝑥+ℎ −𝑓 𝑥
𝑓 𝑥 = 𝑦 = lim
ℎ→0 ℎ

• For Multivariable Functions, 𝑥 and 𝑦 are independent

variables. Hence, we need a new definition for derivatives
of MVFs.
Source: Link
18
 Partial Differentiation – Limits of
RMIT Classification: Trusted

Multivariable Functions
• Given a function 𝑓(𝑥, 𝑦), we can define the limit of 𝑓(𝑥, 𝑦)
as 𝑥, 𝑦 approaches (𝑎, 𝑏) as 𝐿, such as:

𝐥𝐢𝐦 𝒇 𝒙, 𝒚 = 𝑳
𝒙,𝒚 → 𝒂,𝒃

• Similar to a limit in 2D, for the limit 𝐿 to exist, the function

𝑓(𝑥, 𝑦) must approach the same value no matter how (𝑥, 𝑦)
approaches (𝑎, 𝑏).

Source: Link
19
 RMIT Classification: Trusted

Partial Differentiation – Limit Example

𝑥𝑦
Example: Given the multivariable function 𝑓 𝑥, 𝑦 = . Determine if
𝑥 2 +𝑦 2

the limit lim 𝑓(𝑥, 𝑦) exist.

𝑥,𝑦 → 0 ,0 (1) Let’s set 𝑦 = 0:
0
𝑓 𝑥, 0 = 2 = 0
𝑥
(2) Let’s set 𝑥 = 0:
0
𝑓 0, 𝑥 = 2 = 0
𝑦
(3) So far we have check the limit along 2 lines and
they have reached the same number 𝐿 = 0.
However, let’s approach the limit from the line 𝑦 =
Hence the limit
𝒙𝒚 𝑥 = 0.01, 0.001:
𝐥𝐢𝐦 0.01 ∗ 0.01
𝒙,𝒚 → 𝟎,𝟎 𝒙𝟐 + 𝒚𝟐 𝑓 0.01, 0.01 = = 0.5
Does not exist. 0.012 + 0.012
0.001 ∗ 0.001
𝑓 0.001, 0.001 = = 0.5
0.0012 + 0.0012
20
 RMIT Classification: Trusted

Partial Differentiation – Continuity

• Similarly to the 2D case, a function 𝑓(𝑥, 𝑦) can be consider

at point (𝑎, 𝑏) if:

𝐥𝐢𝐦 𝒇 𝒙, 𝒚 = 𝒇(𝒂, 𝒃)
𝒙,𝒚 → 𝒂,𝒃

• In general, we say that 𝑓(𝑥, 𝑦) is continuous on domain D

if 𝑓 is continuous at every point (𝑎, 𝑏) in domain D.

Source: Link
21
 RMIT Classification: Trusted

Partial Differentiation – Definition

• If 𝑓(𝑥, 𝑦) is a function of two independent variables 𝑥 and 𝑦,

we could let only 𝑥 vary while keeping 𝑦 fixed (i.e. 𝑦 = 𝑏,
where 𝑏 is a constant). In such case we can define a single
variable function 𝑔(𝑥) as:
𝑔 𝑥 = 𝑓(𝑥, 𝑏)

• If 𝑔 has a derivative at any point 𝑥 = 𝑎, then we call it the

partial derivative of 𝑓 with respect to 𝑥 at (𝑎, 𝑏), and write it
as 𝑓𝑥 𝑎, 𝑏 .
Source: Link
22
 RMIT Classification: Trusted

Partial Differentiation – Definition

• Hence we have the limit definition of a partial derivative of

𝑓 with respect to 𝑥 at point (𝑥, 𝑦):

𝒇 𝒙 + 𝒉, 𝒚 − 𝒇 𝒙, 𝒚
𝒇𝒙 𝒙, 𝒚 = 𝐥𝐢𝐦
𝒉→𝟎 𝒉

• Similarly, we have the limit definition of a partial derivative

of 𝑓 with respect to 𝑦 at point (𝑥, 𝑦):

𝒇 𝒙, 𝒚 + 𝒉 − 𝒇 𝒙, 𝒚
𝒇𝒚 𝒙, 𝒚 = 𝐥𝐢𝐦
𝒉→𝟎 𝒉
Source: Link
23
 RMIT Classification: Trusted

Partial Differentiation – Notation

• Like derivatives in 2D, there are different ways we can

express the Partial Derivatives with respect to 𝑥 or 𝑦.

𝜕𝑓 𝜕𝑧 𝜕
𝑓𝑥 𝑥, 𝑦 = 𝑓𝑥 = = = 𝑓 𝑥, 𝑦 = 𝐷𝑥 𝑓
𝜕𝑥 𝜕𝑥 𝜕𝑥

𝜕𝑓 𝜕𝑧 𝜕
𝑓𝑦 𝑥, 𝑦 = 𝑓𝑦 = = = 𝑓 𝑥, 𝑦 = 𝐷𝑦 𝑓
𝜕𝑦 𝜕𝑦 𝜕𝑦

• To find 𝑓𝑥 , regard 𝑦 as a constant and differentiate 𝑓(𝑥, 𝑦) with respect to 𝑥.

• To find 𝑓𝑦 , regard 𝑥 as a constant and differentiate 𝑓 𝑥, 𝑦 with respect 𝑦.

Source: Link
24
 Partial Differentiation – What does it
RMIT Classification: Trusted

mean?
• Recall that equation 𝑓 𝑥, 𝑦 represents a surface 𝑆.

• If 𝑓 𝑎, 𝑏 = c, then the point 𝑃(𝑎, 𝑏, 𝑐) is on the surface 𝑆.

• If 𝑦 = 𝑏, then 𝑓(𝑥, 𝑏) is the curve 𝐶1 where plane 𝑦 = 𝑏

intersects surface 𝑆.

• If 𝑥 = 𝑎, then 𝑓(𝑎, 𝑦) is the curve 𝐶2 where plane 𝑦 = 𝑏

intersects surface 𝑆.

• Point 𝑃(𝑎, 𝑏, 𝑐) is the intersection point of curve 𝐶1 and 𝐶2 .

Source: Link
25
 Partial Differentiation – What does it
RMIT Classification: Trusted

mean?

• Hence the partial derivatives 𝑓𝑥 (𝑎, 𝑏) and 𝑓𝑦 (𝑎, 𝑏) can be

understood as the slopes of the tangent lines at 𝑃 𝑎, 𝑏, 𝑐
along curves 𝐶1 and 𝐶2 of surface 𝑆.
Source: Link
26
 RMIT Classification: Trusted

Partial Differentiation – Example

Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 4 − 𝑥 2 − 2𝑦 2 , find
𝑓𝑥 (1, 1) and 𝑓𝑦 (1, 1) and interpret them as slopes.

(1) Let’s find the partial derivative of 𝑥:

𝜕
𝑓𝑥 = 4 − 𝑥 2 − 2𝑦 2 = −2𝑥
𝜕𝑥
(2) Let’s find the partial derivative of 𝑦:
𝜕
𝑓𝑦 = 4 − 𝑥 2 − 2𝑦 2 = −4𝑦
𝜕𝑦
(3) Hence the slopes are:
𝑓𝑥 1, 1 = −2 1 = −𝟐
𝑓𝑦 1, 1 = −4 1 = −𝟒
(4) Let’s find the value of 𝑓(1, 1) for reference:
𝑓 1, 1 = 4 − 1 2 − 2 1 2 = 1
Hence, we are looking at point 𝑃(1, 1, 1) on this
surface 𝑓(𝑥, 𝑦).

27
 RMIT Classification: Trusted

Partial Differentiation – Example

Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 4 − 𝑥 2 − 2𝑦 2 , find
𝑓𝑥 (1, 1) and 𝑓𝑦 (1, 1) and interpret them as slopes.

28
 RMIT Classification: Trusted

End of Part 1

(To be continued…)
 Partial Differentiation – Differentiation
RMIT Classification: Trusted

Formulas
• In general, Partial Derivatives will follow 7 out of 8 rules we discussed
previously for Differentiation.
1. Derivative of a Constant.
2. Power Rule
3. Constant Multiple Rule
4. Sum Rule
5. Difference Rule
6. Product Rule
7. Quotient Rule

Source: Link
30
 RMIT Classification: Trusted

Partial Differentiation – Chain Rule

• The existence of 2 or more independent variables calls for some

special modifications to the Chain Rule.

• Given function 𝑓 𝑥, 𝑦 where 𝑥 = 𝑔(𝑡) and 𝑦 = ℎ(𝑡)

• The derivative of 𝑧 = 𝑓(𝑥, 𝑦) is:

𝑑𝑧 Δ𝑟 𝑓 𝑥 + Δx, y + Δ𝑦 − 𝑓 𝑥, 𝑦
= lim = lim
𝑑𝑡 Δ𝑡→0 Δ𝑡 Δ𝑡→0 Δ𝑡

• Invoking the small change approximation, we can approximate a small

change Δ𝑧 to the value of 𝑧 as:
Δ𝑧 = 𝑓(𝑥 + Δx, y + Δ𝑦) − 𝑓(𝑥, 𝑦)
Source: Link
31
 RMIT Classification: Trusted

Partial Differentiation – Chain Rule

• We can add a term 𝑓 𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 + Δ𝑦 to the previous

equation:

Δ𝑧 1
= 𝑓 𝑥 + Δx, y + Δ𝑦 + 𝑓 𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦
Δ𝑡 Δ𝑡
1
= [ 𝑓 𝑥 + Δ𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 + Δ𝑦 + 𝑓 𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 ]
Δ𝑡

𝑓 𝑥 + Δ𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 + Δ𝑦 Δ𝑥 𝑓 𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 Δ𝑦
= +
Δ𝑥 Δ𝑡 Δ𝑦 Δ𝑡

• Recall that 𝑥 = 𝑔(𝑡), hence if Δ𝑡 → 0 then Δ𝑥 → 0

• Similarly y = ℎ(𝑡), hence if Δ𝑡 → 0 then Δ𝑦 → 0

Source: Link
32
 RMIT Classification: Trusted

Partial Differentiation – Chain Rule

• Finally we take the limit of the function:

Δ𝑧
lim
Δ𝑡→0 Δ𝑡

𝑓 𝑥 + Δ𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 + Δ𝑦 Δ𝑥
= lim lim
Δ𝑥→0 Δ𝑥 Δ𝑡→0 Δ𝑡

𝑓 𝑥, 𝑦 + Δ𝑦 − 𝑓 𝑥, 𝑦 Δ𝑦
+ lim lim
Δ𝑦→0 Δ𝑦 Δ𝑡→0 Δ𝑡

𝒅𝒛 𝝏𝒛 𝒅𝒙 𝝏𝒛 𝒅𝒚 𝒅𝒙 𝒅𝒚
→ = . + . = 𝒇𝒙 . + 𝒇𝒚 .
𝒅𝒕 𝝏𝒙 𝒅𝒕 𝝏𝒚 𝒅𝒕 𝒅𝒕 𝒅𝒕

Source: Link
33
 RMIT Classification: Trusted

Partial Differentiation – Chain Rule

• In general, the process above can be repeated indefinitely for any

function 𝑢 with 𝑛 variables 𝑥1 , 𝑥2 , … , 𝑥𝑛 and each 𝑥𝑗 is a differentiable
function of the 𝑚 variables 𝑡1 , 𝑡2 , … , 𝑡𝑚 :

𝝏𝒖 𝝏𝒖 𝝏𝒙𝟏 𝝏𝒖 𝝏𝒙𝟐 𝝏𝒖 𝝏𝒙𝒏

= + + ⋯+
𝝏𝒕𝒊 𝝏𝒙𝟏 𝝏𝒕𝒊 𝝏𝒙𝟐 𝝏𝒕𝒊 𝝏𝒙𝒏 𝝏𝒕𝒊

For each 𝑖 = 1, 2, … , 𝑚

Source: Link
34
 Partial Differentiation – Chain Rule
RMIT Classification: Trusted

Example
Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 𝑥 2 𝑦 + 3𝑥𝑦 4 , where
𝑑𝑧
𝑥 = sin 2𝑡 and 𝑦 = cos 𝑡 , find when 𝑡 = 0
𝑑𝑡
(1) Invoking the Chain Rule, we have:
𝑑𝑧 𝑑𝑥 𝑑𝑦
= 𝑓𝑥 + 𝑓𝑦
𝑑𝑡 𝑑𝑡 𝑑𝑡
(2) Let’s find the partial derivative with respect to 𝑥
and 𝑦:
𝑓𝑥 = 2𝑥𝑦 + 3𝑦 4
𝑓𝑦 = 𝑥 2 + 12𝑥𝑦 3
(3) Next let’s find the derivative of 𝑥 and 𝑦 with
respect to 𝑡:
𝑑𝑥
= 2 cos 2𝑡
𝑑𝑡
𝑑𝑦
= − sin 𝑡
𝑑𝑡

35
 Partial Differentiation – Chain Rule
RMIT Classification: Trusted

Example
Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 𝑥 2 𝑦 + 3𝑥𝑦 4 , where
𝑑𝑧
𝑥 = sin 2𝑡 and 𝑦 = cos 𝑡 , find when 𝑡 = 0
𝑑𝑡
(4) Now let’s put the terms back into the equation:
𝑑𝑧
= 2𝑥𝑦 + 3𝑦 4 2 cos 2𝑡 + 𝑥 2 + 12𝑥𝑦 3 − sin 𝑡
𝑑𝑡

(5) Let’s see what happens to 𝑥 and 𝑦 when 𝑡 = 0:

𝑥 0 = sin 0 = 0
𝑦 0 = cos 0 = 1
(6) Hence, the equation from (4) becomes:
𝑑𝑧
= 2 ∗ 0 ∗ 1 + 3 ∗ 14 2 ∗ 1 + 02 + 12 0 13 (−0)
𝑑𝑡
= 3∗2 +0=𝟔

36
 Partial Differentiation – Implicit
RMIT Classification: Trusted

Differentiation
• Given a function 𝐹 𝑥, 𝑦 = 0 where 𝑦 = 𝑓(𝑥), then we can invoke the
Chain Rule as:

𝑑𝐹 𝑥, 𝑦 𝜕𝐹 𝑑𝑥 𝜕𝐹 𝑑𝑦
= . + . =0
𝑑𝑥 𝜕𝑥 𝑑𝑥 𝜕𝑦 𝑑𝑥

𝝏𝑭
𝒅𝒚 − 𝑭𝒙
→ = 𝝏𝒙 =−
𝒅𝒙 𝝏𝑭 𝑭𝒚
𝝏𝒚

Note: the result will be identical to the 2D case we learned in topic 3.

Source: Link
37
 Partial Differentiation – Implicit
RMIT Classification: Trusted

Differentiation
• Given a function 𝐹 𝑥, 𝑦, 𝑧 = 0 where 𝑥 and 𝑦 are independent variable
and 𝑧 is given implicitly as a function of 𝑥 and 𝑦, then we can invoke
the Chain Rule as:

𝑑𝐹 𝑥, 𝑦 𝜕𝐹 𝑑𝑥 𝜕𝐹 𝜕𝑦 𝜕𝐹 𝜕𝑧
= . + . + . =0
𝑑𝑥 𝜕𝑥 𝑑𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑧 𝜕𝑥

𝜕 𝑥 𝜕 𝑦
• Recognize that = 1 and = 0, we have:
𝜕𝑥 𝜕𝑥

𝝏𝑭
𝜕𝐹 𝜕𝐹 𝜕𝑧 𝝏𝒛 𝑭𝒙 −
+ . =0→ = 𝝏𝒙 = −
𝜕𝑥 𝜕𝑧 𝜕𝑥 𝝏𝒙 𝝏𝑭 𝑭𝒛
𝝏𝒛
Source: Link
38
 Partial Differentiation – Implicit
RMIT Classification: Trusted

Differentiation
• The partial derivative of 𝑧 with respect to 𝑦 can be obtained in a similar
manner. Thus we have:

𝝏𝑭
𝝏𝒛 − 𝑭𝒚
𝝏𝒚
= =−
𝝏𝒚 𝝏𝑭 𝑭𝒛
𝝏𝒛

Source: Link
39
 Partial Differentiation – Implicit
RMIT Classification: Trusted

Differentiation Example
Example: Given the multivariable function 𝑥 3 + 𝑦 3 + 𝑧 3 + 6𝑥𝑦𝑧 = 1, find
𝜕𝑧 𝜕𝑧
and .
𝜕𝑥 𝜕𝑦
Solution:
(1) First we invoke the formulas for Implicit differentiation:
𝜕𝑧 𝐹𝑥 𝜕𝑧 𝐹𝑦
= − 𝑎𝑛𝑑 = −
𝜕𝑥 𝐹𝑧 𝜕𝑦 𝐹𝑧
(2) Our function 𝐹 𝑥, 𝑦, 𝑧 is:
𝐹 𝑥, 𝑦, 𝑧 = 𝑥 3 + 𝑦 3 + 𝑧 3 + 6𝑥𝑦𝑧 − 1
(3) Let’s find the values of the partial derivatives:
𝐹𝑥 = 3𝑥 2 + 6𝑦𝑧
𝐹𝑦 = 3𝑦 2 + 6𝑥𝑧
𝐹𝑧 = 3𝑧 2 + 6𝑥𝑦

40
 Partial Differentiation – Implicit
RMIT Classification: Trusted

Differentiation Example
Example: Given the multivariable function 𝑥 3 + 𝑦 3 + 𝑧 3 + 6𝑥𝑦𝑧 = 1, find
𝜕𝑧 𝜕𝑧
and .
𝜕𝑥 𝜕𝑦
Solution:
(4) Hence we have:
𝜕𝑧 𝐹𝑥 3𝑥 2 + 6𝑦𝑧 𝒙𝟐 + 𝟐𝒚𝒛
= − =− 2 =− 𝟐
𝜕𝑥 𝐹𝑧 3𝑧 + 6𝑥𝑦 𝒛 + 𝟐𝒙𝒚
𝜕𝑧 𝐹𝑦 3𝑦 2 + 6𝑥𝑧 𝒚𝟐 + 𝟐𝒙𝒛
= − =− 2 =− 𝟐
𝜕𝑦 𝐹𝑧 3𝑧 + 6𝑥𝑦 𝒛 + 𝟐𝒙𝒚

41
 Partial Differentiation – Higher RMIT Classification: Trusted

Derivatives
• Similar to the 2D case, we can take higher derivatives of Multivariable
Functions. However, since there are two Partial 1st Order Partial Derivatives
(for the case of 2 variables), there are four 2nd Order Derivatives:

𝝏 𝝏𝒇 𝝏𝟐 𝒇 𝝏𝟐 𝒛
𝒇𝒙 𝒙 = 𝒇𝒙𝒙 = = 𝟐= 𝟐
𝝏𝒙 𝝏𝒙 𝝏𝒙 𝝏𝒙

𝝏 𝝏𝒇 𝝏𝟐 𝒇 𝝏𝟐 𝒛
𝒇𝒙 𝒚 = 𝒇𝒙𝒚 = = =
𝝏𝒚 𝝏𝒙 𝝏𝒚𝝏𝒙 𝝏𝒚𝝏𝒙

𝝏 𝝏𝒇 𝝏𝟐 𝒇 𝝏𝟐 𝒛
𝒇𝒚 = 𝒇𝒚𝒙 = = =
𝒙 𝝏𝒙 𝝏𝒚 𝝏𝒙𝝏𝒚 𝝏𝒙𝝏𝒚

𝝏 𝝏𝒇 𝝏𝟐 𝒇 𝝏𝟐 𝒛
𝒇𝒚 = 𝒇𝒚𝒚 = = 𝟐= 𝟐
Source: Link
𝒚 𝝏𝒚 𝝏𝒚 𝝏𝒚 𝝏𝒚
42
 Partial Differentiation – Higher
RMIT Classification: Trusted

Derivatives
• Clairaut’s Theorem: (Symmetry of 2nd Derivative) If the functions 𝑓𝑥𝑦 and 𝑓𝑦𝑥
are both continuous on a domain D, then at a point 𝑎, 𝑏 , within domain D, on
the surface 𝑓 where 𝑓𝑥𝑦 and 𝑓𝑦𝑥 exist:
𝒇𝒙𝒚 𝒂, 𝒃 = 𝒇𝒚𝒙 (𝒂, 𝒃)

Source: Link
43
 Partial Differentiation – Multivariable
RMIT Classification: Trusted

Calculus Example
Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 𝑥 3 + 𝑥 2 𝑦 3 − 2𝑦 2 , find
the 2nd partial derivatives of 𝑓(𝑥, 𝑦). Show that 𝑓𝑥𝑦𝑦 = 𝑓𝑦𝑦𝑥 = 𝑓𝑦𝑥𝑦 .
(1) First we find the partial derivatives with respect
to 𝑥 and 𝑦:
𝑓𝑥 = 3𝑥 2 + 2𝑥𝑦 3
𝑓𝑦 = 3𝑥 2 𝑦 2 − 4𝑦
(2) Next let’s find our 2nd Order partial derivatives:
𝒇𝒙𝒙 = 𝟔𝒙 + 𝟐𝒚𝟑
𝒇𝒚𝒚 = 𝟔𝒙𝟐 𝒚 − 𝟒
𝒇𝒙𝒚 = 𝟔𝒙𝒚𝟐
𝒇𝒚𝒙 = 𝟔𝒙𝒚𝟐

44
 Partial Differentiation – Multivariable
RMIT Classification: Trusted

Calculus Example
Example: Given the multivariable function 𝑓 𝑥, 𝑦 = 𝑥 3 + 𝑥 2 𝑦 3 − 2𝑦 2 , find
the 2nd partial derivatives of 𝑓(𝑥, 𝑦). Show that 𝑓𝑥𝑦𝑦 = 𝑓𝑦𝑦𝑥 = 𝑓𝑦𝑥𝑦 .
(3) Next let us find 𝑓𝑥𝑦𝑦 , 𝑓𝑦𝑦𝑥 , 𝑓𝑦𝑥𝑦 :
𝜕 𝑓𝑥𝑦 𝜕 6𝑥𝑦 2
𝑓𝑥𝑦𝑦 = = = 𝟏𝟐𝒙𝒚
𝜕𝑦 𝜕𝑦
𝜕 𝑓𝑦𝑦 𝜕 6𝑥 2 𝑦 − 4
𝑓𝑦𝑦𝑥 = = = 𝟏𝟐𝒙𝒚
𝜕𝑥 𝜕𝑦
𝜕 𝑓𝑦𝑥 𝜕 6𝑥𝑦 2
𝑓𝑦𝑥𝑦 = = = 𝟏𝟐𝒙𝒚
𝜕𝑦 𝜕𝑦

Hence: 𝒇𝒙𝒚𝒚 = 𝒇𝒚𝒚𝒙 = 𝒇𝒚𝒙𝒚 .

45
 Partial Differentiation – RMIT Classification: Trusted

Exercises
1. Find the domain of 𝑧 = 𝑓 𝑥, 𝑦 = ln(4 − 𝑥 2 − 𝑦 2 ) with 𝑝, 𝑥, 𝑦 ∈ 𝑅

2. Find the function 𝑓𝑥 , 𝑓𝑦 , 𝑓𝑥𝑥 , 𝑓𝑦𝑦 , 𝑓𝑥𝑦 , 𝑓 𝑦𝑥 given the function:

𝑓 𝑥, 𝑦 = 3𝑥 3 + 2𝑥𝑦 + ln 𝑦

3. Let 𝑧 = 𝑥 10 + 6 and 𝑥 = 3𝑡 + 𝑣.

𝜕𝑧
Know that 𝑣 = 𝑣(𝑡), write the function .
𝜕𝑡

𝜕𝑤
4. Find , given that:
𝜕𝑡

𝑤 = 𝑥𝑦 2
𝑥 = 𝑒 3𝑡 cos 𝑐 ; 𝑦 = 𝑒 4𝑡 sin 𝑐 ; 𝑐 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡

46
 RMIT Classification: Trusted

Applications of Partial
Differentiations
The useful things we can do
 Small Change Approximation –
RMIT Classification: Trusted

Introduction.
• Recall how the derivative 𝑓′(𝑥) can be estimated by the
formula:

′
𝑓 𝑥 + Δ𝑥 − 𝑓 𝑥
𝑓 𝑥 ≈
Δ𝑥

• For Multivariable function, we can perform a similar

approximation by invoking the tangent plane.

Source: Link
48
 Small Change Approximation –
RMIT Classification: Trusted

Introduction.
• Recall the equation of a tangent plane:
𝒛 − 𝒛𝟎 = 𝒇𝒙 𝒙𝟎 , 𝒚𝟎 𝒙 − 𝒙𝟎 + 𝒇𝒚 (𝒙𝟎 , 𝒚𝟎 )(𝒚 − 𝒚𝟎 )

• Hence, for any point 𝑎, 𝑏 on the surface 𝑓(𝑥, 𝑦), we can

approximate the change to 𝑧 corresponding to small
changes to 𝑥 and 𝑦 as:
𝑧 = 𝑓 𝑎, 𝑏 + 𝑓𝑥 𝑎, 𝑏 𝑥 − 𝑎 + 𝑓𝑦 𝑎, 𝑏 (𝑦 − 𝑏)

Source: Link
49
 Small Change Approximation – RMIT Classification: Trusted

Introduction.
• Thus we can conclude for a general case, the change in 𝑧
can be approximated as Δ𝑧 via this function:
Δ𝑧 ≈ 𝑓𝑥 𝑥, 𝑦 Δ𝑥 + 𝑓𝑦 𝑥, 𝑦 Δ𝑦

Note: This is also called the Linear Approximation or Tangent Plane Approximation of 𝑧.

Source: Link
50
 RMIT Classification: Trusted

The Gradient Vector– Example

Example: Use Linear Approximation to estimate the amount of tin in a
closed tin can with diameter 8 cm and height 12 cm if the tin is 0.04 cm
thick.
(1) First we recognize that the tin can is cylindrical.
Hence its volume is given by the equation:
𝑉 = 𝜋𝑟 2 ℎ

(2) We are trying to find Δ𝑉 which is the amount of

tin needed for 1 can.

(3) We should recognize that a change in volume

(due to thickness) means a change in the can’s
radius and height. Thus 𝑟 and ℎ are our variables.
Hence, we have the equation:
𝑑𝑉 = 𝜋 2𝑟ℎ 𝑑𝑟 + 𝜋 𝑟 2 𝑑ℎ

51
 RMIT Classification: Trusted

The Gradient Vector– Example

Example: Use Linear Approximation to estimate the amount of tin in a
closed tin can with diameter 8 cm and height 12 cm if the tin is 0.04 cm
thick.
(4) The change in radius 𝑑𝑟 = 0.04 and the change
in height is 𝑑ℎ = 0.04 ∗ 2 = 0.08 (due to the top and
bottom lid). Thus we can calculate the change in
volume from the inner side to the outer side of the
can:
𝑑𝑉 = (2 ∗ 𝜋 ∗ 4 ∗ 12 ∗ 0.04) + 𝜋 42 (0.08)
= 16.08 𝑐𝑚3

52
 RMIT Classification: Trusted

Optimization – Introduction.

• Similar to the Single Variable function, it is possible to

perform optimization with Multi Variable functions.

• Due to the existence of more than one variable, we need to

redefine our definition of Minimum and Maximum values

Source: Link
53
 RMIT Classification: Trusted

Optimization – Critical Points

Example of Critical Points on 3D surfaces

54
 RMIT Classification: Trusted

Optimization – Critical Points

• A point (𝑎, 𝑏) is called a critical point (or stationary point) of

𝑓 if 𝑓𝑥 𝑎, 𝑏 = 0 AND 𝑓𝑦 𝑎, 𝑏 = 0 OR if one of these partial
derivatives does not exist.

• At a Critical point, a function could have a local maximum,

or a local minimum, or neither.

Source: Link
55
 Optimization – Critical Points
RMIT Classification: Trusted

Example
Example: Given the MVF: 𝑓 𝑥, 𝑦 = 𝑥 2 + 𝑦 2 − 2𝑥 − 6𝑦 + 14. Find the
critical point(s) on the surface of this MVF. Is it a maximum or a minimum?

(1) First we need to calculate the Partial derivatives

with respect to 𝑥 and 𝑦:
𝑓𝑥 = 2𝑥 − 2
𝑓𝑦 = 2𝑦 − 6

(2) Next we set the partial derivatives to 0 and solve

for 𝑥 and 𝑦:
𝑓𝑥 = 0 = 2𝑥 − 2 → 𝑥 = 1
𝑓𝑦 = 0 = 2𝑦 − 6 → 𝑦 = 3

(3) Hence we can conclude that there is a critical

point at point 1, 3 . The corresponding value for 𝑧 is:
𝑧 = 𝑓 1, 3 = 12 + 32 − 2 1 − 6 3 + 14 = 4

56
 Optimization – Critical Points
RMIT Classification: Trusted

Example
Example: Given the MVF: 𝑓 𝑥, 𝑦 = 𝑥 2 + 𝑦 2 − 2𝑥 − 6𝑦 + 14. Find the
critical point(s) on the surface of this MVF. Is it a maximum or a minimum?

(4) Plotting the surface on 3D space, we can

conclude that point (1, 3, 4) is an absolute minimum
for the function 𝑓 𝑥, 𝑦 .

Can we use the 2nd Derivative to check this fact in

case we don’t have to tool to graph the surface?

57
 RMIT Classification: Trusted

Optimization – 2nd Derivative test

• Similar to the Single Variable Function, it is possible to test the
nature of a Critical point by using the 2nd Derivative.

• The formula for the 2nd Derivative test on MVFs is derived from
the determinant of the Hessian Matrix (to be discussed in Math
2240). For a 2-variables function, the Hessian matrix is:

𝑓𝑥𝑥 𝑓𝑥𝑦
→ 𝐷𝑒𝑡𝑒𝑟𝑚𝑖𝑛𝑎𝑛𝑡 = 𝒇𝒙𝒙 ∗ 𝒇𝒚𝒚 − 𝒇𝒙𝒚 𝒇𝒚𝒙
𝑓𝑦𝑥 𝑓𝑦𝑦

Source: Link
58
 RMIT Classification: Trusted

Optimization – 2nd Derivative test

• To determine the nature of a critical point on a 3D surface, we follow
this procedure:
1. Compute the values of 𝑓𝑥𝑥 , 𝑓𝑦𝑦 , 𝑓𝑥𝑦 at the Critical point.

2. Invoke the Determinant of a Hessian Matrix formula:

2
𝐷 = 𝐷 𝑎, 𝑏 = 𝑓𝑥𝑥 𝑎, 𝑏 𝑓𝑦𝑦 𝑎, 𝑏 − 𝑓𝑥𝑦 (𝑎, 𝑏)
a. If 𝐷 > 0 and 𝑓𝑥𝑥 𝑎, 𝑏 > 0, then 𝑓 𝑎, 𝑏 is a local minimum.
b. If 𝐷 > 0 and 𝑓𝑥𝑥 𝑎, 𝑏 < 0, then 𝑓 𝑎, 𝑏 is a local maximum.
c. If 𝐷 < 0 then 𝑓 𝑎, 𝑏 is neither a local maximum or minimum (saddle point).
d. If 𝐷 = 0 then the test is null as 𝑓(𝑎, 𝑏) can be any of the above.

Source: Link
59
 RMIT Classification: Trusted

Optimization – Saddle Point

60
 Optimization – 2nd Derivative test
RMIT Classification: Trusted

Example
Example: Given the MVF: 𝑓 𝑥, 𝑦 = 𝑥 4 + 𝑦 4 − 4𝑥𝑦 + 1. Find the critical
point(s) on the surface of this MVF.

(1) First we need to calculate the Partial derivatives

with respect to 𝑥 and 𝑦:
𝑓𝑥 = 4𝑥 3 − 4𝑦
𝑓𝑦 = 4𝑦 3 − 4𝑥

(2) Next we set the partial derivatives to 0 and solve

for 𝑥 and 𝑦:
𝑓𝑥 = 0 = 4𝑥 3 − 4𝑦 = 𝑥 3 − 𝑦
𝑓𝑦 = 0 = 4𝑦 3 − 4𝑥 = 𝑦 3 − 𝑥
Hence we need to solve the system of equation:
𝑥3 − 𝑦 = 0
ቊ 3
𝑦 −𝑥 =0
→ 𝑇ℎ𝑒 3 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠 𝑎𝑟𝑒: 0, 0 , 1, 1 , (−1, −1)

61
 Optimization – 2nd Derivative test
RMIT Classification: Trusted

Example
Example: Given the MVF: 𝑓 𝑥, 𝑦 = 𝑥 4 + 𝑦 4 − 4𝑥𝑦 + 1. Find the critical
point(s) on the surface of this MVF.

(3) Next we calculate the 2nd Derivatives:

𝑓𝑥𝑥 = 12 𝑥 2
𝑓𝑦𝑦 = 12𝑦 2
𝑓𝑥𝑦 = −4
(4) Thus we have the Determinant function:
2
𝐷 𝑥, 𝑦 = 𝑓𝑥𝑥 𝑓𝑦𝑦 − 𝑓𝑥𝑦 = 144𝑥 2 𝑦 2 − 16
(5) Next we check the D and 𝑓𝑥𝑥 value for each critical
point:
𝐷 −1, −1 = 144 − 16 = 128 > 0
𝑓𝑥𝑥 −1, −1 = 12 > 0
Hence, point 𝐴(−1, −1) is a local minimum.

62
 Optimization – 2nd Derivative test
RMIT Classification: Trusted

Example
Example: Given the MVF: 𝑓 𝑥, 𝑦 = 𝑥 4 + 𝑦 4 − 4𝑥𝑦 + 1. Find the critical
point(s) on the surface of this MVF.

(5) Next we check the D and 𝑓𝑥𝑥 value for each critical
point (cont.):
𝐷 0, 0 = 0 − 16 = −16 < 0
Hence, point 𝐵(0, 0) is a saddle point.
𝐷 1, 1 = 144 − 16 = 128 > 0
𝑓𝑥𝑥 1, 1 = 12 > 0
Hence, point 𝐶(1, 1) is a local minimum.

63
 RMIT Classification: Trusted

Optimization – Solving Procedures

The procedures to solve Optimization problems in 3D is almost identical to

the 2D case:
1. Read the problem and identify the unknowns and given quantities/condition.

2. Draw a diagram of the problem.

3. Find an equation to relate the unknown (that has to be maximized or

minimized) with the given information.

4. Find the constraint equation and solve for the 3rd variable. Substitute the 3rd
variable back to the equation in step 3.

5. Use the method discussed to find the absolute maximum or absolute

minimum points of the function from step 3.

64
 RMIT Classification: Trusted

Optimization – Example
Example: A rectangular box without a lid is to be made from 12𝑚2 of
cardboard. Find the maximum volume of such a box.

(1) First let us call the length, width, and height of the
box as 𝑥, 𝑦, 𝑧.

(2) Next we recognize that the Volume of the box is

given by:
𝑉 =𝑥∗𝑦∗𝑧

(3) We can also recognize that the area of available

carboard is limited at 12 𝑚2 , hence we have the
constraint equation:
2𝑥𝑧 + 2𝑦𝑧 + 𝑥𝑦 = 12
12 − 𝑥𝑦
→𝑧=
[ 2 𝑥+𝑦

65
 RMIT Classification: Trusted

Optimization – Example
Example: A rectangular box without a lid is to be made from 12𝑚2 of
cardboard. Find the maximum volume of such a box.

(4) We can substitute the 𝑧 equation from (3) to the

Volume equation from (2):
12 − 𝑥𝑦 12𝑥𝑦 − 𝑥 2 𝑦 2
𝑉 = 𝑥𝑦 ∗ =
2∗ 𝑥+𝑦 2 𝑥+𝑦
(5) Now we take the partial derivatives to find 𝑉𝑥 and
𝑉𝑦 :
y 2 12 − 2xy − x 2
𝑉𝑥 =
(2 x + y 2
x 2 12 − 2xy − y 2
𝑉𝑦 =
(2 x + y 2

66
 RMIT Classification: Trusted

Optimization – Example
Example: A rectangular box without a lid is to be made from 12𝑚2 of
cardboard. Find the maximum volume of such a box.

(6) Now we take the partial derivatives to 0 to find the

critical points:
y 2 12 − 2xy − x 2 2 12 − 2𝑥𝑦 − 𝑥 2 = 0
𝑉𝑥 = = 0 → 𝑦
(2 x + y 2
x 2 12 − 2xy − y 2
𝑉𝑦 = 2 = 0 → x 2 12 − 2xy − y 2 = 0
(2 x + y
We recognize that 𝑥 = 0 and 𝑦 = 0 is a solution, thus we
can simplify our system of equation to be:
12 − 2𝑥𝑦 − 𝑥 2 = 0
൝
12 − 2xy − y 2 = 0
Solve the equations above for 𝑥 will yield 2 roots: 𝑥 = 2
and 𝑥 = −2, as 𝑥 cannot be negative, we accept 𝑥 = 2.

67
 RMIT Classification: Trusted

Optimization – Example
Example: A rectangular box without a lid is to be made from 12𝑚2 of
cardboard. Find the maximum volume of such a box.

(7) Thus we can conclude that we have 2 critical points

at (0, 0) and (2, 2). The case of (0, 0) is invalid as the
sides cannot be 0. Hence point (2, 2) must be our
optimum solution.
We can compute the corresponding z value as:
12 − 𝑥𝑦 12 − 2 2 8
𝑧= = = =1
2 𝑥+𝑦 2 2+2 8

(8) Thus the box should have the optimum size of 2 x 2

x 1 meter.

68
 RMIT Classification: Trusted

Questions?

The only stupid question is the one that was not

asked.
 RMIT Classification: Trusted

Copyright © 2019 RMIT University Vietnam