Multivariable Calculus

(Week 5)
Topics: Review of Continuity
Partial Derivatives, Chain Rule, Differentials
Gradients
(Reference Chapters: Chapter 12.3 – 12.8 of Adams and Essex; Chapter 13.3 –
13.6 of Larson and Edwards)

Key References of this file:

Larson, R. and Edwards, B., Multivariable Calculus (Metric Version – 11th Edition),
CENGAGE Learning, 2016.
Adams, R.A. and Essex, C., Calculus – A Complete Course, 9th Edition, Pearson,
2018.

Prepared and Revised by Dr. Hugo MAK

Q & A (from Week 4)
HW 2 Question 5
We define the curvature of a path by 𝐫′′(𝑠) , where 𝐫(𝑠) is the arc-length parametrization of the path.
𝑡
Given a path 𝐫(𝑡), we let 𝐫(𝑠) be its arc-length parametrization so that 𝑠 = ‫׬‬0 𝐫 ′ 𝜏 𝑑𝜏.

Students’ Questions: (1) Why are 𝐫 ′ 𝑠 and 𝐫′′(𝑠) orthogonal vectors?

(2) Why is 𝐫′(𝑠) = 1?
Responses: For Question (2), you may refer to the definition of “arc-length parametrization”
If r(s) is a parametric curve such that 𝐫′(𝑠) = 1 for any s, we say the curve is parametrized by arc-length.
As for Question (1), here you are the claim:
Statement: Let r(t) be a twice continuously differentiable vector-valued function (which describes the
curve), then the speed 𝐫′(𝑡) = C (where C is a constant) if and only if the velocity 𝐫′(𝑡) and the
acceleration vector 𝐫′′(𝑡) are perpendicular for all t. (In the question, our 𝐫′(𝑠) = 1.)
𝑑
Proof: Consider 𝐫′ 𝑡 ∙ 𝐫′ 𝑡 = 𝐫 ′′ 𝑡 ∙ 𝐫 ′ 𝑡 + 𝐫 ′ 𝑡 ∙ 𝐫 ′′ 𝑡 = 2𝐫 ′ 𝑡 ∙ 𝐫 ′′ 𝑡
𝑑𝑡
𝑑(𝐶 2 ) 𝑑
If 𝐫′(𝑡) = C for all t, then 0 = = 𝐫′ 𝑡 ∙ 𝐫′ 𝑡 = 2𝐫 ′ 𝑡 ∙ 𝐫 ′′ 𝑡 implies the two vectors are
𝑑𝑡 𝑑𝑡
perpendicular to each other.
𝑑
On the other hand, if 𝐫 ′ 𝑡 ∙ 𝐫 ′′ 𝑡 = 0, then 0 = 2𝐫 ′ 𝑡 ∙ 𝐫 ′′ 𝑡 = 𝐫′ 𝑡 ∙ 𝐫′ 𝑡 , thus the speed 𝐫′(𝑡) =
𝑑𝑡
′ ′
𝐫 𝑡 ∙ 𝐫 𝑡 is a constant.
Q & A (from Week 4)
Remarks or Q&A from students:
• 𝜀 − 𝛿 approach (Slide 33-35) is NOT required in exam, just for your reference.
• Rectification / Rectifiable (NOT a key word in our content, will NOT be included in exam)
1. Rectification of curve: Find the length of an irregular arc segment of that curve
2. Rectifiable curve: There exists a smallest number L that can serve as an upper bound on the
length of any polygonal approximation of that curve. L: arc-length
3. For further details, please refer to
https://en.wikipedia.org/wiki/Arc_length#Definition_for_a_smooth_curve
4. Non-rectifiable curve: Length of that curve can be made arbitrarily large.
Example: Koch curve (also known as Koch snowflake)
• Recall an example of Continuity
sin 𝑥
𝑓 𝑥 = itself is NOT continuous, however
𝑥
sin 𝑥
𝑔 𝑥 =ቐ 𝑥 , 𝑥≠0
limit at 0 is 1, but f(0) not
well defined. that's why we
redefine to make it
continuous
1, 𝑥=0
is continuous.
1 more example on Continuity
Example: Show that
3𝑥𝑦
2 2
when (𝑥, 𝑦) ≠ (0,0)
𝑓 𝑥, 𝑦 = ൞𝑥 + 𝑦 Ref.: From Thomas Calculus
0 when 𝑥, 𝑦 = (0,0)
is continuous at every point in 𝐑2 except
at the origin.

• 𝑓 is well-defined at all points in 𝐑2 (including the origin).

• The function f is continuous at any point (𝑥, 𝑦) ≠ (0,0), because both 3𝑥𝑦 and
𝑥 2 + 𝑦 2 are polynomials (rational functions of 𝑥 and 𝑦), therefore the limit values
can be obtained by substituting respective (𝑥, 𝑦) into the functional expression.

• At origin, the value of f is well-defined, but f has no limit as (𝑥, 𝑦) → (0,0) .

Reason: Different paths directed towards the origin will lead to different limits.
1 more example on Continuity
Example: Show that
3𝑥𝑦
when (𝑥, 𝑦) ≠ (0,0)
𝑓 𝑥, 𝑦 = ቐ 𝑥 2+𝑦2 is continuous at every point in 𝐑2 except at the origin.
0 when 𝑥, 𝑦 = (0,0)

Note that the function 𝑓 has a constant value on the “punctured” line 𝑦 = 𝑚𝑥 (straight line
with slope 𝑚) for all 𝑚, where 𝑥 ≠ 0.
3𝑥𝑦 3𝑚𝑥 2 3𝑚
𝑓 𝑥, 𝑦 ቚ = 2 ቤ = =
𝑦=𝑚𝑥 𝑥 + 𝑦 2 𝑦=𝑚𝑥 𝑥 2 + 𝑚𝑥 2 1 + 𝑚2
Thus, 𝑓 will have this number as its limit as (𝑥, 𝑦) → (0, 0) along the line.
3𝑚
lim 𝑓 𝑥, 𝑦 =
(𝑥,𝑦)→(0,0) 1 + 𝑚2
𝑎𝑙𝑜𝑛𝑔 𝑦=𝑚𝑥
Such expression (i.e. the limit) changes when 𝑚 changes.
The limit fails to exists, thus conditions (2) and (3) of “continuity” do not hold, therefore
the function is NOT continuous.
1 more example on Continuity
Example: Show that
3𝑥𝑦
when (𝑥, 𝑦) ≠ (0,0)
𝑓 𝑥, 𝑦 = ቐ 𝑥 2+𝑦2 is continuous at every point in 𝐑2 except at the origin.
0 when 𝑥, 𝑦 = (0,0)

Alternative approach (by Polar coordinates)

3𝑥𝑦 3(𝑟 cos 𝜃)(𝑟 sin 𝜃) 3

lim 𝑓 𝑥, 𝑦 = lim = lim+ = lim+ 3 cos 𝜃 sin 𝜃 = lim+ sin 2𝜃
(𝑥,𝑦)→(0,0) (𝑥,𝑦)→(0,0) 𝑥 2 + 𝑦 2 𝑟→0 𝑟2 𝑟→0 𝑟→0 2

(independent of r)
Then, we take two different directions towards the origin (i.e. two values of 𝜃) as follows:
𝜋 3
Path 1 (𝜃 = (straight line with slope 1)): The limit will be .
4 2
Path 2 (𝜃 = 0 (the 𝑥-axis)): The limit will be 0.

Therefore, the above limit does NOT exist, thus violating conditions (2) and (3).
Partial Derivatives
𝑑𝑦 𝑑𝑓
In 1-variable calculus, for 𝑦 = 𝑓(𝑥), we have or (rate of change of f with
𝑑𝑥 𝑑𝑥
respect to x)
In multivariable calculus, one can determine the rate of change of a function f
with respect to one of its several independent variables.
This process is called partial differentiation, and the result is referred to as the
partial derivative of f with respect to the chosen independent variable.

Idea:
When we differentiate a function of several variables with respect to one of the independent variables, we
have to keep all other independent variables constant, then apply product rule, quotient rule, chain rule
etc.
𝜕𝑓 𝜕𝑓 𝜕𝑓
e.g. 𝑧 = 𝑓 𝑥, 𝑦, 𝑡 , we can have 𝑧𝑥 = 𝑓𝑥 = ; 𝑧𝑦 = 𝑓𝑦 = and 𝑧𝑡 = 𝑓𝑡 = respectively.
𝜕𝑥 𝜕𝑦 𝜕𝑡
𝜕𝑓
When we compute , we keep both 𝑦 and 𝑡 constant.
𝜕𝑥
 Partial Derivatives
Motivation: Extreme humidity makes us think that the temperature is higher than actual
temperature, whereas when the air is very dry, we perceive the temperature to be lower than the
indicator on thermometer.
Daily Life Example:
The National Weather Service has devised the heat index (also called the temperature-humidity
index, or humidex) to describe the combined effects of temperature and humidity.

The heat index I is the perceived air temperature when the actual temperature is T and the
relative humidity is H. Thus, I is a function of T and H and we can write I = f(T, H).

Figure from
Stewart Calculus
(National Weather Service)
Partial Derivatives (Example due to Stewart Calculus)
• We now concentrate on the highlighted column in the table, where 𝐻 = 70%.
• Thus, we are now considering the heat index as a function of the single variable T for a fixed value
of H. Therefore, g(T) = f(T, 70). g(T) describes how the heat index I increases as the actual
temperature T increases when the relative humidity is 70%.
• The derivative of g when T = 96F is the rate of change of I with respect to T when T = 96F.
′
𝑔 96 + ℎ − 𝑔(96) 𝑓 96 + ℎ, 70 − 𝑓(96,70)
𝑔 96 = lim = lim
ℎ→0 ℎ ℎ→0 ℎ

We try to approximate 𝑔′(96) by taking ℎ = 2

and ℎ = −2 respectively.
𝑔 98 − 𝑔 96 133 − 125
𝑔′ 96 ≈ = =4
2 2
′
𝑔 94 − 𝑔 96 118 − 125
𝑔 96 ≈ = = 3.5
−2 −2

Averaging these values, we say that the

derivative g(96) is approximately 3.75.

When the actual temperature is 96F and the relative humidity is 70%, the apparent temperature (heat index) rises by
about 3.75F for every degree that the actual temperature rises.
General Definition of Partial Derivatives
• In general, if f is a function of two variables x and y, suppose we only vary x while keeping y fixed,
say y = b, where b is a constant.
• Then, we are actually considering a function of x, namely, g(x) = f(x, b). If g has a derivative at the
point a, then we denote it as the partial derivative of f with respect to x at (a, b) - fx (a, b). Thus,
we have
𝑓𝑥 𝑎, 𝑏 = 𝑔′(𝑎), with 𝑔 𝑥 = 𝑓(𝑥, 𝑏)

𝑔 𝑎+ℎ −𝑔(𝑎)
• By definition, we have 𝑔′ 𝑎 = lim , hence
ℎ→0 ℎ

𝑓 𝑎 + ℎ, 𝑏 − 𝑓(𝑎, 𝑏)
𝑓𝑥 𝑎, 𝑏 = lim
ℎ→0 ℎ

• Similarly, the partial derivative of f with respect to y at (a, b), i.e. fy(a, b), is obtained by keeping x
fixed (at x = a), then we find the ordinary derivative of the function G(y) = f(a, y) at 𝑏.

𝑓 𝑎, 𝑏 + 𝑘 − 𝑓(𝑎, 𝑏)
𝑓𝑦 𝑎, 𝑏 = lim
𝑘→0 𝑘
General Definition of Partial Derivatives
Definition of Partial Derivatives of a Function of Two Variables
If 𝑧 = 𝑓(𝑥, 𝑦), then the first partial derivatives of 𝑓 with respect to 𝑥 and 𝑦 are the functions 𝑓𝑥 and 𝑓𝑦
defined as follows, provided that the limits exist:
𝜕𝑓(𝑥, 𝑦) 𝑓 𝑥 + ℎ, 𝑦 − 𝑓(𝑥, 𝑦)
: = 𝑓𝑥 𝑥, 𝑦 = lim
𝜕𝑥 ℎ→0 ℎ

𝜕𝑓(𝑥, 𝑦) 𝑓 𝑥, 𝑦 + 𝑘 − 𝑓(𝑥, 𝑦)
: = 𝑓𝑦 𝑥, 𝑦 = lim
𝜕𝑦 𝑘→0 𝑘

Example: If 𝑓 𝑥, 𝑦 = 𝑥 3 + 𝑥 2 𝑦 3 − 2𝑦 2 , find 𝑓𝑥 (3, 1) and 𝑓𝑦 (1, 7).

Solution: Holding y constant and differentiating with respect to 𝑥, we have

𝑓𝑥 𝑥, 𝑦 = 3𝑥 2 + 2𝑥𝑦 3
Hence, 𝑓𝑥 3, 1 = 3 3 2 +2 3 1 3 = 33.
Similarly, holding 𝑥 constant and differentiating with respect to 𝑦, we have 𝑓𝑦 𝑥, 𝑦 = 3𝑥 2 𝑦 2 − 4𝑦, thus
𝑓𝑦 1, 7 = 3 1 2 7 2 − 4 7 = 119
General Definition of Partial Derivatives
Definition of Partial Derivatives of a Function of n Variables
If 𝑓: 𝐑𝑛 → 𝐑, i.e. 𝑧 = 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑛 ), then

𝜕𝑓 𝑓 𝑥1 , 𝑥2 , … , 𝑥𝑖 + ∆𝑥𝑖 , 𝑥𝑖+1 , … , 𝑥𝑛 − 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑖 , … , 𝑥𝑛 )

= lim
𝜕𝑥𝑖 ∆𝑥𝑖→0 ∆𝑥𝑖

Example: 𝑧 = 𝑓 𝑥, 𝑦, 𝑡, 𝑤, 𝑎, 𝑏, 𝑐, 𝑑 = 3𝑥 2 𝑦𝑎𝑏𝑐 + 2𝑥𝑤𝑡𝑦 − ln 𝑎𝑐 + 𝑒 𝑎𝑥𝑦

𝜕𝑓 𝜕𝑓
= 6𝑥𝑦𝑎𝑏𝑐 + 2𝑤𝑡𝑦 + 𝑎𝑦𝑒 𝑎𝑥𝑦 ; = 3𝑥 2 𝑎𝑏𝑐 + 2𝑥𝑤𝑡 + 𝑎𝑥𝑒 𝑎𝑥𝑦
𝜕𝑥 𝜕𝑦

𝜕𝑓 𝜕𝑓
= 2𝑥𝑤𝑦 ; = 2𝑥𝑡𝑦
𝜕𝑡 𝜕𝑤

𝜕𝑓 1 𝜕𝑓
= 3𝑥 2 𝑦𝑏𝑐 − + 𝑥𝑦𝑒 𝑎𝑥𝑦 ; = 3𝑥 2 𝑦𝑎𝑐
𝜕𝑎 𝑎 𝜕𝑏

𝜕𝑓 1 𝜕𝑓
= 3𝑥 2 𝑦𝑎𝑏 − ; =0
𝜕𝑐 𝑐 𝜕𝑑
Geometric Interpretation of Partial Derivatives (2 variables)
Given a function of two variables z = f(x, y), we can interpret the partial derivatives of 𝑓
geometrically.
If y = y0, then z = f(x, y0) represents the curve formed by intersecting the
surface z = f(x, y) with the plane y = y0.
𝑓 𝑥0 + ℎ, 𝑦0 − 𝑓(𝑥0 , 𝑦0 )
𝑓𝑥 𝑥0 , 𝑦0 = lim
ℎ→0 ℎ
represents the slope of this curve at the point (x0, y0, f(x0, y0)).

Remarks: Both the curve and the tangent line lie in the plane y = y0.

𝑓 𝑥0 ,𝑦0 +𝑘 −𝑓(𝑥0 ,𝑦0 )

Similarly, 𝑓𝑦 𝑥0 , 𝑦0 = lim
𝑘→0 𝑘
represents the slope of the curve given by the intersection of
z = f(x, y) and the plane x = x0 at (x0, y0, f(x0, y0)).

Figures due to Larson and Edwards

Geometric Interpretation of Partial Derivatives (2 variables)
The partial derivatives of f at (a, b) are
the slopes of the tangents to C1 and C2.

Figure due to Stewart

• The curve C1 is the graph of the function g(x) = f(x, b), therefore the slope of its tangent T1 at P is
g(a) = fx (a, b).

• The curve C2 is the graph of the function G(y) = f(a, y), therefore the slope of its tangent T2 at P is
G(b) = fy (a, b).

• The partial derivatives fx(a, b) and fy(a, b) can be interpreted geometrically as the slopes of the
tangent lines at P(a, b, c) to the traces C1 and C2 of surface S in the planes y = b and x = a.
Geometric Interpretation of Partial Derivatives (2 variables)
Given that f(x, y) = 4 – x2 – 2y2, find fx(1, 1) and fy(1, 1) and
interpret these numbers as slopes.

Solution: Figures due to Stewart

𝑓𝑥 𝑥, 𝑦 = −2𝑥 and 𝑓𝑦 𝑥, 𝑦 = −4𝑦, hence 𝑓𝑥 1, 1 = −2 and

𝑓𝑦 1,1 = −4

• The graph of f is the paraboloid z = 4 – x2 – 2y2 and the vertical

plane y = 1 intersects with the paraboloid in the parabola
z = 2 – x2, with y = 1.
• The slope of the tangent line to this parabola at (1, 1, 1) is
fx(1, 1) = –2. ***Slope not considered in
xy surface

• Similarly, the curve in which the plane x = 1 intersects the

paraboloid is the parabola z = 3 – 2y2 with x = 1.
• The slope of the tangent line to this parabola at (1, 1, 1) is
fy(1, 1) = –4.
More Examples
Example 1: Given that 𝑦𝑧 − ln 𝑧 = 𝑥 2 + 𝑦 defines z as a function of two
𝜕𝑧
independent variables 𝑥 and 𝑦, and all partial derivatives exist. Find 𝑧𝑥 ≔ 𝜕𝑥.
Solution: Differentiate both sides with respect to 𝑥 (but keeping 𝑦 constant), and
notice that 𝑧 is a differentiable function of 𝑥, then we have:
1
𝑦𝑧𝑥 − 𝑧𝑥 = 2𝑥 + 0
𝑧
1
𝑦− 𝑧𝑥 = 2𝑥
𝑧
2𝑥𝑧
𝑧𝑥 =
𝑦𝑧 − 1

Take Home Exercise: How about 𝑧𝑦 ?

(Hint: 𝑥 is a constant this time, and for 𝑦𝑧, you will need some rules.)
More Examples
Example 2: If resistors of 𝑅𝑎 , 𝑅𝑏 and 𝑅𝑐 ohms are connected in parallel to make an 𝑅-ohm
𝜕𝑅
resistor. Find the value of 𝜕𝑅 when 𝑅𝑎 = 30, 𝑅𝑏 = 60 and 𝑅𝑐 = 90 ohms.
𝑏

1 1 1 1 𝜕𝑅
Since = + + , to find , we treat 𝑅𝑎 , 𝑅𝑐 as
𝑅 𝑅𝑎 𝑅𝑏 𝑅𝑐 𝜕𝑅𝑏
constants, and we use implicit differentiation.
Differentiate both sides with respect to 𝑅𝑏 :
1 𝜕𝑅 1
− 2 =0− 2+0
𝑅 𝜕𝑅𝑏 𝑅𝑏
2
𝜕𝑅 𝑅
=
𝜕𝑅𝑏 𝑅𝑏
1 1 1 1 11 180
Consider = + + = , hence 𝑅 = .
𝑅 30 60 90 180 11

180 2
𝜕𝑅 11 9
Therefore, = = .
𝜕𝑅𝑏 60 121
9
Hence, at the given values, a small change in resistance 𝑅𝑏 leads to a change in 𝑅 about as large.
121
Higher order Derivatives
• If f is a function of two variables, then its partial derivatives fx and fy are also functions
of two variables, hence we can consider the four partial derivatives (fx)x, (fx)y, (fy)x, and
(fy)y, which are called the second partial derivatives of f.

If z = f(x, y), we use the following notations:

The third and fourth cases are called mixed partial derivatives.

Theorem (Clairaut’s Theorem): Let 𝑓 be a function of two variables, if 𝑓𝑥 , 𝑓𝑦 , 𝑓𝑥𝑦 , 𝑓𝑦𝑥 are
continuous on an open set (in a neighborhood of (𝑥, 𝑦)), then 𝑓𝑥𝑦 = 𝑓𝑦𝑥 at each point of
the set, i.e. the order of differentiation can be swapped.

Note: The theorem is named after the French mathematician, Alexis Clairaut. To calculate a
mixed second-order derivative, we may differentiate in either order, provided that continuity
conditions are satisfied.
Higher order Derivatives
Proof of Clairaut’s Theorem
Let h and k be sufficiently small values such that the point (𝑎 + ℎ, 𝑏 + 𝑘) lies
in this disk, then so do all points of the rectangle with sides parallel to the
coordinate axes and diagonally opposite corners at (𝑎, 𝑏) and (𝑎 + ℎ, 𝑏 + 𝑘).

Let 𝑄 = 𝑓 𝑎 + ℎ, 𝑏 + 𝑘 − 𝑓 𝑎 + ℎ, 𝑏 − 𝑓 𝑎, 𝑏 + 𝑘 + 𝑓(𝑎, 𝑏) and define

single-variable functions 𝑢(𝑥) and 𝑣(𝑦) as follows:
𝑢 𝑥 = 𝑓 𝑥, 𝑏 + 𝑘 − 𝑓 𝑥, 𝑏
𝑣 𝑦 = 𝑓 𝑎 + ℎ, 𝑦 − 𝑓(𝑎, 𝑦)
Now, we may write 𝑄 = 𝑢 𝑎 + ℎ − 𝑢 𝑎 = 𝑣 𝑏 + 𝑘 − 𝑣(𝑏).
By Mean-Value Theorem, there exists a number 𝜃𝑥 (where 0 < 𝜃𝑥 < 1)
such that 𝑎 + 𝜃𝑥 ℎ lies between 𝑎 and 𝑎 + ℎ, then
𝑄 = 𝑢 𝑎 + ℎ − 𝑢 𝑎 = ℎ𝑢′ 𝑎 + 𝜃𝑥 ℎ
= ℎ 𝑓𝑥 𝑎 + 𝜃𝑥 ℎ, 𝑏 + 𝑘 − 𝑓𝑥 𝑎 + 𝜃𝑥 ℎ, 𝑏
Now, we apply the Mean-Value Theorem to 𝑓𝑥 (considered as a function of
its second variable), and obtain another number 𝜃𝑦 (where 0 < 𝜃𝑦 < 1) such
that
𝑓𝑥 𝑎 + 𝜃𝑥 ℎ, 𝑏 + 𝑘 − 𝑓𝑥 𝑎 + 𝜃𝑥 ℎ, 𝑏 = 𝑘𝑓𝑥𝑦 (𝑎 + 𝜃𝑥 ℎ, 𝑏 + 𝜃𝑦 𝑘)
Thus, 𝑄 = ℎ𝑘𝑓𝑥𝑦 (𝑎 + 𝜃𝑥 ℎ, 𝑏 + 𝜃𝑦 𝑘).
Similar as above, by applying Mean-Value Theorem to 𝑄 = 𝑣 𝑏 + 𝑘 − 𝑣(𝑏) yields
𝑄 = ℎ𝑘𝑓𝑦𝑥 (𝑎 + 𝜃𝑥 ′ ℎ, 𝑏 + 𝜃𝑦′ 𝑘), where 𝜃𝑥 ′ , 𝜃𝑦′ are two numbers, each between 0 and 1.

Equating these two expressions of 𝑄, and cancelling out the common factor ℎ𝑘, we get
𝑓𝑥𝑦 𝑎 + 𝜃𝑥 ℎ, 𝑏 + 𝜃𝑦 𝑘 = 𝑓𝑦𝑥 (𝑎 + 𝜃𝑥 ′ ℎ, 𝑏 + 𝜃𝑦′ 𝑘)

Since 𝑓𝑥𝑦 and 𝑓𝑦𝑥 are continuous at (𝑎, 𝑏), we may simply let ℎ, 𝑘 → 0 to obtain 𝑓𝑥𝑦 𝑎, 𝑏 =
𝑓𝑦𝑥 (𝑎, 𝑏).

Mean-value Theorem
Let 𝑓(𝑥) be differentiable on (𝑎, 𝑏) and continuous on [𝑎, 𝑏], then there exists
𝑓 𝑏 −𝑓 𝑎
at least one point 𝑐 ∈ (𝑎, 𝑏), where 𝑓 ′ 𝑐 = .
𝑏−𝑎
𝑓 𝑎+ℎ −𝑓 𝑎
If 𝑏 = 𝑎 + ℎ, then = 𝑓 ′ 𝑐 = 𝑓′(𝑎 + 𝜃ℎ), where 𝜃 ∈ (0,1).
ℎ
Higher order Derivatives
Example 1: Find 𝑓𝑦𝑥𝑦𝑧 if 𝑓 𝑥, 𝑦, 𝑧 = 1 − 2𝑥𝑦 2 𝑧 2 + 𝑥 2 𝑦 4 .

𝑓𝑦 = −4𝑥𝑦𝑧 2 + 4𝑥 2 𝑦 3
𝑓𝑦𝑥 = −4𝑦𝑧 2 + 8𝑥𝑦 3
𝑓𝑦𝑥𝑦 = −4𝑧 2 + 24𝑥𝑦 2
𝑓𝑦𝑥𝑦𝑧 = −8𝑧

Example 2 (Wave Equation – 2nd order PDE):

The wave equation 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥 describes the motion of a waveform, which could be an ocean wave, a sound
wave, a light wave, or a wave traveling along a vibrating string. Here the constant c depends on the density
of the string and the tension in the string.

u(x, t) represents the displacement of a vibrating violin string at time t and at a distance x from one end
of the string.
U(X,T) REPRESENTS DISPLACEMENT
FROM MEAN POSITION
X=DISTACE FROM THE END OF STRING
T=TIME
C=SPEED
Higher order Derivatives
Show that any function of the form 𝑢 𝑥, 𝑡 = 𝑓 𝑥 + 𝑐𝑡 + 𝑔(𝑥 − 𝑐𝑡) is a solution of the wave equation
𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥

Here we assume both 𝑓 and 𝑔 are twice-differentiable functions of one variable.

Solution:

Using Chain Rule for functions of one variable, we have

𝑢𝑡 = 𝑐𝑓′ 𝑥 + 𝑐𝑡 − 𝑐𝑔′(𝑥 − 𝑐𝑡)

𝑢𝑥 = 𝑓 ′ 𝑥 + 𝑐𝑡 + 𝑔′ 𝑥 − 𝑐𝑡
𝑢𝑡𝑡 = 𝑐 2 𝑓 ′′ 𝑥 + 𝑐𝑡 + 𝑐 2 𝑔′′(𝑥 − 𝑐𝑡)
𝑢𝑥𝑥 = 𝑓 ′′ 𝑥 + 𝑐𝑡 + 𝑔′′(𝑥 − 𝑐𝑡)

Hence, 𝑢𝑡𝑡 = 𝑐 2 𝑢𝑥𝑥 .

Remark: If t measures time, then

𝑓(𝑥 + 𝑐𝑡) represents a waveform travelling to the left with speed c.
𝑔(𝑥 − 𝑐𝑡) represents a waveform travelling to the right with speed c.
Figure from Adams and Essex
Chain Rule (Case 1)
Motivation: Imagine you are hiking in a mountainous region. Let (𝑥, 𝑦) be the coordinates of your
position on the map. Let 𝑧 = 𝑓(𝑥, 𝑦) be the height above sea level at position (𝑥, 𝑦). Suppose you are
𝑥 = 𝑢(𝑡)
walking along a trail such that your position at time 𝑡 is given by ቊ (Parametric equations of the
𝑦 = 𝑣(𝑡)
trail). Then the altitude above sea level at time 𝑡 is : 𝑧 = 𝑓 𝑢 𝑡 , 𝑣 𝑡 = 𝑔(𝑡) , i.e. a function of only one
variable.

Question: How fast does your altitude change with respect to time at time 𝑡?

Answer: Simply the derivative of 𝑔(𝑡).

𝑔 𝑡+ℎ −𝑔(𝑡) 𝑓 𝑢 𝑡+ℎ ,𝑣 𝑡+ℎ −𝑓 𝑢 𝑡 ,𝑣 𝑡
Consider 𝑔′ 𝑡 = lim = lim
ℎ→0 ℎ ℎ→0 ℎ

𝑓 𝑢 𝑡+ℎ ,𝑣 𝑡+ℎ −𝑓 𝑢 𝑡 ,𝑣 𝑡+ℎ 𝑓 𝑢 𝑡 ,𝑣 𝑡+ℎ −𝑓 𝑢 𝑡 ,𝑣 𝑡

= lim + lim
ℎ→0 ℎ ℎ→0 ℎ

By single-variable Chain Rule, we have 𝑔′ 𝑡 = 𝑓𝑥 𝑢 𝑡 , 𝑣 𝑡 𝑢′ 𝑡 + 𝑓𝑦 𝑢 𝑡 , 𝑣 𝑡 𝑣′(𝑡).

Using Leibniz notation, if 𝒛 is a function of 𝒙 and 𝒚 with continuous first partial derivatives, and if
𝒅𝒛 𝝏𝒛 𝒅𝒙 𝝏𝒛 𝒅𝒚 𝝏𝒛 𝝏𝒛 𝒅𝒙 𝒅𝒚
𝒙 and 𝒚 are differentiable functions of 𝒕, then = + = , ∙ , .
𝒅𝒕 𝝏𝒙 𝒅𝒕 𝝏𝒚 𝒅𝒕 𝝏𝒙 𝝏𝒚 𝒅𝒕 𝒅𝒕
Chain Rule (Case 2)
Now, consider a function 𝑓 of two variables (x and y), each of them is a function of two other variables, s
and t, then we can form the composite function
𝑧 = 𝑓 𝑥, 𝑦 = 𝑓 𝑥 𝑠, 𝑡 , 𝑦 𝑠, 𝑡 = 𝑔(𝑠, 𝑡)

Assume that 𝑓, 𝑥, 𝑦 all have first partial derivatives with respect to respective variables, and those partial
derivatives are continuous. Then
𝑧𝑠 = 𝑧𝑥 𝑥𝑠 + 𝑧𝑦 𝑦𝑠 and 𝑧𝑡 = 𝑧𝑥 𝑥𝑡 + 𝑧𝑦 𝑦𝑡

Quick Example: If 𝑓 𝑥, 𝑦 = 𝑥 6 + 3𝑦, where 𝑥 𝑠, 𝑡 = 𝑠𝑡 2 and 𝑦 𝑠, 𝑡 = 𝑠 + 𝑡, then

𝑔 𝑠, 𝑡 = 𝑠 6 𝑡 12 + 3𝑠 + 3𝑡

𝑓𝑠 = 𝑓𝑥 𝑥𝑠 + 𝑓𝑦 𝑦𝑠 = 6𝑥 5 𝑡 2 + 3 1 = 6𝑠 5 𝑡 12 + 3
𝑓𝑡 = 𝑓𝑥 𝑥𝑡 + 𝑓𝑦 𝑦𝑡 = 6𝑥 5 2𝑠𝑡 + 3 1 = 12𝑠 6 𝑡11 + 3

Question: How do you draw out the tree?

In general, if 𝑓(𝑥1 , 𝑥2 , 𝑥3 , … , 𝑥𝑛 ) and 𝑥𝑖 = 𝑥𝑖 (𝑡1 , 𝑡2 , . . , 𝑡𝑚 ) are differentiable, then

𝑛
𝜕𝑓 𝜕𝑓 𝜕𝑥𝑖
=෍
to find partialz/partialt1, all 𝜕𝑡𝑗 𝜕𝑥𝑖 𝜕𝑡𝑗
partial symbol, not d(see 𝑖=1
where 𝑗 = 1,2,3, … , 𝑚. lecture vid)
Chain Rule (Case 2)
In general, if 𝑓(𝑥1 , 𝑥2 , 𝑥3 , … , 𝑥𝑛 ) and 𝑥𝑖 = 𝑥𝑖 (𝑡1 , 𝑡2 , . . , 𝑡𝑚 ) are differentiable, then
𝑛
𝜕𝑓 𝜕𝑓 𝜕𝑥𝑖
=෍
𝜕𝑡𝑗 𝜕𝑥𝑖 𝜕𝑡𝑗
𝑖=1
where 𝑗 = 1,2,3, … , 𝑚.

Particular case: If 𝑤 = 𝑓(𝑥 𝑡 , 𝑦 𝑡 , 𝑧 𝑡 ), then

𝑑𝑤 𝜕𝑓 𝑑𝑥 𝜕𝑓 𝑑𝑦 𝜕𝑓 𝑑𝑧 𝑑𝑥 𝑑𝑦 𝑑𝑧
= + + = 𝑓𝑥 , 𝑓𝑦 , 𝑓𝑧 ∙ , ,
𝑑𝑡 𝜕𝑥 𝑑𝑡 𝜕𝑦 𝑑𝑡 𝜕𝑧 𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡

Examples
(1) Let w = x2y – y2, where x = sin t and y = et. Find dw/dt when t = 0.
𝑑𝑤 𝜕𝑤 𝑑𝑥 𝜕𝑤 𝑑𝑦
= + = 2xy cos t + x 2 − 2y et = 2sin t cos t et + sin2 t − 2et et
𝑑𝑡 𝜕𝑥 𝑑𝑡 𝜕𝑦 𝑑𝑡
At t = 0, the answer will be – 2.
Chain Rule
Example 2. Atmospheric temperature depends on position and time. Denote the spatial position as
(𝑥, 𝑦, 𝑧) (in km) and time by 𝑡 (in hrs), then the temperature is a function of four variables, 𝑇(𝑥, 𝑦, 𝑧, 𝑡). If
a thermometer is attached to a weather balloon that moves through the atmosphere along the curve
𝑥𝑦 𝑥𝑦
𝑥 = 𝑡, 𝑦 = 2𝑡, 𝑧 = 𝑡 − 𝑡 2 , and 𝑇 𝑥, 𝑦, 𝑧, 𝑡 = + 𝑡, find the rate of change of the recorded
1+𝑧 1+𝑧
temperature at 𝑡 = 3.
Solution:
The rate of change of the thermometer reading will depend on the change of
position of the thermometer, as well as increasing time.
𝑑𝑇 𝑑𝑥 𝑑𝑦 𝑑𝑧
Rate = = 𝑇𝑥 + 𝑇𝑦 + 𝑇𝑧 + 𝑇𝑡 ………(*)
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡

The first three terms are arising from the balloon motion, while the last term
refers to the rate of change of the temperature with respect to time at a fixed
position in atmosphere.

At 𝑡 = 3, 𝑥 = 3, 𝑦 = 6, 𝑧 = −6
𝑦 24 𝑥 12
𝑇𝑥 = 1 + 𝑡 = − ; 𝑇𝑦 = 1+𝑡 =− ;
1+𝑧 5 1+𝑧 5
𝑥𝑦 72 𝑥𝑦 18
𝑇𝑧 = − 1 + 𝑡 = − ; 𝑇 𝑡 = = −
1+𝑧 2 25 1+𝑧 5
𝑑𝑥 𝑑𝑦 𝑑𝑧
Also, = 1, = 2, = 1 − 2𝑡 = −5
𝑑𝑡 𝑑𝑡 𝑑𝑡

Substitute everything into (*) and calculate the answer…

Schematic Diagram of Chain Rule

VERY VERY
IMPORTANT

Due to Page 813, Thomas Calculus

𝑑𝑦 𝑑𝑦
𝑑𝑟 𝑑𝑥 𝑑𝑠
𝑑𝑥 𝑑𝑧 𝑑𝑧
𝑑𝑠 𝑑𝑠
𝑑𝑟 𝑑𝑟
Implicit Partial Differentiation
Let x and y be related by the equation F(x, y) = 0, where y = f(x) is a differentiable function of x.

We try to adopt the approach of Chain Rule.

Consider the function w = F(x, y) = F(x, f(x)), we have

𝑑𝑤 𝑑𝑥 𝑑𝑦
= 𝐹𝑥 𝑥, 𝑦 + 𝐹𝑦 𝑥, 𝑦
𝑑𝑥 𝑑𝑥 𝑑𝑥

𝑑𝑤
Since w = F(x, y) = 0 for all x in the domain of f, we have = 0. Also, if Fy(x, y) ≠ 0, we can use the
𝑑𝑥
𝑑𝑦 𝐹𝑥 𝑥,𝑦
fact that dx/dx = 1 to conclude that =− .
𝑑𝑥 𝐹𝑦 𝑥,𝑦

A similar procedure can be used to find the partial derivatives of functions of three variables that are
defined implicitly.

If 𝐹 𝑥, 𝑦, 𝑧 = 0 defines z implicitly as a differentiable function of 𝑥 and 𝑦, we have

𝜕𝑧 𝐹𝑥 𝑥, 𝑦, 𝑧 𝜕𝑧 𝐹𝑦 𝑥, 𝑦, 𝑧
=− ; =−
𝜕𝑥 𝐹𝑧 𝑥, 𝑦, 𝑧 𝜕𝑦 𝐹𝑧 𝑥, 𝑦, 𝑧
Here, we assume 𝐹𝑧 (𝑥, 𝑦, 𝑧) is non-zero.
Examples on Implicit Partial Differentiation
Example 1 (2-variable case)
𝑑𝑦
Find for y3 + 2y2 – 7y – x2 + 12 = 3.
𝑑𝑥

Solution
We let F(x, y) = y3 + 2y2 – 7y – x2 + 9.
Then Fx (x, y) = –2x and Fy (x, y) = 3y2 + 4y – 7.

𝑑𝑦 𝐹𝑥 𝑥,𝑦 2𝑥
Hence, =− = holds for all values of 𝑥 and 𝑦 except when 3𝑦 2 + 4𝑦 − 7 = 0.
𝑑𝑥 𝐹𝑦 𝑥,𝑦 3𝑦 2 +4𝑦−7

Example 2 (3-variable case)

Find 𝑧𝑥 and 𝑧𝑦 if 𝑥 2 𝑧 − 𝑥 2 𝑦 2 + 2𝑧 3 = 1.

Solution
We let 𝐹 𝑥, 𝑦, 𝑧 = 𝑥 2 𝑧 − 𝑥 2 𝑦 2 + 2𝑧 3 − 1, then
𝐹𝑥 𝑥, 𝑦, 𝑧 = 2𝑥𝑧 − 2𝑥𝑦 2 , 𝐹𝑦 = −2𝑥 2 𝑦, 𝐹𝑧 = 𝑥 2 + 6𝑧 2

𝐹 𝑥,𝑦,𝑧 2𝑥𝑦 2 −2𝑥𝑧 𝐹𝑦 𝑥,𝑦,𝑧 2𝑥 2 𝑦

Hence, 𝑧𝑥 = − 𝑥 = and 𝑧𝑦 = − = .
𝐹𝑧 𝑥,𝑦,𝑧 𝑥 2 +6𝑧 2 𝐹𝑧 𝑥,𝑦,𝑧 𝑥 2 +6𝑧 2
Differentials
Increments and Differentials
For a differentiable function of one variable, 𝑦 = 𝑓(𝑥), we define the differential dx to be an
independent variable; i.e. 𝑑𝑥 can be given the value of any real number. The differential of
𝑦 is then defined as
𝑑𝑦 = 𝑓 (𝑥) 𝑑𝑥
Relationship between the increment y and the differential dy: y represents the
change in height of the curve y = f(x) and dy represents the change in height of the tangent
line when x changes by an amount dx = x.
del y=bd, dy=cd
For a differentiable function of two variables, z = f(x, y), (diagram)

we define the differentials dx and dy to be independent

variables; that is, we can assign any real values to
Figure from b
them. Then the differential dz, also called the total
Stewart c
differential, is defined as follows:
d
𝜕𝑧 𝜕𝑧
𝑑𝑧 = 𝑑𝑥 + 𝑑𝑦 = 𝑓𝑥 𝑥, 𝑦 𝑑𝑥 + 𝑓𝑦 𝑥, 𝑦 𝑑𝑦
𝜕𝑥 𝜕𝑦

Now, take dx = x = x – a and dy = y = y – b, then the

differential of z is dz = fx(a, b)(x – a) + fy(a, b)(y – b).
Differentials
Geometric Interpretation

Differential dz and the increment z:

dz represents the change in height of the
tangent plane, whereas z represents the
change in height of the surface z = f(x, y)
when (x, y) changes from (a, b) to
(a + x, b + y).

General Expression of Differential

If 𝑤 = 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑛 ), then the increment of 𝑤 is Figure from Stewart
∆𝑤 = 𝑓 𝑥1 + ∆𝑥1 , 𝑥2 + ∆𝑥2 , … , 𝑥𝑛 + ∆𝑥𝑛 − 𝑓(𝑥1 , 𝑥2 , … , 𝑥𝑛 )

The differential dw is defined in terms of the differentials

𝑑𝑥1 , 𝑑𝑥2 , … , 𝑑𝑥𝑛 of the independent variables by the following
expression:
𝜕𝑤
𝑑𝑤 = 𝑤𝑥1 𝑑𝑥1 + 𝑤𝑥2 𝑑𝑥2 + ⋯ + 𝑤𝑥𝑛 𝑑𝑥𝑛 , where 𝑤𝑥𝑖 : = 𝜕𝑥
𝑖
Examples of Differentials
Example 1
The dimensions of a rectangular box are measured to be 75 cm, 60 cm, and 40 cm, and
each measurement is correct to be within 0.2 cm. Estimate the largest possible error when
the volume of the box is calculated based on all these measurements.

Solution
Let the dimensions of the box be x, y, and z, then its volume is V = xyz.
𝑑𝑉 = 𝑉𝑥 𝑑𝑥 + 𝑉𝑦 𝑑𝑦 + 𝑉𝑧 𝑑𝑧 = 𝑦𝑧 𝑑𝑥 + 𝑥𝑧 𝑑𝑦 + 𝑥𝑦 𝑑𝑧
Since | x |  0.2, | y |  0.2, and | z |  0.2, and our aim is to estimate the largest error in
the volume, we choose 𝑑𝑥 = 0.2, 𝑑𝑦 = 0.2, 𝑑𝑧 = 0.2.
Hence V  dV = (60)(40)(0.2) + (75)(40)(0.2) + (75)(60)(0.2) = 1980
From the above calculation, we notice that an error of only 0.2 cm in each dimension could
lead to an error of approximately 1980 cm3 in the calculated volume.
However, we notice that 𝑉 = 75 60 40 = 180000 𝑐𝑚3 , hence the error is only about
1%.
Examples of Differentials
Example 2
Find the approximate value of the function 𝑓 𝑥, 𝑦 = 𝑥 2 𝑦 3 at (3.1, 0.9) using the concepts of
differential.

Solution:
Let 𝑥 = 3, 𝑦 = 1, then ∆𝑥 = 0.1, ∆𝑦 = −0.1.
Consider 𝑓 𝑥 + ∆𝑥, 𝑦 + ∆𝑦 ≈ 𝑓 𝑥, 𝑦 + 𝑓𝑥 ∆𝑥 + 𝑓𝑦 ∆𝑦
= 𝑥 2 𝑦 3 + 2𝑥𝑦 3 ∆𝑥 + (3𝑥 2 𝑦 2 )(∆𝑦)
Therefore, 𝑓 3.1, 0.9 ≈ 32 1 3 + 2 ∙ 3 ∙ 13 0.1 + 3 ∙ 32 ∙ 12 −0.1 = 6.9

Exact answer = 𝑓 3.1,0.9 = 7.00569

Error = 6.9 − 7.00569 = −0.10569
Differentiability
Definition: In 1-variable calculus, 𝑓 is differentiable at 𝑥0 if there exists a number 𝑓′(𝑥0 ) such that the
following holds:
∆𝑓 = 𝑓 𝑥 + ∆𝑥 − 𝑓 𝑥 = 𝑓 ′ 𝑥 ∆𝑥 + 𝜀∆𝑥, with 𝜀 → 0 as ∆𝑥 → 0

For 2-variable functions, 𝑧 = 𝑓(𝑥, 𝑦) is differentiable at

(𝑥0 , 𝑦0 ) if both 𝑓𝑥 (𝑥0 , 𝑦0 ) and 𝑓𝑦 (𝑥0 , 𝑦0 ) exist, then the
change
∆𝑓 = 𝑓 𝑥0 + ∆𝑥, 𝑦0 + ∆𝑦 − 𝑓(𝑥0 , 𝑦0 )
satisfies the following equation:

∆𝑓 = 𝑓𝑥 𝑥0 , 𝑦0 ∆𝑥 + 𝑓𝑦 𝑥0 , 𝑦0 ∆𝑦 + 𝜀1 ∆𝑥 + 𝜀2 ∆𝑦

Here 𝜀1 , 𝜀2 → 0 as (∆𝑥, ∆𝑦) → (0,0).

Alternative, we say the function 𝑓(𝑥, 𝑦) is differentiable

at the point (𝑥0 , 𝑦0 ) if the partial derivatives 𝑓𝑥 𝑥0 , 𝑦0 and
𝑓𝑦 (𝑥0 , 𝑦0 ) exist, and
𝑓 𝑥0 + ∆𝑥, 𝑦0 + ∆𝑦 − 𝑓 𝑥0 , 𝑦0 − ∆𝑥𝑓𝑥 𝑥0 , 𝑦0 − ∆𝑦𝑓𝑦 𝑥0 , 𝑦0
lim =0
(∆𝑥,∆𝑦)→(0,0) 2
(∆𝑥) + (∆𝑦) 2
Differentiability
Corollary : If the partial derivatives 𝑓𝑥 and 𝑓𝑦 of a function 𝑓(𝑥, 𝑦) are continuous throughout an open region
𝐷, then 𝑓 is differentiable at every point of 𝐷.

If 𝑧 = 𝑓(𝑥, 𝑦) is differentiable, then the definition of differentiability gives us the following result:
∆𝑧 = 𝑓 𝑥0 + ∆𝑥, 𝑦0 + ∆𝑦 − 𝑓(𝑥0 , 𝑦0 ) → 0 as ∆𝑥 and ∆𝑦 both tend to 0.

Theorem (Differentiability implies Continuity): If a function 𝑓(𝑥, 𝑦) is differentiable at a point (𝑥0 , 𝑦0 ), then
𝑓 is continuous at that point.

Example (Show that a function is Differentiable) Show that the function f(x, y) = x2 + 5y is differentiable at
every point in the plane.
Solution:
Let z = f(x, y), the increment of z at an arbitrary point (x, y) in the plane is
∆z = f(x + ∆x, y + ∆y) – f(x, y) = (x + ∆x)2 + 5(y + ∆y) − (x2 + 5y)
= x2 + 2x∆x + (∆x) 2 + 5y + 5∆y – x2 – 5y
= 2x∆x + (∆x)2 + 5∆y
= 2x(∆x) + 5(∆y) + ∆x(∆x) + 0(∆y), which is in the form fx(x, y) ∆x + fy(x, y) ∆y + ε1∆x + ε2∆y

Here ε1= ∆x and ε2= 0. Since ε1 → 0 and ε2 → 0 as (∆x, ∆y) → (0, 0), f is differentiable at every point in the
plane.
Differentiability
𝑥𝑦
if (𝑥, 𝑦) ≠ (0,0)
Consider 𝑓 𝑥, 𝑦 = ቐ𝑥 2+𝑦2 𝑦 = −𝑥
0 if 𝑥, 𝑦 = (0,0)
(1) 𝒇 is not differentiable at (0, 0)
𝑥2 1
Along the line 𝑦 = 𝑥, lim 𝑓 𝑥, 𝑦 = lim =
(𝑥,𝑦)→(0,0) (𝑥,𝑥)→(0,0) 2𝑥 2 2
−𝑥 2 1 𝑦=𝑥
Along the line 𝑦 = −𝑥, lim 𝑓 𝑥, 𝑦 = lim =−
(𝑥,𝑦)→(0,0) (𝑥,−𝑥)→(0,0) 2𝑥 2 2
𝑥𝑦
Hence, limit of 𝑓(𝑥, 𝑦) as 𝑥, 𝑦 → 0,0 does not exist. if (𝑥, 𝑦) ≠ (0,0)
Therefore, f is not continuous at (0, 0), thus not differentiable. 𝑓 𝑥, 𝑦 = ൞𝑥 2 + 𝑦 2
0 if 𝑥, 𝑦 = (0,0)
(2) 𝒇𝒙 (𝟎, 𝟎) and 𝒇𝒚 (𝟎, 𝟎) both exist

𝑓 ∆𝑥,0 −𝑓(0,0)
By definition, 𝑓𝑥 0,0 = lim =0
∆𝑥→0 ∆𝑥
𝑓(0,∆𝑦)−𝑓(0,0)
and 𝑓𝑦 0,0 = lim =0
∆𝑦→0 ∆𝑦

i.e. Partial derivatives at (0, 0) exist.

Gradients of Functions
(1-variable function): Let 𝑦 = 𝑓(𝑥), then the gradient of the function at 𝑥 = 𝑥0 is 𝑓′(𝑥0 ), i.e. the slope of
𝑓(𝑥) at 𝑥 = 𝑥0 .

(2-variable function): Let 𝑧 = 𝑓(𝑥, 𝑦) be a function of 𝑥 and 𝑦 such that both 𝑓𝑥 and 𝑓𝑦 exist. Then the
gradient of 𝑓 is the vector:
∇𝑓 𝑥, 𝑦 = 𝑓𝑥 𝑥, 𝑦 𝐢 + 𝑓𝑦 𝑥, 𝑦 𝐣

Note that for each (𝑥, 𝑦) , the gradient is a vector in the plane (not in space).

(n-variable function): Let 𝑧 = 𝑓 𝑥1 , 𝑥2 , … , 𝑥𝑛 be a continuously differentiable

function. Then the gradient of this function is:
∇𝑓 𝑥1 , 𝑥2 , . . , 𝑥𝑛 = 𝑓𝑥1 , 𝑓𝑥2 , … , 𝑓𝑥𝑛
𝜕 𝜕 𝜕
Here ∇= , ,…,
𝜕𝑥1 𝜕𝑥2 𝜕𝑥𝑛
implies fx, fy exists

Due to Larson and Edwards

Gradients of Functions
Examples
(1) Let 𝑓 𝑥, 𝑦, 𝑧 = 𝑥 3 𝑦 + 𝑦 2 𝑧, then ∇𝑓 = 𝑓𝑥 , 𝑓𝑦 , 𝑓𝑧 = 3𝑥 2 𝑦, 𝑥 3 + 2𝑦𝑧, 𝑦 2 .
At the point (1, 1, 1), ∇𝑓 1, 1, 1 = 3,3,1

𝑦
(2) For 𝑓 𝑥, 𝑦 = 𝑦 ln 𝑥 + 𝑥𝑦 3 , then ∇𝑓 = 𝑓𝑥 , 𝑓𝑦 = + 𝑦 3 , ln 𝑥 + 3𝑥𝑦 2
𝑥
At the point (1, 2), ∇𝑓(1, 2) = 10, 12 .

Note: In general, ∇𝑓 is a vector with 𝑛 components.

Gradient of a differentiable function of two variables at a point (𝑥0 , 𝑦0 )

is always normal to the function’s level curve passing through it.
Gradients of Functions
Theorem: At every point (𝑥0 , 𝑦0 ) in the domain of a differentiable function 𝑓(𝑥, 𝑦), the gradient of 𝑓 is
normal to the level curve of 𝑓 through the point (𝑥0 , 𝑦0 ).

Proof: Suppose 𝑧 = 𝑓 𝑥, 𝑦 = 𝐶 is the level curve (C is a constant).

Let 𝐫 𝑡 = 𝑥 𝑡 , 𝑦(𝑡) be a parametric curve of the level curve, then 𝑓 𝑥 𝑡 , 𝑦 𝑡 = 𝐶.

𝑑 𝑑
Differentiating with respect to 𝑡, we have 𝑓 𝑥 𝑡 ,𝑦 𝑡 = (𝐶).
𝑑𝑡 𝑑𝑡
𝑑𝑥 𝑑𝑦
𝑓𝑥 + 𝑓𝑦 =0
𝑑𝑡 𝑑𝑡
𝑑𝑥 𝑑𝑦
𝑓𝑥 , 𝑓𝑦 ∙ , =0
𝑑𝑡 𝑑𝑡

𝑑𝐫
Hence, gradient of f is normal to the tangent vector , so it is
𝑑𝑡
normal to the curve.

Similarly, if 𝑓 is a function of three variables, then ∇𝑓(𝑥0 , 𝑦0 , 𝑧0 ) is

normal to the level surface of 𝑓 through the point 𝑥0 , 𝑦0 , 𝑧0 .
Gradients of Functions
Example (Ref.: Pg. 926 of Larson and Edwards)
Sketch the level curve corresponding to c = 0 for the function given by f(x, y) = y – sin x and find a normal
vector at several points on the curve.

Solution
The level curve for 𝑐 = 0 is given by

The gradient of f at (x, y) is

f(x, y) = fx(x, y)i + fy(x, y)j = –cos xi + j

Therefore, f(x, y) is normal to the level curve at the point (x, y).

Examples of Gradients:
Gradient Rules
(Ref.: Page 824, Thomas’ Calculus)
END OF WEEK 5

Revised by Dr. Hugo MAK