Functions of several variables

Samy Tindel

Purdue University

Multivariate calculus - MA 261

Mostly taken from Calculus, Early Transcendentals

by Briggs - Cochran - Gillett - Schulz

Samy T. Several variables Multivariate calculus 1 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 2 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 3 / 145

Recalling functions of 1 variable (1)

Example of function:
√
y = f (x ) = 9 − x2

Questions:
1 Domain of f ?
2 Range of f ?

Samy T. Several variables Multivariate calculus 4 / 145

Recalling functions of 1 variable (2)

Recalling the function:

√
y = f (x ) = 9 − x2

Domain:
x ∈ [−3, 3]

Range:
y ∈ [0, 3]

Samy T. Several variables Multivariate calculus 5 / 145

Functions of 2 variables: example (1)

Example of function:
√ q
z = f (x , y ) = 9 − x2 − 25 − y 2

Questions:
1 Domain of f ?
2 Range of f ?

Samy T. Several variables Multivariate calculus 6 / 145

Functions of 2 variables: example (2)

Recalling the function:

√ q
z = f (x , y ) = 9 − x2 − 25 − y 2

Domain:
(x , y ) ∈ [−3, 3] × [−5, 5]

Range: Looking at lines x = ±3 and y = ±5, we get

y ∈ [−5, 3]

Samy T. Several variables Multivariate calculus 7 / 145

Contour and level curves

Definition 1.
Contour curve:
Intersection of the surface (x , y , f (x , y )) and plane z = z0

Level curve:
Projection of contour curve on xy -plane

Samy T. Several variables Multivariate calculus 8 / 145

Contour and level curves: illustration

Samy T. Several variables Multivariate calculus 9 / 145

Example of level curves (1)

Function:
f (x , y ) = y − x 2 − 1

Samy T. Several variables Multivariate calculus 10 / 145

Example of level curves (2)
Function:
f (x , y ) = y − x 2 − 1

Level curves: For z0 ∈ R, we get the parabola

y = x 2 + 1 + z0

Samy T. Several variables Multivariate calculus 11 / 145

Example 2 of level curves (1)

Function:  
f (x , y ) = exp −x 2 − y 2

Samy T. Several variables Multivariate calculus 12 / 145

Example 2 of level curves (2)
Function:  
f (x , y ) = exp −x 2 − y 2

Level curves: For z0 ∈ (0, 1], we get the circle

x 2 + y 2 = − ln(z0 )

Samy T. Several variables Multivariate calculus 13 / 145

Example 3 of level curves (1)

Function:
f (x , y ) = 2 + sin(x − y )

Samy T. Several variables Multivariate calculus 14 / 145

Example 3 of level curves (2)
Function:
f (x , y ) = 2 + sin(x − y )

Level curves:
For z0 ∈ [1, 3], we get a family of lines

Level curves for z0 = 2:

y = x − k π, k∈Z

Level curves for z0 = 1:

π
y =x− + 2k π, k∈Z
2

Samy T. Several variables Multivariate calculus 15 / 145

Example 3 of level curves (3)
Function:
f (x , y ) = 2 + sin(x − y )

Depiction of level curves:

Samy T. Several variables Multivariate calculus 16 / 145

Application of functions of 2 variables (1)

Situation:
Fraction of students infected by FV is r on 9/12
We have n random encounters with students on 9/12

Function:
The probability of meeting at least one student with FV is

p(n, r ) = 1 − (1 − r )n

This requires probability theory and is admitted

Question:
Draw level curves

Samy T. Several variables Multivariate calculus 17 / 145

Application of functions of 2 variables (2)

Function:
p(n, r ) = 1 − (1 − r )n

Useful values of z:
For p0 ∈ [0, 1], the curve p(n, r ) = p0 is non empty

Level curves for p0 ∈ [0, 1]:

r = 1 − (1 − p)1/n

Samy T. Several variables Multivariate calculus 18 / 145

Application of functions of 2 variables (3)
Function:
p(n, r ) = 1 − (1 − r )n

Depiction of level curves:

Samy T. Several variables Multivariate calculus 19 / 145

Application of functions of 2 variables (4)
Function:
p(n, r ) = 1 − (1 − r )n

Depiction of function:

Samy T. Several variables Multivariate calculus 20 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 21 / 145

Continuity for functions of 1 variable (1)

Limit: The assertion

lim f (x ) = L
x →a

means that f (x ) can be made as close to L as we wish

,→ by making x close to a

Remark: If limx →a f (x ) = L, then

the limit should not depend on the way x → a

Samy T. Several variables Multivariate calculus 22 / 145

Continuity for functions of 1 variable (2)

Continuity: The function f is continuous at point a if

lim f (x ) = f (a)
x →a

Examples of continuous functions:

Polynomials
sin, cos, exponential
Rational functions (on their domain)
Log functions (on their domain)

Samy T. Several variables Multivariate calculus 23 / 145

Continuity for functions of 2 variables (1)

Limit: The assertion

lim f (x , y ) = L
(x ,y )→(a,b)

means that f (x , y ) can be made as close to L as we wish

,→ by making (x , y ) close to (a, b)

Remark: If lim(x ,y )→(a,b) f (x , y ) = L, then

the limit should not depend on the way (x , y ) → (a, b)

Samy T. Several variables Multivariate calculus 24 / 145

Continuity for functions of 2 variables (2)

Continuity: The function f is continuous at point a if

lim f (x , y ) = f (a, b)
(x ,y )→(a,b)

Examples of continuous functions:

Polynomials
sin, cos, exponential
Rational functions (on their domain)
Log functions (on their domain)

Samy T. Several variables Multivariate calculus 25 / 145

Logarithmic example (1)

Function:
1 + y2
!
ln
x2

Problem: Continuity at point

(1, 0)

Samy T. Several variables Multivariate calculus 26 / 145

Logarithmic example (2)

Continuity: f is the log of a rational function

,→ Continuous wherever it is defined

Definition at point (1, 0): We have

f (1, 0) = 0

This is well defined

Conclusion: f is continuous at (1, 0), that is

lim f (x , y ) = f (1, 0) = 0
(x ,y )→(1,0)

Samy T. Several variables Multivariate calculus 27 / 145

Rational function example (1)

Function:
y 2 − 4x 2
f (x , y ) =
2x 2 + y 2

Problem: Continuity at point

(0, 0)

Samy T. Several variables Multivariate calculus 28 / 145

Rational function example (2)

Continuity: f is a rational function

,→ Continuous wherever it is defined

Definition at point (0, 0): We have

0
f (0, 0) =
0
This is not well defined, therefore general result cannot be applied

Samy T. Several variables Multivariate calculus 29 / 145

Rational function example (3)

Two paths: We have

y 2 − 4x 2
Along x = 0, lim = 1
(x ,y )→(0,0), x =0 2x 2 + y 2
y 2 − 4x 2
Along y = 0, lim = −2
(x ,y )→(0,0), y =0 2x 2 + y 2
We get 2 different limits

Conclusion:
f is not continuous at point (0, 0)

Samy T. Several variables Multivariate calculus 30 / 145

Another rational function example (1)

Function:
x2 − y2
f (x , y ) =
x +y

Problem: Continuity at point

(0, 0)

Samy T. Several variables Multivariate calculus 31 / 145

Another rational function example (2)

Continuity: f is a rational function

,→ Continuous wherever it is defined

Definition at point (0, 0): We have

0
f (0, 0) =
0
This is not well defined, therefore general result cannot be applied

Samy T. Several variables Multivariate calculus 32 / 145

Another rational function example (3)

Two paths: We have

x2 − y2
Along x = 0, lim = 0
(x ,y )→(0,0), x =0 x +y
x2 − y2
Along y = 0, lim = 0
(x ,y )→(0,0), y =0 x +y
We get the same limit

Partial conclusion:
This is not enough!

Samy T. Several variables Multivariate calculus 33 / 145

Another rational function example (4)
Next steps: Try different paths
y = x 2 , y = x 3 , etc
Those all give a 0 limit
This is still not enough

Key remark: If (x , y ) ̸= (0, 0) we have

x2 − y2
f (x , y ) = =x −y
x +y
The rhs above is continuous

Conclusion: We have

lim f (x , y ) = 0
(x ,y )→(0,0)

Samy T. Several variables Multivariate calculus 34 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 35 / 145

Motivation
Derivative for functions of 1 variable: Captures the rate of change

f (x + h) − f (x )
f ′ (x ) = lim
h→0 h

Rate of change in the 2-d case: Can be different in x and y directions

,→ Captured by partial derivatives

Samy T. Several variables Multivariate calculus 36 / 145

Partial derivatives

Definition 2.
Consider
f function of 2 variables
Then we set
f (x + h, y ) − f (x , y )
fx (x , y ) = lim
h→0 h
f (x , y + h) − f (x , y )
fy (x , y ) = lim
h→0 h

Samy T. Several variables Multivariate calculus 37 / 145

Some remarks on partial derivatives

Frozen and live variables:

In order to compute fx (x , y )
,→ the x variable is alive and the y variable is frozen
In order to compute fy (x , y )
,→ the y variable is alive and the x variable is frozen

Funny notation: For partial derivatives we also use

∂f ∂f
(x , y ) = fx (x , y ), (x , y ) = fy (x , y )
∂x ∂y

Samy T. Several variables Multivariate calculus 38 / 145

Example of computation (1)

Function:
f (x , y ) = x 8 y 5 + x 3 y

Samy T. Several variables Multivariate calculus 39 / 145

Example of computation (2)

Recall:
f (x , y ) = x 8 y 5 + x 3 y

Partial derivative fx :
fx = 8x 7 y 5 + 3x 2 y

Partial derivative fy :
fy = 5x 8 y 4 + x 3

Samy T. Several variables Multivariate calculus 40 / 145

Second example of computation (1)

Function:
f (x , y ) = e x sin(y )

Samy T. Several variables Multivariate calculus 41 / 145

Second example of computation (2)

Recall:
f (x , y ) = e x sin(y )

Partial derivative fx :
fx = e x sin(y )

Partial derivative fy :
fy = e x cos(y )

Samy T. Several variables Multivariate calculus 42 / 145

Second derivatives
Second derivative fxx , fyy :

∂ 2f ∂ 2f
fxx = (fx )x = , fyy = (fy )y =
∂x 2 ∂y 2

Second derivative fxy :

∂ 2f
fxy = (fx )y =
∂x ∂y

Second derivative fyx :

∂ 2f
fyx = (fy )x =
∂y ∂x

Samy T. Several variables Multivariate calculus 43 / 145

Example of second derivatives

Function:
f (x , y ) = e x sin(y )

Second derivative fxx :

fxx = (fx )x = e x sin(y )

Second derivative fxy :

fxy = (fx )y = e x cos(y )

Samy T. Several variables Multivariate calculus 44 / 145

Order of derivatives

On our running example: We have

fyx = (fy )x = e x cos(y ) = fxy

General result (Clairaut’s theorem):

For a smooth f , the order of the derivatives does not matter

fyx = fxy

Samy T. Several variables Multivariate calculus 45 / 145

Example of order of derivatives (1)

Function:
2y
f (x , y ) = e x

Problem: Check that

fyx = fxy

Samy T. Several variables Multivariate calculus 46 / 145

Example of order of derivatives (2)

Recall:
2y
f (x , y ) = e x

Partial derivative fx :
2y
fx = 2xy e x

Partial derivative fy :
2y
fy = x 2 e x

Mixed derivatives:
  2y
fyx = fxy = 2x x 2 y + 1 e x

Samy T. Several variables Multivariate calculus 47 / 145

Functions of 3 variables (1)

Basic rule: Functions of 3 variables are handled

,→ in the same way as functions of 2 variables

Example:
f (x , y , z) = xyz

First derivatives:

fx = yz, fy = xz, fz = xy

Samy T. Several variables Multivariate calculus 48 / 145

Functions of 3 variables (2)

Second derivatives: We have for instance

fxy = fyx = z

Third derivatives: The only non zero derivatives are

fxyz = fxzy = · · · = fzyx = 1

Samy T. Several variables Multivariate calculus 49 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 50 / 145

Chain rule for functions of 1 variable

Situation: We have
y = f (x )
x = g(t)

Chain rule:
dy dy dx
=
dt dx dt

Samy T. Several variables Multivariate calculus 51 / 145

Chain rule with 1 independent variable

Theorem 3.
Let
z = z(x , y )
x = x (t) and y = y (t)
z, x , y differentiable

Then
dz ∂z dx ∂z dy
= +
dt ∂x dt ∂y dt

Samy T. Several variables Multivariate calculus 52 / 145

Tree representation of chain rule (2d)

Samy T. Several variables Multivariate calculus 53 / 145

Tree representation of chain rule (3d)

Samy T. Several variables Multivariate calculus 54 / 145

Example of computation (1)
Functions: We consider

z = x 2 − 3y 2 + 20, x = 2 cos(t), y = 2 sin(t)

Derivative: We find
dz ∂z dx ∂z dy
= +
dt ∂x dt ∂y dt
= −16 sin(2t)

Particular value: It t = π4 , then

dz π
 
= −16
dt 4

Samy T. Several variables Multivariate calculus 55 / 145

Example of computation (2)
Other possible strategy:
1 Express z(x (t), y (t)) as a function F (t)
2 Differentiate as usual
Problem: this becomes impractical very soon.

Samy T. Several variables Multivariate calculus 56 / 145

Implicit differentiation

Theorem 4.
Let F (x , y ) be such that
F differentiable
The equation F (x , y ) = 0 defines y = y (x )
x 7→ y (x ) differentiable
Fy ̸= 0
Then we have
dy Fx
=−
dx Fy

Samy T. Several variables Multivariate calculus 57 / 145

Example of implicit differentiation (1)

Equation:
e y sin(x ) = x + xy

Problem: Find
dy
dx

Samy T. Several variables Multivariate calculus 58 / 145

Example of implicit differentiation (2)

Reformulation of the equation: F (x , y ) = 0 with

F (x , y ) = e y sin(x ) − x − xy

Implicit differentiation:

dy Fx e y cos(x ) − 1 − y
=− =−
dx Fy e y sin(x ) − x

Samy T. Several variables Multivariate calculus 59 / 145

Implicit differentiation with 3 variables (1)

Implicit equation: We consider

F (x , y , z) = xy + yz + xz
Equation: F (x , y ) = 3
The equation defines z = z(x , y )

Problem: Find
∂z
∂y

Samy T. Several variables Multivariate calculus 60 / 145

Implicit differentiation with 3 variables (2)

Implicit differentiation:
∂z Fy x +z
=− =−
∂y Fz y +x

Samy T. Several variables Multivariate calculus 61 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 62 / 145

Objective
Aim: Understand variations of a function
,→ In directions which are not parallel to the axes

Samy T. Several variables Multivariate calculus 63 / 145

Directional derivative

Definition 5.
Let
f differentiable function at (a, b)
u = ⟨u1 , u2 ⟩ unit vector in xy -plane
Then the directional derivative of f
in the direction of u at (a, b) is

f (a + hu1 , b + hu2 ) − f (a, b)

Du f (a, b) = lim
h→0 h

Samy T. Several variables Multivariate calculus 64 / 145

Computation of the directional derivative

Proposition 6.
Let
f differentiable function at (a, b)
u = ⟨u1 , u2 ⟩ unit vector in xy -plane
Then the directional derivative of f
in the direction of u at (a, b) is given by

Du f (a, b) = fx (a, b)u1 + fy (a, b)u2

Remark: One can also write

Du f (a, b) = ⟨fx (a, b), fy (a, b)⟩ · ⟨u1 , u2 ⟩

Samy T. Several variables Multivariate calculus 65 / 145

Example of directional derivative (1)

Function: Paraboloid of the form

1 2 
z = f (x , y ) = x + 2y 2 + 2
4

Unit vector: * +
1 1
u= √ ,√
2 2

Problem: Compute the directional derivative

Du f (3, 2)

Samy T. Several variables Multivariate calculus 66 / 145

Example of directional derivative (2)

Function: Paraboloid of the form

1 2 
z = f (x , y ) = x + 2y 2 + 2
4

Unit vector: * +
1 1
u= √ ,√
2 2

Directional derivative: We get

 * +
3 1 1 7

Du f (3, 2) = ,2 · √ , √ = √ ≃ 2.47
2 2 2 2 2

Samy T. Several variables Multivariate calculus 67 / 145

Example of directional derivative (3)

Samy T. Several variables Multivariate calculus 68 / 145

Gradient

Definition 7.
Let
f differentiable function at (x , y )
Then the gradient of f at (x , y ) is

∇f (x , y ) = ⟨fx (x , y ), fy (x , y )⟩

Samy T. Several variables Multivariate calculus 69 / 145

Example of gradient (1)

Function:
x 2 xy 2
f (x , y ) = 3 − +
10 10

Problem:
1 Compute ∇f (3, −1)
2 Compute the directional derivative of f
,→ at (3, −1) in the direction of the vector ⟨3, 4⟩

Samy T. Several variables Multivariate calculus 70 / 145

Example of gradient (2)

Gradient:
y 2 xy
* +
x
∇f (x , y ) = − + ,
5 10 5
Thus
1 3
 
∇f (3, −1) = − , −
2 5

Samy T. Several variables Multivariate calculus 71 / 145

Example of gradient (3)

Directional derivative: Unit vector in direction of ⟨3, 4⟩ is

3 4
 
u= ,
5 5
Thus directional derivative in direction of ⟨3, 4⟩ is

Du f (x , y ) = ∇f (x , y ) · u

We get
39
Du f (3, −1) = −
50

Samy T. Several variables Multivariate calculus 72 / 145

Interpretation of gradient
Remark: If
u is a unit vector
θ ≡ angle between u and ∇f (x , y )
Then
Du f (x , y ) = |∇f (x , y )| cos(θ)

Information given by the gradient

1 |∇f (x , y )| is the maximal possible directional derivative
∇f (x ,y )
2 The direction u = |∇f (x ,y )|
is the one of maximal ascent
∇f (x ,y )
3 The direction u = − |∇f (x ,y )|
is the one of maximal desccent
4 If u ⊥ ∇f (x , y ), the directional derivative is 0

Samy T. Several variables Multivariate calculus 73 / 145

Interpretation of gradient: illustration

Samy T. Several variables Multivariate calculus 74 / 145

Example of steepest descent (1)
Function:
f (x , y ) = 4 + x 2 + 3y 2

Questions:
1 If you are located on the paraboloid at the point (2, − 21 , 35
4
)
,→ In which direction should you move in order to ascend on the
surface at the maximum rate?
2 If you are located on the paraboloid at the point (2, − 21 , 35
4
)
,→ In which direction should you move in order to descend on
the surface at the maximum rate?
3 At the point (3, 1, 16), in what direction(s) is there no change
in the function values?

Samy T. Several variables Multivariate calculus 75 / 145

Example of steepest descent (2)
Gradient:
∇f (x , y ) = ⟨2x , 6y ⟩
Thus
1
 
∇f 2, − = ⟨4, −3⟩
2

Steepest ascent direction: We get

4 3
 
u= ,− ,
5 5
with rate of ascent
1
 
∇f 2, − =5
2

Samy T. Several variables Multivariate calculus 76 / 145

Example of steepest descent (3)

Steepest descent direction: We get

4 3
 
v = −u = − , ,
5 5
with rate of descent
1
 
− ∇f 2, − = −5
2

Samy T. Several variables Multivariate calculus 77 / 145

Example of steepest descent (4)

Samy T. Several variables Multivariate calculus 78 / 145

Example of steepest descent (5)

Gradient at point (3, 1): Recall that

∇f (x , y ) = ⟨2x , 6y ⟩

Thus
∇f (3, 1) = ⟨6, 6⟩

Direction of 0 change: Any direction ⊥ ⟨6, 6⟩

,→ Unit vectors given by
1 1
u = √ ⟨1, −1⟩ , √ ⟨−1, 1⟩
2 2

Samy T. Several variables Multivariate calculus 79 / 145

Example of steepest descent (6)

Samy T. Several variables Multivariate calculus 80 / 145

Gradient and level curves

Theorem 8.
Let
f differentiable function at (x , y )
Hypothesis: ∇f (a, b) ̸= 0
Then:

The line tangent to the level curve of f at (a, b)

is
orthogonal to ∇f (a, b)

Samy T. Several variables Multivariate calculus 81 / 145

Hyperboloid example (1)

Function: q
z = f (x , y ) = 1 + 2x 2 + y 2

Questions:
1 Verify that the gradient at (1, 1) is orthogonal to the
corresponding level curve at that point.
2 Find an equation of the line tangent to the level curve at (1, 1)

Samy T. Several variables Multivariate calculus 82 / 145

Hyperboloid example (2)

Samy T. Several variables Multivariate calculus 83 / 145

Hyperboloid example (3)
Point on surface:
Given by (1, 1, 2) =⇒ On level curve z = 2

Equation for level curve: Ellipse of the form

1 + 2x 2 + y 2 = 4 ⇐⇒ 2x 2 + y 2 = 3

Implicit derivative:
dy Fx 2x
=− =−
dx Fy y
Thus
dy
(1) = −2
dx

Samy T. Several variables Multivariate calculus 84 / 145

Hyperboloid example (4)
Tangent vector: Proportional to

t = ⟨1, −2⟩

Gradient of f :
* +
2x y
∇f (x , y ) = √ 2 2
, √
1 + 2x + y 1 + 2x 2 + y 2

Thus
1
 
∇f (1, 1) = 1,
2
Orthogonality: We have

t · ∇f (1, 1) = 0

Samy T. Several variables Multivariate calculus 85 / 145

Hyperboloid example (5)

Tangent line to level curve: At point (1, 1) we get

fx (1, 1)(x − 1) + fy (1, 1)(y − 1) = 0,

that is
y = −2x + 3

Samy T. Several variables Multivariate calculus 86 / 145

Generalization to 3 variables
Situation:
We have a function w = f (x , y , z)
Each w0 results in a level surface
f (x , y , z) = w0

Gradient on level surface:

Will be ⊥ to level surface

Samy T. Several variables Multivariate calculus 87 / 145

Example of tangent plane (1)

Function:
f (x , y , z) = xyz

Gradient:
∇f (x , y , z) = ⟨yz, xz, xy ⟩
Thus
∇f (1, 2, 3) = ⟨6, 3, 2⟩

Samy T. Several variables Multivariate calculus 88 / 145

Example of tangent plane (2)

Plane tangent to level surface:

⟨6, 3, 2⟩ · ⟨x − 1, y − 2, z − 3⟩ = 0

We get
6x + 3y + 2z = 18

Samy T. Several variables Multivariate calculus 89 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 90 / 145

Linear approximation for functions of 1 variable
Situation: We have
y = f (x )

Tangent vector at a:
t = (1, f ′ (a))

Linear approximation: Near a we have

f (x ) ≃ f (a) + f ′ (a)(x − a)

Samy T. Several variables Multivariate calculus 91 / 145

Tangent plane for F (x , y , z) = 0

Definition 9.
Let F (x , y , z) be such that
F differentiable at P(a, b, c)
∇F ̸= 0
S is the surface F (x , y , z) = 0
Then the tangent plane at (a, b, c) is given by

Fx (a, b, c)(x − a) + Fy (a, b, c)(y − b) + Fz (a, b, c)(z − c) = 0

Samy T. Several variables Multivariate calculus 92 / 145

Interpretation of tangent plane
Tangent plane as collection of tangent vectors: If
S is the surface F (x , y , z) = 0
r is a curve passing through (a, b, c) at time t
Then r′ (t) ∈ tangent plane

Samy T. Several variables Multivariate calculus 93 / 145

Example of tangent plane (1)

Surface: Ellipsoid of the form

x2 y2
F (x , y , z) = + + z2 − 1 = 0
9 25

Questions:
1 Tangent plane at (0, 4, 53 )
2 What tangent planes to S are horizontal?

Samy T. Several variables Multivariate calculus 94 / 145

Example of tangent plane (2)

Gradient: We have
2x 2y
 
∇F (x , y , z) = , , 2z
9 25
Thus
3 8 6
 
∇F (0, 4, ) = 0, ,
5 25 5

Tangent plane:
4y + 15z = 25

Samy T. Several variables Multivariate calculus 95 / 145

Example of tangent plane (3)

Samy T. Several variables Multivariate calculus 96 / 145

Example of tangent plane (4)

Horizontal plane: When the normal vector is of the form

n = (0, 0, c), with c ̸= 0

Horizontal tangent plane: When the normal vector ∇F is of the form

∇F (x , y , z) = (0, 0, c) ⇐⇒ Fx = 0, Fy = 0, Fz ̸= 0

Solutions: Horizontal tangent plane for

(0, 0, 1) and (0, 0, −1)

Samy T. Several variables Multivariate calculus 97 / 145

Tangent plane for z = f (x , y )

Definition 10.
Let f (x , y ) be such that
f differentiable at (a, b)
S is the surface z = f (x , y )
Then the tangent plane to S at (a, b, f (a, b)) is given by

z = fx (a, b) (x − a) + fy (a, b) (y − b) + f (a, b)

Samy T. Several variables Multivariate calculus 98 / 145

Example of tangent plane for z = f (x , y ) (1)

Surface: Paraboloid of the form

z = f (x , y ) = 32 − 3x 2 − 4y 2

Question:
Tangent plane at (2, 1, 16)

Samy T. Several variables Multivariate calculus 99 / 145

Example of tangent plane for z = f (x , y ) (2)

Partial derivatives: We have

fx = 6x , fy = −8y

Thus
fx (2, 1) = −12, fy (2, 1) = −8

Tangent plane:
z = −12x − 8y + 48

Samy T. Several variables Multivariate calculus 100 / 145

Linear approx for functions of 1 variable (Repeat)
Situation: We have
y = f (x )

Tangent vector at a:
t = (1, f ′ (a))

Linear approximation: Near a we have

f (x ) ≃ f (a) + f ′ (a)(x − a)

Samy T. Several variables Multivariate calculus 101 / 145

Linear approximation for functions of 2 variables

Definition 11.
Let f (x , y ) be such that
f differentiable at (a, b)
S is the surface z = f (x , y )
Then the linear approximation to S at (a, b, f (a, b)) is given by

L(x , y ) = fx (a, b) (x − a) + fy (a, b) (y − b) + f (a, b)

Remark: Another popular form of the linear approximation is

∆z ≃ fx dx + fy dy

Samy T. Several variables Multivariate calculus 102 / 145

Example of infinitesimal change (1)

Function:
z = f (x , y ) = x 2 y

Question: Evaluate the percentage of change in z if

x is increased by 1%
y is decreased by 3%

Samy T. Several variables Multivariate calculus 103 / 145

Example of infinitesimal change (2)
Small change in z:

dz ≃ fx dx + fy dy = 2xy dx + x 2 dy

Small percentage change in z:

dz 2xy x2 2 1
= dx + dy = dx + dy
z z z x y

dx dy
If x
= .01 and y
= −.03:

dz
= −.01 = −1%
z

Samy T. Several variables Multivariate calculus 104 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 105 / 145

Max and min for functions of 1 variable

Situation: We have
y = f (x )

Critical point: (c, f (c)) whenever

f ′ (c) = 0

Second derivative test: If (c, f (c)) is critical then

1 If f ′′ (c) > 0, there is a local minimum
2 If f ′′ (c) < 0, there is a local maximum
3 If f ′′ (c) = 0, the test is inconclusive

Samy T. Several variables Multivariate calculus 106 / 145

Critical points for functions of 2 variables

Definition 12.
Let
f function of 2 variables
(a, b) interior point in the domain of f
Then (a, b) is a critical point of f if

fx (a, b) = 0, and fy (a, b) = 0,

or if one of the partial derivatives fx , fy does not exist at (a, b)

Samy T. Several variables Multivariate calculus 107 / 145

Second derivative test

Theorem 13.
For f twice diff. function, define the discriminant of f as

D(x , y ) = fxx (x , y )fyy (x , y ) − (fxy (x , y ))2

Then for a critical point (a, b) the following holds true:

1 If D(a, b) > 0 and fxx (a, b) < 0, we have a local max
2 If D(a, b) > 0 and fxx (a, b) > 0, we have a local min
3 If D(a, b) < 0, we have a saddle point
4 If D(a, b) = 0, the test is inconclusive

Samy T. Several variables Multivariate calculus 108 / 145

Saddle point for an hyperboloid

Samy T. Several variables Multivariate calculus 109 / 145

Hyperboloids in architecture

Samy T. Several variables Multivariate calculus 110 / 145

Hyperboloids in the food industry

Samy T. Several variables Multivariate calculus 111 / 145

Example of critical points analysis (1)

Function:
f (x , y ) = x 2 + 2y 2 − 4x + 4y + 6

Problem:
Use second derivative test to classify the critical points of f

Samy T. Several variables Multivariate calculus 112 / 145

Example of critical points analysis (2)

Partial derivatives:

fx = 2x − 4, fy = 4y + 4

Critical point:
(2, −1)

Critical value of f :
f (2, −1) = 0

Samy T. Several variables Multivariate calculus 113 / 145

Example of critical points analysis (3)

Second derivatives:

fxx = 2, fxy = fyx = 0, fyy = 4

Discriminant:
D(x , y ) = 8 > 0

Second derivative test: We have

D(2, −1) > 0, fxx (2, −1) > 0 =⇒ Local minimum at (2, −1)

Samy T. Several variables Multivariate calculus 114 / 145

Example of critical points analysis (4)

Samy T. Several variables Multivariate calculus 115 / 145

Second example (1)

Function:
f (x , y ) = xy (x − 2)(y + 3)

Problem:
Use second derivative test to classify the critical points of f

Samy T. Several variables Multivariate calculus 116 / 145

Second example (2)

Partial derivatives:

fx = 2y (x − 1)(y + 3), fy = x (x − 2)(2y + 3)

Critical points:
3
(0, 0), (2, 0), (1, − ), (0, −3), (2, −3),
2

Samy T. Several variables Multivariate calculus 117 / 145

Second example (3)
Second derivatives:

fxx = 2y (y + 3), fxy = 2(2y + 3)(x − 1), fyy = 2x (x − 2)

Analysis of critical points:

(x , y ) D(x , y ) fxx Conclusion
(0, 0) −36 0 Saddle point
(2, 0) −36 0 Saddle point
(1, −3/2) 9 −9/2 Local maximum
(0, −3) −36 0 Saddle point
(2, −3) −36 0 Saddle point

Samy T. Several variables Multivariate calculus 118 / 145

Second example (4)

Samy T. Several variables Multivariate calculus 119 / 145

Absolute maximum

Proposition 14.
Let
f continuous function of 2 variables
R closed region of R2
In order to find the maximum of f in R, we proceed as follows:
1 Determine the values of f at all critical points in R.
2 Find the maximum and minimum values of f on the
boundary of R.
3 The greatest function value found in Steps 1 and 2 is the
absolute maximum value of f on R.

Samy T. Several variables Multivariate calculus 120 / 145

Example of global maximum (1)

Function:
z = f (x , y ) = x 2 + y 2 − 2x − 4y

Region:

R = {(x , y ); (x , y ) within triangle with vertices (0, 0), (0, 4), (4, 0)}

Question:
Find global maximum of f on region R

Samy T. Several variables Multivariate calculus 121 / 145

Example of global maximum (2)

Partial derivatives:

fx = 2x − 2, fy = 2y − 4

Critical point:
(1, 2), with f (1, 2) = −5

Samy T. Several variables Multivariate calculus 122 / 145

Example of global maximum (3)

Boundary 1: On y = 0, 0 ≤ x ≤ 4 we have

f (x , y ) = x 2 − 2x ≡ g(x ), g ′ (x ) = 2(x − 1)

Points of interest on boundary 1: We get

(0, 0), (1, 0), (0, 4)

and
f (0, 0) = 0, f (1, 0) = −1, f (4, 0) = 8

Samy T. Several variables Multivariate calculus 123 / 145

Example of global maximum (4)

Boundary 2: On y = 4 − x , 0 ≤ x ≤ 4 we have

f (x , y ) = 2x 2 − 6x ≡ h(x ), h′ (x ) = 4x − 6

Points of interest on boundary 2: We get

3 5
 
(0, 4), , , (4, 0)
2 2
and
3 5 9
 
f (0, 4) = 0, f , =− , f (4, 0) = 8
2 2 2

Samy T. Several variables Multivariate calculus 124 / 145

Example of global maximum (5)

Boundary 3: On x = 0, 0 ≤ y ≤ 4 we have

f (x , y ) = y 2 − 4y ≡ k(y ), k ′ (y ) = 2(y − 2)

Points of interest on boundary 3: We get

(0, 0), (0, 2), (0, 4)

and
f (0, 0) = 0, f (0, 2) = −4, f (0, 4) = 0

Samy T. Several variables Multivariate calculus 125 / 145

Example of global maximum (6)
Summary of points of interest:

f (0, 0) = 0, f (1, 0) = −1, f (4, 0) = 8

3 5 9
 
f (0, 4) = 0, f , =− , f (4, 0) = 8
2 2 2
f (0, 0) = 0, f (0, 2) = −4, f (0, 4) = 0, f (1, 2) = −5

Absolute minimum: at (1, 2) and

f (1, 2) = −5

Absolute maximum: at (4, 0) and

f (4, 0) = 8

Samy T. Several variables Multivariate calculus 126 / 145

Outline
1 Graphs and level curves
2 Limits and continuity
3 Partial derivatives
4 The chain rule
5 Directional derivatives and the gradient
6 Tangent plane and linear approximation
7 Maximum and minimum problems
8 Lagrange multipliers

Samy T. Several variables Multivariate calculus 127 / 145

Global aim

Objective function:
f = f (x , y )

Constraint: We are moving on a curve of the form

g(x , y ) = 0

Optimization problem: Find

max f (x , y ), subject to g(x , y ) = 0

Samy T. Several variables Multivariate calculus 128 / 145

Optimization problem: illustration

Samy T. Several variables Multivariate calculus 129 / 145

Lagrange multipliers intuition (1)

Samy T. Several variables Multivariate calculus 130 / 145

Lagrange multipliers intuition (2)
Some observations from the picture:
1 P(a, b) on the level curve of f
=⇒ Tangent to level curve ⊥ ∇f (a, b)
2 P(a, b) gives a maximum of f on curve C
=⇒ Tangent to level curve ∥ Tangent to constraint curve
3 Constraint is g(x , y ) = 0
=⇒ Tangent to constraint curve ⊥ ∇g(a, b)

Conclusion (Lagrange’s idea):

At the maximum under constraint we have

∇f (a, b) ∥ ∇g(a, b)

Samy T. Several variables Multivariate calculus 131 / 145

Lagrange multipliers procedure

Optimization problem: Find

max f (x , y ), subject to g(x , y ) = 0

Recipe:
1 Find the values of x , y and λ such that

∇f (x , y ) = λ∇g(x , y ), and g(x , y ) = 0

2 Select the largest and smallest corresponding function values.

,→ We get absolute max and min values of f s.t constraint.

Samy T. Several variables Multivariate calculus 132 / 145

Example of Lagrange multipliers (1)

Optimization problem: Find

max f (x , y ), with f (x , y ) = x 2 + y 2 + 2,

subject to the constraint

g(x , y ) = x 2 + xy + y 2 − 4 = 0

Samy T. Several variables Multivariate calculus 133 / 145

Example of Lagrange multipliers (2)

Computing the gradients: We get

∇f (x , y ) = ⟨2x , 2y ⟩ , ∇g(x , y ) = ⟨2x + y , x + 2y ⟩

Lagrange constraint 1:

fx = λgx ⇐⇒ 2x = λ (2x + y ) (1)

Lagrange constraint 2:

fy = λgy ⇐⇒ 2y = λ (x + 2y ) (2)

Samy T. Several variables Multivariate calculus 134 / 145

Example of Lagrange multipliers (3)

System for x , y : Gathering (1) and (2), we get

2(λ − 1)x + λy = 0, λx + 2(λ − 1)y = 0

This has solution (0, 0) unless

2
λ = 2, or λ =
3

Samy T. Several variables Multivariate calculus 135 / 145

Example of Lagrange multipliers (4)

Case λ = 2: We get x = −y . The constraint

x 2 + xy + y 2 − 4 = 0

becomes
x2 − 4 = 0
Solutions:
x = 2, and x = −2

Corresponding values of f : We have

f (2, −2) = f (−2, 2) = 10

Samy T. Several variables Multivariate calculus 136 / 145

Example of Lagrange multipliers (5)
Case λ = 23 : We get x = y . The constraint

x 2 + xy + y 2 − 4 = 0

becomes
3x 2 − 4 = 0
Solutions:
2 2
x=√ , and x = − √
3 3
Corresponding values of f : We have
! !
2 2 2 2 14
f √ ,√ =f −√ , −√ =
3 3 3 3 3

Samy T. Several variables Multivariate calculus 137 / 145

Example of Lagrange multipliers (6)

Absolute maximum:
For function f on the curve C defined by g = 0,

Maximum = 10, obtained for (2, −2) , (−2, 2)

Absolute minimum:
For function f on the curve C defined by g = 0,
! !
14 2 2 2 2
Minimum = , obtained for √ ,√ , −√ , −√
3 3 3 3 3

Samy T. Several variables Multivariate calculus 138 / 145

Example of Lagrange multipliers (7)

Samy T. Several variables Multivariate calculus 139 / 145

Optimization in dimension 3 (1)

Problem: Find the point on the sphere

x 2 + y 2 + z 2 = 1,

closest to the point

(1, 2, 3)

Samy T. Several variables Multivariate calculus 140 / 145

Optimization in dimension 3 (2)

Related minimization problem:

Find

min f (x , y ), with f (x , y ) = (x − 1)2 + (y − 2)2 + (z − 3)2 ,

subject to the constraint

g(x , y ) = x 2 + y 2 + z 2 − 1 = 0

Samy T. Several variables Multivariate calculus 141 / 145

Optimization in dimension 3 (3)

Computing the gradients: We get

∇f (x , y ) = ⟨2(x − 1), 2(y − 2), 2(z − 3)⟩

∇g(x , y ) = ⟨2x , 2y , 2z⟩

Lagrange constraint: We have

∇f (x , y ) = λ ∇g(x , y )
⇐⇒
(λ − 1)x = −1, (λ − 1)y = −2, (λ − 1)z = −3

Samy T. Several variables Multivariate calculus 142 / 145

Optimization in dimension 3 (4)

Solutions of Lagrange constraints:

The Lagrange system has unique solution whenever λ ̸= 1. We get
1 2 1
x =− , y =− = 2x , z =− = 3x
λ−1 λ−1 λ−1

Reporting in constraint g: We have

y = 2x , z = 3x , g(x , y ) = 0,

Thus we get
14x 2 = 1

Samy T. Several variables Multivariate calculus 143 / 145

Optimization in dimension 3 (5)

Solutions:
1 1
x=√ , and x = − √
14 14

Corresponding values of f : We have

!
1 2 3
f √ ,√ ,√ ≃ 7.51
14 14 14
!
1 2 3
f −√ , −√ , −√ ≃ 22.48
14 14 14

Samy T. Several variables Multivariate calculus 144 / 145

Optimization in dimension 3 (6)

Absolute maximum:
Maximal distance from (1, 2, 3) to a point on the sphere is
!
1 2 3
Maximum = 4.74, obtained for −√ , −√ , −√
14 14 14

Absolute minimum:
Minimal distance from (1, 2, 3) to a point on the sphere is
√
!
1 2 3
Minimum = 2.74 = 7.51, obtained for √ ,√ ,√
14 14 14

Samy T. Several variables Multivariate calculus 145 / 145