APSC248 – Engineering Analysis III

Lecture 10 Gradient Vectors

(1) Gradient Vector

Recall the chain rule for multivariable function 𝑤 = 𝑓(𝑥, 𝑦, 𝑧) , where 𝑥 = 𝑥(𝑡), 𝑦 =

𝑦(𝑡), 𝑧 = 𝑧(𝑡)

𝑑𝑤 𝑑𝑥 𝑑𝑦 𝑑𝑧
= 𝑓! + 𝑓" + 𝑓#
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡

We can rewrite this with a new notation. Define a (position) vector 𝑟⃑(𝑡) = 𝑥(𝑡)𝚤̂ +

𝑦(𝑡)𝚥̂ + 𝑧(𝑡)𝑘3. Then its derivative (velocity vector) is

𝑑𝑟⃑(𝑡) 𝑑𝑥(𝑡) 𝑑𝑦(𝑡) 𝑑𝑧(𝑡)

= 𝚤̂ + 𝚥̂ + 𝑘3
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡

Define another vector, ∇𝑤 or ∇𝑓, as

𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓 𝜕𝑓
∇𝑓 = 𝚤̂ + 𝚥̂ + 𝑘3 = 〈 , , 〉
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦 𝜕𝑧

$% $'⃑(&)
Then $&
= ∇𝑓 ⋅ $&
. This newly defined vector ∇𝑓 = 〈𝑓! , 𝑓" , 𝑓# 〉 is called “Gradient

vector” (or Gradient) of function 𝑓(𝑥, 𝑦, 𝑧) at some point (𝑥, 𝑦, 𝑧). Gradient refers to the

derivative of functions of multivariable.

Remarks:

• At each point (𝑥, 𝑦, 𝑧), ∇𝑓 gives a vector. For a certain region of points, ∇𝑓 gives a

so called “vector field”.

• ∇𝑓 is read as “grad 𝑓”, or “gradient of 𝑤”, or “del 𝑓”. ∇ by itself is read as “del”.

• The journey of calculus, 𝑑𝑓 → 𝜕𝑓 → ∇𝑓 .

+ + +
• ∇: vector differential operator: ∇= +! 𝚤̂ + +" 𝚥̂ + +# 𝑘3

@Dr. Yang Cao Page 1

 APSC248 – Engineering Analysis III

• ∇ is a linear operator. Superposition rule is satisfied,

∇(𝑐, 𝑓 + 𝑐- 𝑔) = 𝑐, ∇𝑓 + 𝑐- ∇𝑔

• Recall that function 𝑓(𝑥, 𝑦) around (𝑥. , 𝑦. ) can be approximated with a plane, i.e.,

𝑓(𝑥, 𝑦) ≈ 𝑓(𝑥. , 𝑦. ) + 𝑓! (𝑥. , 𝑦. )(𝑥 − 𝑥. ) + 𝑓" (𝑥. , 𝑦. )(𝑦 − 𝑦. )

= 𝑓(𝑥. , 𝑦. ) + ∇𝑓(𝑥. , 𝑦. ) ⋅ 〈𝑥 − 𝑥. 𝑦 − 𝑦. 〉

From point (𝑥. , 𝑦. ) to nearby point (𝑥, 𝑦) = (𝑥. + Δ𝑥, 𝑦. + Δ𝑦), we have

Δ𝑓 = 𝑓(𝑥. + Δ𝑥, 𝑦. + Δ𝑦) − 𝑓(𝑥. , 𝑦. ) ≈ ∇𝑓(𝑥. , 𝑦. ) ⋅ 〈Δ𝑥 Δ𝑦〉

Example 1: 𝑤 = 𝑓 (𝑥, 𝑦, 𝑧) = 𝑥 ! + 𝑦 ! − 2𝑧 ! + 𝑧 ln 𝑥, find ∇𝑓 and ∇𝑓 at (1,1,1)

Solution:

𝜕𝑓 𝜕𝑓 𝜕𝑓
∇𝑓 = 𝚤̂ + 𝚥̂ + 𝑘3
𝜕𝑥 𝜕𝑦 𝜕𝑧
𝑧
= @2𝑥 + B 𝚤̂ + (2𝑦)𝚥̂ + (−4𝑧 + ln 𝑥)𝑘3
𝑥

∇𝑓|(,,,,,) = (2 + 1)𝚤̂ + (2)𝚥̂ + (−4 + 0)𝑘3 = 3𝚤̂ + 2𝚥̂ − 4𝑘3

@Dr. Yang Cao Page 2

 APSC248 – Engineering Analysis III

Example 2: 𝑤 = 𝑓 (𝑥, 𝑦) = 𝑥 ! − 𝑦 ! , find ∇𝑓 and ∇𝑓 at (1, −1)

Solution:

∇𝑓 = (2𝑥)𝑖̂ + (−2𝑦)𝑗̂

∇𝑓|(,,0,) = 2𝚤̂ + 2𝚥̂

∇𝑓 = 2𝚤̂ + 2𝚥̂

@Dr. Yang Cao Page 3

 APSC248 – Engineering Analysis III

In vector calculus, the gradient of a scalar field is a vector field that points in the direction

of the greatest rate of increase of the scalar field, and whose magnitude is that rate of increase.

In simple terms, the variation in space of any quantity can be represented (e.g. graphically) by a

slope.

Interpretations:

• For example, consider a room in which the temperature is given by a scalar field 𝑇, so at

each point (𝑥, 𝑦, 𝑧) the temperature is 𝑇(𝑥, 𝑦, 𝑧) (we will assume that the temperature

does not change in time). Then, at each point in the room, the gradient of 𝑇 at that point

will show the direction in which the temperature rises most quickly. The magnitude of

the gradient will determine how fast the temperature rises in that direction.

• Consider a surface whose height above sea level at a point (𝑥, 𝑦) is 𝐻(𝑥, 𝑦). The gradient

of 𝐻 at a point is a vector pointing in the direction of the steepest slope at that point. The

steepness of the slope at that point is given by the magnitude of the gradient vector |∇𝑓|

(we’ll learn this in directional derivative).

Summary:

The gradient is a fancy word for derivative, or the rate of change of a function. It’s a vector (a

direction to move) that

• points in the direction of greatest increase of a function

• is zero at a local maximum or local minimum (because there is no single direction of

increase)

We can use concept of directional derivative in Lecture 15 to explain why gradient vector gives

the direction of greatest increase.

@Dr. Yang Cao Page 4

 APSC248 – Engineering Analysis III

Example 3: Watch out for which gradient you are looking for

Given a surface 𝑧 = 𝑓(𝑥, 𝑦) = 𝑥 - + 𝑦 - . We can also say that this surface is one of the level

surfaces of function 𝑤(𝑥, 𝑦, 𝑧) = 𝑧 − 𝑥 - − 𝑦 - when 𝑤 = 0.

• The Gradient vector of function 𝑓(𝑥, 𝑦) gives a planar vector field

∇𝑓 = 2𝑥𝚤̂ + 2𝑦𝚥̂

• The Gradient vector of function 𝑤(𝑥, 𝑦, 𝑧) gives a 3D vector field

∇𝑤 = −2𝑥𝚤̂ − 2𝑦𝚥̂ + 𝑘3

• Given a point 𝑃. (1,1) which is on level curve 𝑧 = 2. We have the gradient vector of

function 𝑓(𝑥, 𝑦) at this point is

∇𝑓|(,,,) = (2𝑥𝚤̂ + 2𝑦𝚥̂)|(,,,) = 2𝚤̂ + 2𝚥̂

This vector points to the direction of greatest increase for variable 𝑧 or function 𝑓(𝑥, 𝑦) at

location 𝑃. (1,1).

@Dr. Yang Cao Page 5

 APSC248 – Engineering Analysis III

• Given a point 𝑃, (1,1,2) which is on the level surface 𝑤 = 0, or the surface 𝑧 = 𝑥 - + 𝑦 - .

We have the gradient vector of function 𝑤(𝑥, 𝑦, 𝑧) at this point is

∇𝑤|(,,,,-) = O−2𝑥𝚤̂ − 2𝑦𝚥̂ + 𝑘3 PQ(,,,,-) = −2𝚤̂ − 2𝚥̂ + 𝑘3 = 〈−2, −2,1〉

This vector points to the direction of greatest increase for variable 𝑤 or function

𝑤(𝑥, 𝑦, 𝑧) at location 𝑃, (1,1,2).

The graphs showing here were generated using Maple to illustrate the differences.

Question: What are the gradient vectors of ∇𝑓 and ∇𝑤 at points (0,0) and (0,0,0), respectively?

Can you interpret the meaning?

@Dr. Yang Cao Page 6

 APSC248 – Engineering Analysis III

Theorem:

(a) Given a differentiable function 𝑧 = 𝑓(𝑥, 𝑦), then

∇𝑓 ⊥ level curves 𝑓(𝑥, 𝑦) = 𝐶

∇𝑓 ⊥ tangent lines of level curves 𝑓(𝑥, 𝑦) = 𝐶

(b) Given a differentiable function 𝑤 = 𝑓(𝑥, 𝑦, 𝑧), then

∇𝑓 ⊥ level surfacess 𝑓(𝑥, 𝑦, 𝑧) = 𝐶

∇𝑓 ⊥ tangent planes of level surfacess 𝑓(𝑥, 𝑦, 𝑧) = 𝐶

For example: Given 𝑧 = 𝑓(𝑥, 𝑦). Suppose its level curves or contour plot is shown

below. Then a gradient at any point (𝑥, 𝑦) gives a vector perpendicular to that level

curve.
∇𝑓
𝑦

Example 4: 𝑤 = 𝑓 (𝑥, 𝑦, 𝑧) = 𝑎" 𝑥 + 𝑎! 𝑦 + 𝑎# 𝑧, find ∇𝑤 and show that ∇𝑤 is

perpendicular to level surfaces.

∇𝑤 = 〈𝑎" , 𝑎! , 𝑎# 〉

The level surface is 𝑓 (𝑥, 𝑦, 𝑧 ) = 𝐶 or 𝑎" 𝑥 + 𝑎! 𝑦 + 𝑎# 𝑧 = 𝐶

This is a plane equation in 3D space and the normal vector to this plane is

〈𝑎, , 𝑎- , 𝑎1 〉

Hence the gradient vector ∇𝑤 is perpendicular to the level surface 𝑤 = 𝐶.

@Dr. Yang Cao Page 7

 APSC248 – Engineering Analysis III

Example 5: 𝑤 = 𝑓 (𝑥, 𝑦) = 𝑥 ! + 𝑦 ! , find ∇𝑤 and show that ∇𝑤 is

perpendicular to level curves 𝑤 = 𝐶.

Solution:

∇𝑤 = 〈2𝑥 2𝑦〉

The level curves are given by 𝑤 = 𝐶 or

𝑥- + 𝑦- = 𝐶

which is a cluster of concentric circles. And the gradient vectors are in the direction from origin

to the tangent point, which are perpendicular to the tangent line or we can say they are

perpendicular to the level curve at point (𝑥, 𝑦).

𝑦 ∇𝑤

@Dr. Yang Cao Page 8

 APSC248 – Engineering Analysis III

Example 6: 𝑧 = 𝑓 (𝑥, 𝑦) = 4 − (cos ! 𝑥 + cos ! 𝑦 )! , find ∇𝑓.

The gradient of the function 𝑓(𝑥, 𝑦) = 4 − (cos - 𝑥 + cos - 𝑦 )- depicted as a vector field on the

bottom plane. Its gradient

∇𝑓 = 𝑓! 𝚤̂ + 𝑓" 𝚥̂

= −2(cos - 𝑥 + cos - 𝑦)(2 cos 𝑥)(− sin 𝑥)𝚤̂ − 2(cos - 𝑥 + cos - 𝑦)(2 cos 𝑦)(− sin 𝑦)𝚥̂

= 2(cos - 𝑥 + cos - 𝑦) sin 2𝑥 𝚤̂ + 2(cos - 𝑥 + cos - 𝑦) sin 2𝑦 𝚥̂

Apparently, this function has a global minimum at (0,0). At point (0,0), the gradient is

∇𝑓|(.,.) = 0

@Dr. Yang Cao Page 9

 APSC248 – Engineering Analysis III

Matlab Code:

x = -2:0.2:2; y = x;

[X,Y] = meshgrid(x,y);

f = 4 - (cos(X).^2+cos(Y).^2).^2

mesh(X,Y,f);

hold on

hidden off;

%contour(x,y, f, 'color','r'); hold on

%hidden off;

[Dx,Dy] = gradient(f);

quiver(x , y, Dx, Dy, 'linewidth', 1);

title('f(x,y) = 4 - (cos(x)^2+cos(y)^2)^2, gradient of f');

@Dr. Yang Cao Page 10

 APSC248 – Engineering Analysis III

(2) Tangent Planes and Normal Lines for Level Surfaces

Given a function of three variables, 𝑤 = 𝑓(𝑥, 𝑦, 𝑧). If 𝑟⃑(𝑡) = 𝑥(𝑡)𝚤̂ + 𝑦(𝑡)𝚥̂ + 𝑧(𝑡)𝑘3 is a
smooth curve on the level surface 𝑓(𝑥, 𝑦, 𝑧) = 𝐶, then 𝑓O𝑥(𝑡), 𝑦(𝑡), 𝑧(𝑡)P = 𝐶. Differentiating
both sides with respect to 𝑡 and with chain rule, we have
𝑑 𝑑
𝑓O𝑥(𝑡), 𝑦(𝑡), 𝑧(𝑡)P = (𝐶)
𝑑𝑡 𝑑𝑡
𝑑𝑥 𝑑𝑦 𝑑𝑧
𝑓! + 𝑓" + 𝑓# =0
𝑑𝑡 𝑑𝑡 𝑑𝑡
𝑑𝑟⃑ 𝑑𝑟⃑
∇𝑓 ⋅ = 0 ⟹ ∇𝑓 ⊥ c = 𝑣⃑e
𝑑𝑡 𝑑𝑡
Comments:

• At every point along the curve, ∇𝑓 is perpendicular to the curve’s velocity vector. The

result also applies for any other curves on the surface.

• Now let’s focus on a point 𝑃. (𝑥. , 𝑦. , 𝑧. ) on the surface and curves passing through 𝑃. .

For all curves on the surface through 𝑃. , the velocity vectors at 𝑃. are orthogonal to ∇𝑓 at

𝑃. .

• Hence the curves’ tangent lines all lie in the plane through 𝑃. and normal to ∇𝑓. We call

this plane “the tangent plane of the surface at 𝑃. ”.

@Dr. Yang Cao Page 11

 APSC248 – Engineering Analysis III

Definition: Tangent Plane

The tangent plane at the point 𝑃(𝑥. , 𝑦. , 𝑧. ) on the level surface 𝑓O𝑥(𝑡), 𝑦(𝑡), 𝑧(𝑡)P = 𝐶 is the

plane through 𝑃. normal to ∇𝑓|2! . The plane equation is

𝑓! (𝑃. )(𝑥 − 𝑥. ) + 𝑓" (𝑃. )(𝑦 − 𝑦. ) + 𝑓# (𝑃. )(𝑧 − 𝑧. ) = 0

Definition: Normal Line

The normal line of the surface at 𝑃(𝑥. , 𝑦. , 𝑧. ) is the line through 𝑃. parallel to vector ∇𝑓|2! . The

normal line equation is

𝑥 = 𝑥. + 𝑓! (𝑃. )𝑡
f𝑦 = 𝑦. + 𝑓" (𝑃. )𝑡
𝑧 = 𝑧. + 𝑓# (𝑃. )𝑡

Note, if ∇𝑓|2! = 0, then we are at local minimum or maximum. The tangent plane will be

parallel to 𝑥𝑦-plane. The equation for tangent plane is

𝑧 = 𝑧.

The equation for normal line is

𝑥 = 𝑥.
g𝑦 = 𝑦. , − ∞ < 𝑡 < ∞
𝑧=𝑡

@Dr. Yang Cao Page 12

 APSC248 – Engineering Analysis III

Example 7: Find tangent plane to surface 𝑥 ! + 𝑦 ! − 𝑧 ! = 4 at (2,1,1)

Solution:

Here 𝑓(𝑥, 𝑦, 𝑧) = 𝑥 - + 𝑦 - − 𝑧 - . Then the gradient vector for this surface at (2,1,1) is

∇𝑓|(-,,,,) = 〈2𝑥, 2𝑦, −2𝑧〉|(-,,,,)

= 〈4, 2, −2〉

The tangent plant is then

4(𝑥 − 2) + 2(𝑦 − 1) + (−2)(𝑧 − 1) = 0

4𝑥 + 2𝑦 − 2𝑧 = 8

The normal line is

𝑥 = 2 + 4𝑡
k𝑦 = 1 + 2𝑡
𝑧 = 1 − 2𝑡

@Dr. Yang Cao Page 13

 APSC248 – Engineering Analysis III

We can also use gradient of a function 𝑓(𝑥, 𝑦) to obtain the its linearization or approximation of

𝑓(𝑥, 𝑦) at point 𝑃. (𝑥. , 𝑦. ). Suppose that from a level surface 𝑓(𝑥, 𝑦, 𝑧) = 𝐶, we can explicitly

express 𝑧 as function 𝑓(𝑥, 𝑦). In another word, the level surface is rewritten as 𝑧 − 𝑓(𝑥, 𝑦) = 0.

Define 𝑤 = 𝑧 − 𝑓(𝑥, 𝑦). For this level surface, the gradient vector is

𝜕𝑤 𝜕𝑤 𝜕𝑤 𝜕𝑓 𝜕𝑓
∇𝑤 = 𝚤̂ + 𝚥̂ + 𝑘3 = − 𝚤̂ − 𝚥̂ + 𝑘3
𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑦

At 𝑃. ,

𝜕𝑓 𝜕𝑓
∇𝑤|2! = − l 𝚤̂ − l 𝚥̂ + 𝑘3 = −𝑓! (𝑃. )𝚤̂ − 𝑓" (𝑃. )𝚥̂ + 𝑘3
𝜕𝑥 2! 𝜕𝑦 2
!

Then the tangent plane at 𝑃. is

−𝑓! (𝑃. )(𝑥 − 𝑥. ) − 𝑓" (𝑃. )(𝑦 − 𝑦. ) + (𝑧 − 𝑧. ) = 0

𝑧 = 𝑧. + 𝑓! (𝑃. )(𝑥 − 𝑥. ) + 𝑓" (𝑃. )(𝑦 − 𝑦. )

which is an approximation of the surface 𝑧 = 𝑓(𝑥, 𝑦) at 𝑃. . We can also say the above equation

is a linearization of the surface 𝑧 = 𝑓(𝑥, 𝑦) at 𝑃. . This is the same result as we derived in

Lecture 13.

Recall that in 2D space, the approximation of a curve 𝑦 = 𝑓(𝑥) at point 𝑃. (𝑥. , 𝑦. ) is

𝑓(𝑥) ≈ 𝑓(𝑥. ) + 𝑓 3 (𝑥. )(𝑥 − 𝑥. )

@Dr. Yang Cao Page 14

 APSC248 – Engineering Analysis III

(3) Tangent Lines for Level Curves (Optional)

Given a differentiable function, 𝑧 = 𝑓(𝑥, 𝑦). The level curve are then 𝑓(𝑥, 𝑦) = 𝐶.

𝑦 ∇𝑤

𝑟⃑(𝑡)

If part of or whole of a level curve 𝑓(𝑥, 𝑦) = 𝐶 is represented using parametric function or

vector 𝑟⃑(𝑡) = 𝑥(𝑡)𝚤̂ + 𝑦(𝑡)𝚥̂, then 𝑓(𝑥(𝑡), 𝑦(𝑡)) = 𝐶. Using chain rule, we have

𝑑 𝑑
𝑓O𝑥(𝑡), 𝑦(𝑡)P = (𝐶)
𝑑𝑡 𝑑𝑡
𝑑𝑥 𝑑𝑦
𝑓! + 𝑓" =0
𝑑𝑡 𝑑𝑡
𝑑𝑥 𝑑𝑦
O𝑓! 𝚤̂ + 𝑓" 𝚥̂P ⋅ m 𝚤̂ + 𝚥̂n
𝑑𝑡 𝑑𝑡
𝑑𝑟⃑ 𝑑𝑟⃑
∇𝑓 ⋅ = 0 ⟹ ∇𝑓 ⊥ c = 𝑣⃑e
𝑑𝑡 𝑑𝑡
At every point (𝑥. , 𝑦. ) in the domain of the differentiable function 𝑓(𝑥, 𝑦), the gradient of

𝑓 is normal to the level curve through (𝑥. , 𝑦. ).

• Equation of tangent line to level curve at point 𝑃.

At the point 𝑃. (𝑥. , 𝑦. ) on the curve, the gradient vector is ∇𝑓 = 𝑓! (𝑥. , 𝑦. )𝚤̂ + 𝑓" (𝑥. , 𝑦. )𝚥̂.

∇𝑓 is perpendicular to the tangent line through 𝑃. (𝑥. , 𝑦. ).

@Dr. Yang Cao Page 15

 APSC248 – Engineering Analysis III

o⃑ = 𝑃
If 𝑃(𝑥, 𝑦) is another point on the tangent line, then we can define 𝑇 ooooooo⃑
. 𝑃 = (𝑥 − 𝑥. )𝚤̂ +

(𝑦 − 𝑦. )𝚥̂, which is a vector parallel to the tangent line. Thus

o⃑ = 0
∇𝑓 ⋅ 𝑇

𝑓! (𝑥. , 𝑦. )(𝑥 − 𝑥. ) + 𝑓" (𝑥. , 𝑦. )(𝑦 − 𝑦. ) = 0

This is the equation of the tangent line.

$!
Example 8: Given a level curve %
+ 𝑦 ! = 2. Find the tangent line at point
(−2,1)

𝑥-
𝑓(𝑥, 𝑦) = + 𝑦-
4
𝑥
∇𝑓|(0-,,) = @ 𝚤̂ + 2𝑦𝚥̂Bp = −𝚤̂ + 2𝚥̂
2 (0-,,)

Thus the tangent line equation is

(−1)(𝑥 + 2) + 2(𝑦 − 1) = 0

𝑥 − 2𝑦 = −4

@Dr. Yang Cao Page 16