Functions of Several Variables

1 Planes & Surfaces

One-variable functions y = f (x): curves in R2
Two-variable functions z = f (x, y): surfaces in R3

General Equations of Planes in R3

A plane in space is determined by a point P0 (x0 , y0 , z0 ) in the plane and a vector n = ⟨a, b, c⟩
that is orthogonal to the plane (specify the “direction” of the plane). This orthogonal vector
is called a normal vector. Let P (x, y, z) be an arbitrary point in the plane, and let r0 and r
ÐÐ→
be the position vectors of P0 and P , then P0 P = r − r0 . Since the normal vector is orthogonal

1
to every vector in the given plane, we have n ⋅ (r − r0 ) = 0, that is

⟨a, b, c⟩ ⋅ ⟨x − x0 , y − y0 , z − z0 ⟩ = 0 or
a(x − x0 ) + b(y − y0 ) + c(z − z0 ) = 0 or
ax + by + cz + d = 0 with d = ax0 + by0 + cz0

EXAMPLE 1 Find an equation of the plane through the point (2, 4, −1) with normal vector
n = ⟨2, 3, 4⟩. Find the intercepts and sketch the plane.
Solution. Put a = 2, b = 3, c = 4 and x0 = 2, y0 = 4, z0 = −1 into the equation of plane, we have
2(x − 2) + 3(y − 4) + 4(z + 1) = 0 or 2x + 3y + 4z = 12. To find the x-intercept we set y = z = 0
in this equation and obtain x = 6. Similarly, the y-intercept is 4 and the z-intercept is 3. This
enables us to sketch the portion of the plane that lies in the first octant.

EXAMPLE 2 Find an equation of the plane that passes through the points P (1, 3, 2),
Q(3, −1, 6) and R(5, 2, 0).
Ð→ Ð→
Solution. Since both P Q and P R lie in the plane, their cross product is orthogonal to the plane
RR i j k RRR
Ð→ Ð→ RRRR RR
and can be taken as the normal vector. Then n = P Q × P R = RRR 2 −4 4 RRRR = 12i + 20j + 14k.
RRR R
RR 4 −1 −2 RRRR
A plane equation is 12(x − 1) + 20(y − 3) + 14(z − 2) = 0 or 6x + 10y + 7z = 50

Parallel & Orthogonal Planes

Two distinct plane with normal vectors n1 and n2

ˆ The planes are parallel if n1 and n2 are parallel

ˆ The planes are orthogonal if n1 ⋅ n2 = 0

2
EXAMPLE 3 Find an equation of the plane that passes through the points (−2, 4, 1) and is
parallel to the plane 3x − 2y + z = 4.
Solution. A normal vector is n = ⟨3, −2, 1⟩. Then a equation of the plane is 3(x + 2) − 2(y −
4) + (z − 1) = 0 or 3x − 2y + z = −13.

EXAMPLE 4 Determine the line of intersection L of two planes x+y +z = 1 and x−2y +3z = 1
Solution. The normal of the two planes are n1 = ⟨1, 1, 1⟩, n2 = ⟨1, −2, 3⟩, which are not parallel,
i.e., the two planes do intersect.
To determine the intersection line, its direction vector v and a point P0 on the line are
required. Because L is in both planes, v is orthogonal to both of the normal vectors. Thus a
vector v parallel to L is given by the cross product
RRR i j k RRR
RR RR
v = n1 × n2 = RRRR 1 1 1 RRRR = 5i − 2j − 3k.
RRR R
RR 1 −2 3 RRRR
For a point P0 on L, we can find it where the line intersects the x − y plane by setting z = 0
in the equations of both planes. This gives x + y = 1 and x − 2y = 1, whose solution is x = 1 and
y = 0. So P0 (1, 0, 0).
Eventually, L is given as r(t) = r0 + vt = ⟨1, 0, 0⟩ + t⟨5, −2, −3⟩

3
General Cylindrical Surfaces
DEFINITION Given a curve C in a plane S and a line L not in S, a cylinder is the surface
consisting of all the lines parallel to L that pass through C.

A common situation: the curve C is in a coordinate plane and the line L is in the direction
of the 3rd coordinate axis, i.e., direction that orthogonal to the coordinate plane where C is.
Then we can easily write down the cylindrical surface generated by C and L.

EXAMPLE 5 The curve z = x2 is a parabola in the x − z plane, and the line L is parallel
to the y-axis. The equation for the cylindrical surface generated by the curve and the line is
z = x2 . The variable corresponding to the coordinate axis parallel to L is missing, because for
arbitrary point P (x, y, z) on this surface, constrain is only set to its x and z coordinates.

EXAMPLE 6 Graphing cylinders

(a) x2 + y 2 = 1 a circular cylinder parallel to z-axis
(b) z − cos x = 0 a cosine cylinder parallel to y-axis
(c) z − tan y = 0 a tangent cylinder parallel to x-axis

4
Quadric Surfaces
The General equation (quadratic equation)

Ax2 + By 2 + Cz 2 + Dxy + Exz + F yz + Gx + Hy + Iz + J = 0

where A, B, C, D, E, F, G, H, I, J are not zero at the same time. Making hand sketches can be
challenging, we usually need the help of

ˆ Intercepts: where the surface intersects the coordinate axes

ˆ Traces: the curves of intersection of the surface with planes parallel to the coordinate
planes. Traces in the coordinate planes are called the xy-trace, yz-trace, and xz-trace.
y2 z2
EXAMPLE 7 Use traces to sketch the quadric surface with equation x2 + + = 1.
9 4
y2
Solution. Substituting z = 0, the xy-trace is x2 + = 1 which is an ellipse. In general, the
9
y 2 k 2
horizontal trace in the plane z = k is x2 + = 1 − , which is an ellipse, if k 2 < 4. (If k 2 = 4, the
9 4
trace consists of a single point (0, 0, ±2), the intercepts at z-axis. If k 2 > 4, the trace is empty.)
y2 z2
Similarly, vertical traces parallel to the y − z plane and x − z plane are also ellipses: + =
9 4
z2 k2
1 − k 2 for x = k, −1 < k < 1 and x2 + = 1− for y = k, −3 < k < 3. The intercepts at x- and y-
4 9
axes are (±1, 0, 0) and (0, ±3, 0)
The surface is an ellipsoid.

y2 z2
Graph of an ellipsoid x2 + + =1
9 4

EXAMPLE 8 Use traces to sketch the quadric surface with equation z = 4x2 + y 2 .
Solution. If we put x = 0, we get z = y 2 , so the yz-trace is a parabola. If we put x = k (a
constant), we get z = y 2 +4k 2 . This means that if we slice the graph with any plane parallel to the
yz-plane, we obtain a parabola that opens upward. Similarly, if y = k, the trace is z = 4x2 + k 2 ,
which is again a parabola that opens upward. If we put z = k, we get the horizontal traces
4x2 + y 2 = k, which are a family of ellipses for k > 0. Because of the elliptical and parabolic
traces, the quadric surface is called an elliptic paraboloid.
x2 z2
EXAMPLE 9 Sketch the quadric surface with equation + y2 − = 1.
4 4
x2 k2
Solution. The trace in any horizontal plane z = k is the ellipse + y 2 = 1 + . But the traces
4 4
x2 z 2 2
z2
in the xz- and yz- planes are the hyperbolas − = 1, y = 0 and y − = 1, x = 0. This
4 4 4
surface is called a hyperboloid of one sheet.

5
 Graph of an elliptic paraboloid z = 4x2 + y 2

x2 z2
Graph of a hyperboloid of one sheet + y2 − =1
4 4

Graphs of Quadric Surfaces

6
7
2 Graphs & Level Curves
Functions of Two Variables
DEFINITION A function of two variables z = f (x, y) is a rule that assigns to each ordered
pair of real numbers (x, y) a unique real number z.

ˆ Independent variables: x, y

ˆ dependent variables: z

ˆ domain of the function D: the set of (x, y), i.e. a subset of R2

ˆ range of the function: the set of z, i.e. a subset of R

If a function f is given by an explicit formula and no domain is specified, then the domain
of f is understood to be the set of all pairs (x, y) for which the given expression defined a real
number.
√
EXAMPLE 1 Find the domain and range of z = f (x, y) = 9 − x2 − y 2
Solution. Domain is the set of (x, y) such that z is real: 9−x2 −y 2 ⩾ 0 ⇒ D = {(x, y) ∣ x2 +y 2 ⩽ 9}
Range is the set of z: {z ∣ 0 ⩽ z ⩽ 3} = [0, 3]

8
Graphs
DEFINITION If f is a function of two variables with domain D, then the graph of f is the
set of all points (x, y, z) in R3 such that z = f (x, y) and (x, y) is in D.

EXAMPLE 2 Graphing z = f (x, y) = 6 − 3x − 2y

Solution. A plane (linear function) with x-intercept x = 2 (by setting y = z = 0), y-intercept
y = 3 (by setting x = z = 0) and z-intercept z = 6 (by setting x = y = 0)

√
EXAMPLE 3 Graphing z = f (x, y) = 9 − x2 − y 2
Solution. A hemisphere generated by rotating the half circle x2 + z 2 = 9 for z ⩾ 0 on the x − z
plane about the z- axis.
EXAMPLE 4 Graphing
(a) z = f (x, y) = x2 + y 2 a paraboloid by rotating the parabola z = x2 on the x − z plane
about the z-axis

9
 √
(b) z = f (x, y) = 1 + x2 + y 2 a hyperboloid by rotating the upper branch of the hyperbola
z 2 = 1 + y 2 on the y − z plane about the z-axis

Rotation can only be used to graph functions whose horizontal traces are circles.
ˆ Graphs of z = f (x, y) should pass the vertical line test: every line parallel to z-axis
intersects the graph at most once

Level Curves & Contour Maps

An alternative method for visualizing functions, borrowed from mapmakers, is a contour map
on which points of constant elevation are joined to form contour curves, or level curves.

DEFINITION The level curves of a function f of two variables are the curves with equations
f (x, y) = k, where k is a constant (in the range of f ).

10
 A level curve f (x, y) = k is the set of all points in the domain of f at which f takes on a
given value k. In other words, it is a curve in the x − y plane that shows where the graph of f
has height k (above or below the x − y plane). A collection of level curves is called a contour
map. From a contour map, one can visualize a function z = f (x, y) by lifting up (k > 0) or
putting down (k < 0) the level curves with k.
Contour maps are most descriptive when the level curves f (x, y) = k are drawn for equally
spaced values of k, and we assume that this is the case unless indicated otherwise.

EXAMPLE 5 Find the level curves of (a) z = y − x2 − 3 (b) z = 4x2 + y 2 + 1

Solution. (a) Level curves are parabola y − x2 − 3 = k or y = x2 + 3 + k
(b) Level curves are ellipses 4x2 + y 2 + 1 = k or 4x2 + y 2 = k − 1 for k > 1

Functions of Three or More Variables

DEFINITION A function of three variables w = f (x, y, z) is a rule that assigns to each ordered
triple (x, y, z) in a domain D ∈ R3 a unique real number w in its range.

EXAMPLE 6 Find the domain of w = f (x, y, z) = ln(z − y) + xy sin z.

11
Solution. The expression for f (x, y, z) is defined as long as z − y > 0, therefore, the domain is
D = {(x, y, z) ∈ R3 ∣ z > y}

Graphs of functions of three variables lie in a four-dimensional space, which can be examined
by its level surfaces, which are the surfaces with equations f (x, y, z) = k with k a constant in
the range of f .

EXAMPLE 7 Find the level surfaces of the functions (a) f (x, y, z) = x2 +y 2 +z 2 (b) f (x, y, z) =
x2 − y − z 2 √
Solution. (a) Level surfaces are a family of concentric spheres with radius k: x2 + y 2 + z 2 = k
(b) Level surfaces are a family of hyperbolic paraboloids x2 − y − z 2 = k or y = x2 − z 2 − k

DEFINITION A function of n variables y = f (x1 , x2 , . . . , xn ) is a rule that assigns to an

n-tuple (x1 , x2 , . . . , xn ) in a domain D ∈ Rn a unique real number y in its range.

3 Limits & Continuity

Limits of a Function of Two Variables
DEFINITION Let f be a function of two variables whose domain D includes points arbitrarily
close to (a, b). Then we say the limit of f as (x, y) approaches (a, b) is L and we write

lim f (x, y) = L
(x,y)→(a,b)

if for
√ every number ε > 0 there is a corresponding number δ > 0 such that if (x, y) ∈ D and
0 < (x − a)2 + (y − b)2 < δ then ∣f (x, y) − L∣ < ε.

12
Illustration 1 If any small interval (L − ε, L + ε) is given around L, then we can find a disk
Dδ with center (a, b) and radius δ > 0 such that f maps all the points in Dδ [except possibly
(a, b)] into the interval (L − ε, L + ε).
Illustration 2 If ε > 0 is given, we can find δ > 0 such that if (x, y) is restricted to lie in the
disk Dδ and (x, y) ≠ (a, b), then the corresponding part of S lies between the horizontal planes
z = L − ε and z = L + ε.
Remark For functions of a single variable, if we let x → a, there are only two possible directions:
x → a− and x → a+ . However, for functions of two variables, (x, y) can approach (a, b) from
infinite numbers of directions and in infinite number of manners in the domain of the function.
Therefore, if lim(x,y)→(a,b) f (x, y) exists, f (x, y) → L no matter how (x, y) → (a, b)

THEOREM Limits of constant and linear functions.

Let a, b and c be real numbers

1. Constant function f (x, y) = c ∶ lim f (x, y) = lim c=c

(x,y)→(a,b) (x,y)→(a,b)

2. Linear function f (x, y) = x ∶ lim f (x, y) = lim x=a

(x,y)→(a,b) (x,y)→(a,b)

3. Linear function f (x, y) = y ∶ lim f (x, y) = lim y=b

(x,y)→(a,b) (x,y)→(a,b)

THEOREM Laws of Limits for functions of two variables.

Let L and M be real numbers and suppose that lim(x,y)→(a,b) f (x, y) = L, lim(x,y)→(a,b) g(x, y) =
M . Assume c is a constant, and m and n are integers.

1. Sum lim [f (x, y) + g(x, y)] = lim f (x, y) + lim g(x, y) = L + M

(x,y)→(a,b) (x,y)→(a,b) (x,y)→(a,b)

2. Difference lim [f (x, y) − g(x, y)] = lim f (x, y) − lim g(x, y) = L − M

(x,y)→(a,b) (x,y)→(a,b) (x,y)→(a,b)

3. Constant multiple lim cf (x, y) = c [ lim f (x, y)] = cL

(x,y)→(a,b) (x,y)→(a,b)

4. Product lim f (x, y)g(x, y) = [ lim f (x, y)] [ lim g(x, y)] = LM
(x,y)→(a,b) (x,y)→(a,b) (x,y)→(a,b)

f (x, y) lim(x,y)→(a,b) f (x, y) L

5. Quotient lim = = , provided M ≠ 0
(x,y)→(a,b) g(x, y) lim(x,y)→(a,b) g(x, y) M
n
n
6. Power lim [f (x, y)] = [ lim f (x, y)] = Ln
(x,y)→(a,b) (x,y)→(a,b)

7. m/n power. If m and n have no common factors and n ≠ 0

m/n
m/n
lim [f (x, y)] =[ lim f (x, y)] = Lm/n , provided L > 0 if n is even
(x,y)→(a,b) (x,y)→(a,b)

√
EXAMPLE 1 Find lim(x,y)→(2,8) (3x2 y + xy)

13
Solution.
√ √
lim (3x2 y + xy) = lim 3x2 y + lim xy
(x,y)→(2,8) (x,y)→(2,8) (x,y)→(2,8)
2 1/2 1/2
= 3[ lim x] [ lim y] + [ lim x] [ lim y]
(x,y)→(2,8) (x,y)→(2,8) (x,y)→(2,8) (x,y)→(2,8)

= 3 × (2)2 × 8 + (2)1/2 × (8)1/2 = 100

Limits at boundary points

DEFINITION Interior and boundary points
Let R be a region in R2 . An interior point P of R lies entirely within R, which means it
is possible to find a disk centered at P that contains only points of R. A boundary point Q
lies on the edge of R in the sense that every disk centered at Q contains at least one point in
R and at least one point not in R.

DEFINITION Open and closed sets

A region is open is it consists entirely of interior points. A region is closed if it contains
all its interior and boundary points.

EXAMPLE 2 R defined by x2 + y 2 < 9. Boundary points of R lie on the circle x2 + y 2 = 9.

Interior points lie inside the circle.
EXAMPLE 3 {(x, y) ∣ x2 + y 2 < 9} is an open set. {(x, y) ∣ ∣x∣ ⩽ 1, ∣y∣ ⩽ 1} is a closed set.

xy − 4y 2
EXAMPLE 4 Find lim(x,y)→(4,1) f (x, y), where f (x, y) = √ √
x−2 y
Solution. The domain of f (x, y) is D = {(x, y) ∣ x > 0, y > 0, x ≠ 4y}
√ √ √ √
(xy − 4y 2 ) ( x + 2 y) y (x − 4y) ( x + 2 y)
lim f (x, y) = lim √ √ √ lim
√ = (x,y)→(4,1)
(x,y)→(4,1) (x,y)→(4,1) ( x − 2 y) ( x + 2 y) x − 4y
x≠4y x≠4y
√ √
= lim y ( x + 2 y) = 4
(x,y)→(4,1)

14
Non-existence of Limits: Two-Path Test
If f (x, y) approaches two different values as (x, y) approaches (a, b) along two different paths
in the domain of f , then lim(x,y)→(a,b) f (x, y) does not exist.

(x + y)2
EXAMPLE 5 Does lim(x,y)→(0,0) exist?
x2 + y 2
Solution. Let (x, y) approaches (a, b) along y = mx, where m is a constant.
2 2 2
(x + y) (x + mx) x2 (1 + m) (1 + m)2
lim = lim = lim =
(x,y)→(0,0) x2 + y 2 x→0 x2 + (mx)2 x→0 x2 (1 + m2 ) 1 + m2
y=mx

m-dependent answer, thus the limit does NOT exist!

Continuity of Functions of Two Variables

DEFINITION The function f (x, y) is continuous at the point (a, b) provided

1. f (x, y) is defined at (a, b)

2. lim(x,y)→(a,b) f (x, y) exists

3. lim(x,y)→(a,b) f (x, y) = f (a, b)

EXAMPLE 6 Check the continuity of f (x, y) at (0, 0)

⎧ xy 2
⎪ x2 +y4 ,
⎪ if (x, y) ≠ (0, 0)
f (x, y) = ⎨
⎩0, if (x, y) = (0, 0)
⎪
⎪

Solution. Let (x, y) approaches (a, b) along y = mx and (x, y) ≠ (0, 0)

2
x (mx) m2 x
lim f (x, y) = lim 2 = lim =0
(x,y)→(0,0) x→0 x + (mx)4 x→0 1 + m4 x2
y=mx

Let (x, y) approaches (a, b) along x = my 2 and (x, y) ≠ (0, 0)

(my 2 ) y 2 m
lim f (x, y) = lim 2 2 4
= 2
(x,y)→(0,0) y→0 (my ) + y m +1
x=my 2

lim(x,y)→(0,0) f (x, y) does NOT exist. Therefore, the function is NOT continuous at (0, 0). The
graph of f (x, y) is shown below.

15
Continuity of Composite Functions
THEOREM If u = g(x, y) is continuous at (a, b) and z = f (u) is continuous at g(a, b), then
the composite function z = f (g(x, y)) is continuous at (a, b).

EXAMPLE 7 Where is the function h(x, y) = ln(x2 + y 2 + 4) continuous?

Solution. Let u = g(x, y) = x2 + y 2 + 4, f (u) = ln u. Since u = g(x, y) = x2 + y 2 + 4 > 0 is
continuous for all (x, y), and f (u) is continuous for u > 0, h(x, y) = f (g(x, y)) is continuous for
all (x, y) ∈ R2

EXAMPLE 8 Where is the function h(x, y) = ex/y continuous?

x x
Solution. Let u = g(x, y) = , f (u) = eu . Since u = g(x, y) = is continuous for all (x, y)
y y
provided y ≠ 0, and f (u) is continuous for any u ∈ R, h(x, y) = f (g(x, y)) is continuous on
{(x, y) ∣ y ≠ 0}

Functions of Three and More Variables

DEFINITION Let f (x) be defined on a subset D = {x = ⟨x1 , x2 , . . . , xn ⟩} of Rn . f has the
limit L as x → a = ⟨a1 , a2 , . . . , an ⟩, written as limx→a f (x) = L, if for every number ε > 0 there
is a corresponding number δ > 0 such that if x ∈ D and 0 < ∣x − a∣ < δ then ∣f (x) − L∣ < ε.

DEFINITION f (x) is continuous at x = a, if (1) f (x) is defined at a; (2) limx→a f (x) exists;
(3) limx→a f (x) = f (a)
x2 sin y π
EXAMPLE 9 Find the limit of f (x, y, z) = 2 at (2, , 0)
z +4 2
Solution.
x2 sin y 22 sin π2
lim = 2 =1
(x,y,z)→(2, π2 ,0) z 2 + 4 0 +4
√
EXAMPLE 10 Where is the function h(x, y, z) = x2 √ + y 2 + z 2 − 1 continuous?
Solution. Let u = g(x, y, z) = x2 + y 2 + z 2 − 1, f (u) = u. Since u = g(x, y, z) is continuous
and non-negative on {(x, y, z) ∣ x2 + y 2 + z 2 ⩾ 1}, and f (u) is continuous for u ⩾ 0, h(x, y, z) =
f (g(x, y, z)) is continuous on or outside the unit sphere centered at the origin.

4 Partial Derivatives
Derivatives of Two-Variable Functions
If f is a function of two variables x and y, suppose we let only x vary while keeping y fixed,
say y = b, where b is a constant. Then we are really considering a function of a single variable
x, namely, g(x) = f (x, b). If g has a derivative at a, then we call it the partial derivative of
f with respect to x at (a, b), and denote it by fx (a, b). Thus

fx (a, b) = g ′ (a), where g(x) = f (x, b)

By the definition of a derivative, we have

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

16
Similarly, the partial derivative of f with respect to y at (a, b), denoted by fy (a, b), is
obtained by keeping x fixed (x = a) and finding the ordinary derivative at b of the function
G(y) = f (a, y):
f (a, b + h) − f (a, b)
fy (a, b) = lim
h→0 h
The geometrical interpretation of the partial derivatives are illustrated below.

If we replace (a, b) by variable point (x, y), then fx and fy are functions of x and y.
DEFINITION If f is a function of two variables, its partial derivatives are the functions
fx and fy defined by
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
NOTATIONS If z = f (x, y), we write
∂f ∂ ∂z
fx (x, y) = fx = = f (x, y) =
∂x ∂x ∂x
∂f ∂ ∂z
fy (x, y) = fy = = f (x, y) =
∂y ∂y ∂y
RULES for finding partial derivatives of z = f (x, y)

17
 ˆ To find fx , regard y as a constant and differentiate f (x, y) with respect to x. That is
d
taking g(x), where g(x) = f (x, y) by keeping y fixed.
dx
ˆ To find fy , regard x as a constant and differentiate f (x, y) with respect to y. That is
d
taking G(y), where G(y) = f (x, y) by keeping x fixed.
dy
REMARK All the rules and results for ordinary derivative can be used to compute partial
derivatives.

EXAMPLE 1 Calculate the partial derivatives of f (x, y) = x2 − y 2 + 4 at (2, −4)

Solution. fx (x, y) = 2x, fx (2, −4) = 2 × 2 = 4
fy (x, y) = −2y, fy (2, −4) = −2 × (−4) = 8

EXAMPLE 2 Calculate the partial derivatives of (a) f (x, y) = sin(xy) and (b) f (x, y) = x2 exy
∂f ∂ ∂
Solution. (a) = [sin(xy)] = cos(xy) ⋅ (xy) = y cos(xy) ⇐ applying chain rule
∂x ∂x ∂x
∂f ∂ ∂
= [sin(xy)] = cos(xy) ⋅ (xy) = x cos(xy)
∂y ∂y ∂y
∂f ∂ ∂ ∂
(b) = [x2 exy ] = [ (x2 )] exy + x2 [ (exy )] ⇐ applying rule of product
∂x ∂x ∂x ∂x
= 2xexy + x2 yexy = x(2 + xy)exy ⇐ applying chain rule
∂f ∂ 2 xy ∂
= [x e ] = x2 (exy ) = x2 ⋅ xexy = x3 exy ⇐ applying rule of constant
∂y ∂y ∂y
multiple & chain rule

Second Partial Derivatives

Since fx and fy are also functions of x and y, we can consider their partial derivatives, which
are called the second partial derivatives of f . We use the following notations

∂ ∂f ∂ 2f ∂ 2z
( )= = = (fx )x = fxx
∂x ∂x ∂x2 ∂x2
∂ ∂f ∂ 2f ∂ 2z
( )= = = (fx )y = fxy
∂y ∂x ∂y∂x ∂y∂x
∂ ∂f ∂ 2f ∂ 2z
( )= = = (fy )x = fyx
∂x ∂y ∂x∂y ∂x∂y
∂ ∂f ∂ 2f ∂ 2z
( )= = = (fy )y = fyy
∂y ∂y ∂y 2 ∂y 2

EXAMPLE 3 Calculate the second derivatives of f (x, y) = 3x4 y − 2xy + 5xy 3

Solution. fx = 12x3 y − 2y + 5y 3 fy = 3x4 − 2x + 15xy 2
fxx = 36x y 2 fyx = 12x3 − 2 + 15y 2
fxy = 12x3 − 2 + 15y 2 fyy = 30xy
where fxy = fyx . Is this a general result?

THEOREM Equality of mixed partial derivatives

Assume f is defined on an open set D on R2 , and fxy and fyx are continuous throughout
D, then fxy = fyx at all points of D.

18
Functions of Three or More Variables
In general, if y = f (x1 , x2 , . . . , xn ) is a function of n variables, its partial derivative with respect
to the i-th variable xi is defined as
∂f f (x1 , x2 , . . . , xi−1 , xi + h, xi+1 , . . . , xn ) − f (x1 , x2 , . . . , xi−1 , xi , xi+1 , . . . , xn )
fxi = = lim
∂xi h→0 h
If the limit exists.

EXAMPLE 4 Calculate the partial derivatives of f (x, y, z) = e−xy cos z

Solution. fx = −ye−xy cos z keep y and z fixed
xy
fy = −xe cos z keep x and z fixed
xy
fz = −e sin z keep x and y fixed

EXAMPLE 5 The idea gas law gives the relationship of gas pressure P , volume V , and
temperature T as P V = kT , where k > 0 is a constant. Discuss the behaviour of the pressure.
T
Solution. We can rewrite P as a function of V and T : P = P (V, T ) = k , providing V > 0
V
and T > 0.
∂P T
= −k 2 < 0, P is a decreasing function of volume V .
∂V V
∂P 1
= k > 0, P is an increasing function of temperature T .
∂T V

Differentiability
f (a + ∆x) − f (a)
One-variable function y = f (x) is differentiable at x = a, if the limit lim∆x→0 =
∆x
f (a + ∆x) − f (a)
f ′ (a) exits. We can define an error function ε(∆x) = − f ′ (a). Then, if f (x)
∆x
is differentiable at x = a, we have lim∆x→0 ε = 0. The change in y as x varies from a to a + ∆x
is expressed as ∆y = f (a + ∆x) − f (a) = f ′ (a)∆x + ε∆x with lim∆x→0 ε = 0. This implies that
to be truly differentiable a function must be locally linear: when x changes around a,
i.e., x = a + ∆x with ∆x → 0, f (x) = f (a + ∆x) = f (a) + f ′ (a)∆x + ε∆x ≃ f (a) + f ′ (a)∆x. In
other words, if a one-variable function is differentiable at a, it can be approximated by a piece
of straight line in the vicinity of f (a) and the slope of the line is f ′ (a).
For two-variable functions, z = f (x, y) is differentiable at (a, b), if fx (a, b) and fy (a, b) exist
and the change in z as (x, y) varies in the vicinity of (a, b) is equal to

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

= fx (a, b)∆x + fy (a, b)∆y + ε1 ∆x + ε2 ∆y

where ε1 and ε2 are functions of ∆x and ∆y, and (ε1 , ε2 ) → (0, 0) as (∆x, ∆y) → (0, 0).

19
THEOREM Suppose f has fx and fy defined on an open set containing (a, b), and fx and fy
are continuous at (a, b), then f is differentiable at (a, b).

REMARK For two-variable functions, the existence of fx and fy at (a, b) is NOT enough to
guarantee the differentiability of f at (a, b).

REMARK Differentiability implies continuity

lim ∆z = lim [f (a + ∆x, b + ∆y) − f (a, b)] = 0
(∆x,∆y)→(0,0) (∆x,∆y)→(0,0)

lim f (a + ∆x, b + ∆y) = f (a, b) → continuous at (a, b)

(∆x,∆y)→(0,0)

EXAMPLE 6 Is the function

⎧ xy
⎪ x2 + y 2 ,
⎪
⎪ if (x, y) ≠ (0, 0)
f (x, y) = ⎨
⎪
⎩0, if (x, y) = (0, 0)
⎪
⎪
differentiable at (0, 0)?
Solution. Let (x, y) → (0, 0) along y = mx, then
xy mx2 m
lim(x,y)→(0,0) 2 = limx→0 = .
y=mx x + y2 x2 + m2 x2 1 + m2
m-dependence → no limit → not continuous → not differentiable.

f (x, y) is not differentiable at (0, 0), however fx and fy exist at (0, 0):
f (∆x, 0) − f (0, 0) 1 ∆x ⋅ 0
fx (0, 0) = lim∆x→0 = lim∆x→0 [ − 0] = 0
∆x ∆x (∆x)2 + 02
f (0, ∆y) − f (0, 0) 1 ⋅0 ⋅ ∆y
fy (0, 0) = lim∆y→0 = lim∆y→0 [ 2 − 0] = 0.
∆x ∆y 0 + (∆y)2
f is not differentiable, since fx and fy are not continuous at (0, 0). See the graph of f (x, y)

5 The Chain Rule

The chain rule for one-variable function gives the rule for differentiating a composite function:
if y = f (x) and x = g(t), where f and g are differentiable functions, then y is indirectly a
differentiable function of t and dy dy dx
dt = dx dt . For functions of two independent variables, the chain
rule has several versions to deal with different cases.

The Chain Rule: Case 1

Theorem 1 Suppose that z = f (x, y) is a differentiable function of x and y, where x = g(t)
and y = h(t) are both differentiable functions of t. Then z is a dfferentiable function of t and
dz ∂f dx ∂f dy
= +
dt ∂x dt ∂y dt

20
PROOF A change of ∆t in t produces changes of ∆x in x and ∆y in y. These, in turn, produce
a change of ∆z in z, which is equal to
∂f ∂f
∆z = ∆x + ∆y + ε1 ∆x + ε2 ∆y
∂x ∂y

where ε1 → 0 and ε2 → 0 as (∆x, ∆y) → (0, 0). [z is differentiable.] Dividing both sides of this
equation by ∆t, we have
∆z ∂f ∆x ∂f ∆y ∆x ∆y
= + + ε1 + ε2
∆t ∂x ∆t ∂y ∆t ∆t ∆t

We now let ∆t → 0, then ∆x = g(t + ∆t) − g(t) → 0 and ∆y = h(t + ∆t) − h(t) → 0 because g
and h are differentiable. This, in turn, means that ε1 → 0 and ε2 → 0, so
dz ∆z
= lim
dt ∆t→0 ∆t
∂f ∆x ∂f ∆y ∆x ∆y
= lim + lim + lim ε1 ⋅ lim + lim ε2 ⋅ lim
∂x ∆t→0 ∆t ∆t→0 ∂y ∆t ∆t→0 ∆t→0 ∆t ∆t→0 ∆t→0 ∆t
∂f dx ∂f dy dx dy ∂f dx ∂f dy
= + +0⋅ +0⋅ = + ◻
∂x dt ∂y dt dt dt ∂x dt ∂y dt

Extension Functions with three independent variables w = f (x, y, z), where x, y, and z are
functions of the single variable t, then
dw ∂w dx ∂w dy ∂w dz
= + +
dt ∂x dt ∂y dt ∂z dt
dz π
EXAMPLE 1 If z = x2 − 3y 2 + 20 with x = 2 cos t and y = 2 sin t, calculate at t = .
dt 4
∂z ∂z dx dy
Solution. = 2x, = −6y, = −2 sin t, = 2 cos t, then
∂x ∂y dt dt
dz ∂z dx ∂z dy
= + = (2x)(−2 sin t) + (−6y)(2 cos t) = −8 sin t cos t − 24 sin t cos t = −16 sin 2t.
dt ∂x dt ∂y dt
dz π
Therefore, ∣ = −16 sin = −16. The geometric picture is shown below.
dt t= π 2
4

21
The Chain Rule: Case 2
Theorem 2 Suppose that z = f (x, y) is a differentiable function of x and y, where x = g(s, t)
and y = h(s, t) are differentiable functions of s and t. Then
∂z ∂z ∂x ∂z ∂y ∂z ∂z ∂x ∂z ∂y
= + = +
∂s ∂x ∂s ∂y ∂s ∂t ∂x ∂t ∂y ∂t
Theorem 2 could be directly obtained by Theorem 1: in computing ∂z/∂s we hold t fixed and
compute the ordinary derivative of z with respect to s. A similar argument holds for ∂z/∂t.

EXAMPLE 2 Given z = sin 2x cos 3y, x = s + t and y = s − t, calculate ∂z/∂s and ∂z/∂t.
Solution.
∂z ∂z ∂x ∂z ∂y
= + = (2 cos 2x cos 3y)(1) + (−3 sin 2x sin 3y)(1)
∂s ∂x ∂s ∂y ∂s
= 2 cos 2x cos 3y − 3 sin 2x sin 3y = 2 cos[2(s + t)] cos[3(s − t)] − 3 sin[2(s + t)] sin[3(s − t)]

∂z ∂z ∂x ∂z ∂y
= + = (2 cos 2x cos 3y)(1) + (−3 sin 2x sin 3y)(−1)
∂t ∂x ∂t ∂y ∂t
= 2 cos 2x cos 3y + 3 sin 2x sin 3y = 2 cos[2(s + t)] cos[3(s − t)] + 3 sin[2(s + t)] sin[3(s − t)]

Implicit Differentiation
We suppose that an equation of the form F (x, y) = 0 defines y implicitly as a differentiable
function of x. If F is differentiable, we can apply Theorem 1 to differentiate both sides of the
equation F (x, y) = 0 with respect to x. Since both x and y are functions of x, we obtain
∂F dx ∂F dy
+ =0
∂x dx ∂y dx
But dx/dx = 1, so if ∂F /∂y ≠ 0 we solve for dy/dx and obtain
∂F
dy Fx
= − ∂x = −
dx ∂F Fy
∂y
EXAMPLE 3 If F (x, y) = sin xy + πy 2 − x = 0, find dy/dx.
dy Fx y cos xy − 1
Solution. Fx = y cos xy − 1, Fy = x cos xy + 2πy, then =− =−
dx Fy x cos xy + 2πy
EXAMPLE 4 A fluid flow √ with velocity v = ⟨u, v⟩, where u = 2 sin πx cos πy and v = −2 cos πx sin π[Link]
speed of the flow is s = u2 + v 2 , apply the chain rule to find ∂s/∂x and ∂s/∂y.
∂s 2u u ∂s 2v v
Solution. = √ = , = √ =
∂u 2 u2 + v 2 s ∂v 2 u2 + v 2 s
∂s ∂s ∂u ∂s ∂v u v 2π
= + = (2π cos πx cos πy)+ (2π sin πx sin πy) = (u cos πx cos πy+v sin πx sin πy)
∂x ∂u ∂x ∂v ∂x s s s
∂s ∂s ∂u ∂s ∂v u v 2π
= + = (−2π sin πx sin πy)+ (−2π cos πx cos πy) = − (u sin πx sin πy+v cos πx cos πy)
∂y ∂u ∂y ∂v ∂y s s s

22
6 Directional Derivatives & Gradient
Directional Derivatives
Recall that if z = f (x, y), then the partial derivatives fx and fy are defined as

f (x0 + h, y0 ) − f (x0 , y0 ) f (x0 , y0 + h) − f (x0 , y0 )

fx (x0 , y0 ) = lim , fy (x0 , y0 ) = lim
h→0 h h→0 h
and represent the rates of change of z in the x- and y-directions, that is, in the directions of
the unit vector i and j.
We now find the rate of change of z at (x0 , y0 ) in the direction of an arbitrary unit
vector u = ⟨a, b⟩. Consider the surface S with the equation z = f (x, y) and let z0 = f (x0 , y0 ).
Then the point P (x0 , y0 , z0 ) lies on S. The vertical plane that passes through P in the direction
of u intersects S in a curve C. The slope of the tangent line T to C at the point P is the
rate of change of z in the direction of u. If Q(x, y, z) is another point on C and P ′ , Q′

ÐÐ→
are the projections of P , Q onto the xy-plane, then the vector P ′ Q′ is parallel to u and so
ÐÐ→
P ′ Q′ = hu = ⟨ha, hb⟩ for some scalar h. Therefore, x = x0 + ha, y = y0 + hb, and

∆z z − z0 f (x0 + ha, y0 + hb) − f (x0 , y0 )

= =
h h h
If we take the limit as h → 0, we obtain the rate of change of z with respect to distance in the
direction of u, which is called the directional derivative of f in the direction of u.

DEFINITION The directional derivative of f at (x0 , y0 ) in the direction of a unit vector

u = ⟨a, b⟩ is

f (x0 + ha, y0 + hb) − f (x0 , y0 )

Du f (x0 , y0 ) = lim
h→0 h
if this limit exists.

REMARK The partial derivatives of f with respect to x and y are just special cases of the
directional derivative: fx = Di f and fy = Dj f .

23
THEOREM If f is a differentiable function of x and y, then f has a directional derivative in
the direction of any unit vector u = ⟨a, b⟩ and

Du f (x, y) = fx (x, y)a + fy (x, y)b = ⟨fx (x, y), fy (x, y)⟩ ⋅ ⟨a, b⟩

PROOF If we define a function g of the single variable h by g(h) = f (x0 + ha, y0 + hb), then,
by the definition of the directional derivative at (x0 , y0 ), we have
f (x0 + ha, y0 + hb) − f (x0 , y0 ) g(h) − g(0)
Du f (x0 , y0 ) = lim = lim = g ′ (0)
h→0 h h→0 h
On the other hand, we can write g(h) = f (x, y), where x = x0 + ha, y = y0 + hb, so the Chain
Rule gives
∂f dx ∂f dy
g ′ (h) = + = fx (x, y)a + fy (x, y)b
∂x dh ∂y dh
If we now put h = 0, then x = x0 , y = y0 and

Du f (x0 , y0 ) = g ′ (0) = fx (x0 , y0 )a + fy (x0 , y0 )b ◻

If the unit vector u makes an angle θ with the positive x-axis, then we can write u =
⟨cos θ, sin θ⟩ and the formula becomes

Du f (x, y) = fx (x, y) cos θ + fy (x, y) sin θ = ⟨fx (x, y), fy (x, y)⟩ ⋅ ⟨cos θ, sin θ⟩

√
1 2 1 2 1 1 1 3
EXAMPLE 1 z = f (x, y) = x + y + 2 (paraboloid), u = ⟨ √ , √ ⟩, v = ⟨ , − ⟩ (already
4 2 2 2 2 2
unit vectors). Find Du f and Dv f at P0 (3, 2).
1 3
Solution. fx = x, fy = y, then ⟨fx (P0 ), fy (P0 )⟩ = ⟨ , 2⟩
2 2
3 1 1 7
Du f (P0 ) = ⟨ , 2⟩ ⋅ ⟨ √ , √ ⟩ = √
2 2√2 2 2
3 1 3 3 √
Dv f (P0 ) = ⟨ , 2⟩ ⋅ ⟨ , − ⟩= − 3
2 2 2 4

The Gradient Vector

DEFINITION If f is a function of two variables x and y, then the gradient of f is the vector
function ∇f defined by
∂f ∂f
∇f (x, y) = ⟨fx (x, y), fy (x, y)⟩ = i+ j
∂x ∂y

24
 Therefore Du f (x, y) = ∇f (x, y) ⋅ u with u a unit vector.

EXAMPLE 2 f (x, y) = x2 + 2xy − y 3 . Find ∇f at (3, 2).

∂f ∂f
Solution. ∇f (x, y) = i+ j = (2x + 2y)i + (2x − 3y 2 )j. Then, ∇f (3, 2) = 10i − 6j.
∂x ∂y
x2 xy 2
EXAMPLE 3 f (x, y) = 3 − + . Compute the directional derivative of f at (3, −1) in
10 10
the direction of u = ⟨3, 4⟩.
1 1 1
Solution. fx (x, y) = − x + y 2 , fy (x, y) = xy
5 10 5
3 1 3 1 3
∇f (3, −1) = fx (3, −1)i + fy (3, −1)j = (− + ) i + (− ) j = ⟨− , − ⟩
5 10 5 2 5
u 3 4
The unit vector in the direction of u is v = =⟨ , ⟩
∣u∣ 5 5
1 3 3 4 39
The directional derivative is Dv f (3, −1) = ⟨− , − ⟩ ⋅ ⟨ , ⟩ = −
2 5 5 5 50

Interpretation of the gradient

The interpretation based on the directional derivative Du f (x0 , y0 ) = ∇f (x0 , y0 )⋅u = ∣∇f (x0 , y0 )∣ cos α,
where α is the angle between ∇f (x0 , y0 ) and u. Note that ∣u∣ = 1.
1. Maximum rate of increase of f : if u is in the direction of ∇f (x0 , y0 ), i.e., α = 0
max Du f (x0 , y0 ) = ∣∇f (x0 , y0 )∣ > 0

2. Maximum rate of decrease of f : if u is in the opposite direction of ∇f (x0 , y0 ), i.e., α = π

min Du f (x0 , y0 ) = −∣∇f (x0 , y0 )∣ < 0
1
EXAMPLE 4 Find the directions at (2, − ) in which z = f (x, y) = 4 + x2 + 3y 2 has steepest
2
ascent and descent.
1 1 1
Solution. Since ∇f (2, − ) = ⟨fx (2, − ) , fy (2, − )⟩ = ⟨4, −3⟩
2 2 2
4 3
steepest ascent is in the direction of ⟨4, −3⟩ or ⟨ , − ⟩
5 5
4 3
steepest descent is in the direction of − ⟨4, −3⟩ or ⟨− , ⟩
5 5

The gradient and level curves

A level curve of z = f (x, y) is a curve in the x − y plane defined by f (x, y) = z0 . Suppose the
curve is parametrised by t as
⎧
⎪x = x(t)
⎪
⎨
⎩y = y(t)
⎪
⎪
or it can be represented by the vector-valued function r(t) = ⟨x(t), y(t)⟩ with
f (x(t), y(t)) = z0
Differentiate the above equation and apply the chain rule, we have
d dx dy
f (x(t), y(t)) = fx (x, y) + fy (x, y) = ⟨fx , fy ⟩ ⋅ ⟨x′ (t), y ′ (t)⟩ = ∇f ⋅ r ′ (t) = 0
dt dt dt
Therefore, the tangent vector of r(t) is orthogonal to ∇f .

25
Equation of Tangent Line

√
EXAMPLE 5 Given z = f (x, y) = 1 + 2x2 + y 2 the upper sheet of a two-sheet hyperboloid.
(a) Calculate ∇f (1, 1).
(b) Find the slope of the line tangent to the level curve at (1, 1).
(c) Find the equation of the tangent line.
2x y 1
Solution. (a) fx (x, y) = √ , fy (x, y) = √ , then ∇f (1, 1) = ⟨1, ⟩
1 + 2x2 + y 2 1 + 2x2 + y 2√ 2
(b) Since f (1, 1) = 2, the level curve through (1, 1) is given as 1 + 2x2 + y 2 = 2 or 2x2 +y 2 = 3,
which is an ellipse. We find the slope of the tangent line to the level curve by taking d/dx to
dy dy 2x
the equation of the level curve: 4x + 2y = 0, therefore ∣ =− ∣ = −2
dx dx (1,1) y (1,1)
1
The tangent vector is T = ⟨1, −2⟩ or any of its scalar multiple. ∇f (1, 1)⋅T = ⟨1, ⟩⋅⟨1, −2⟩ = 0.
2
Orthogonality between ∇f and T verified.
(c) Equation of the tangent line at (1, 1) is fx (1, 1)(x−1)+fy (1, 1)(y−1) = 0, thus y = −2x+3
(slope= −2)

26
Gradient in Three Dimensions
Let w = f (x, y, z) be differentiable, the gradient of f is the vector function ∇f defined by
∂f ∂f ∂f
∇f (x, y, z) = i+ j+ k
∂x ∂y ∂z
The directional derivative of f at (x0 , y0 , z0 ) in the direction of a unit vector u = ⟨u1 , u2 , u3 ⟩ is

Du f (x0 , y0 , z0 ) = ∇f (x0 , y0 , z0 ) ⋅ u

Let w0 = f (x0 , y0 , z0 ), f (x, y, z) = w0 is a level surface of w = f (x, y, z) and the point (x0 , y0 , z0 )
is on the level surface. Then, ∇f (x0 , y0 , z0 ) is orthogonal to the level surface at (x0 , y0 , z0 ).

EXAMPLE 6 For w = f (x, y, z) = x2 + 2y 2 + 4z 2 − 1, the level surface for w = 3 is given

x2 + 2y 2 + 4z 2 = 4, which is an ellipsoid. Three intercepts of the level surface are (2, 0, 0),
by √
(0, 2, 0) and (0, 0, 1).
The gradient vector of f is ∇f (x, y, z) = ⟨2x, 4y, 8z⟩, which is orthogonal to the level surface
at each point.

7 Tangent Plane & Linear Approximation

If z = f (x, y) is differentiable, the graph is a smooth surface S in R3 . Focus on a single point
P (x0 , y0 , z0 ) on S with z0 = f (x0 , y0 ), in the vicinity of P , the surface appears like a plane, i.e.,
it can be approximated by the tangent plane to the surface S at P .

Tangent Plane to a Surface with Equation F (x, y, z) = k

Suppose that a surface S is implicitly defined as a level surface of w = F (x, y, z) for w = k. Let
P (x0 , y0 , z0 ) be a point on S. Let C be any curve that lies on the surface S and passes through
the point P . Recall that C can be described by a vector-valued function r(t) = ⟨x(t), y(t), z(t)⟩.
Let t0 be the parameter value corresponding to P ; that is r(t0 ) = ⟨x0 , y0 , z0 ⟩. Since C lies on
the surface, any point (x(t), y(t), z(t)) must satisfy the equation of S

F (x(t), y(t), z(t)) = k

If x, y and z are differentiable functions of t and F is also differentiable, then we can use the
Chain Rule to differentiate both side of the above equation as
∂F dx ∂F dy ∂F dz
+ + = 0 or ∇F ⋅ r ′ (t) = 0
∂x dt ∂y dt ∂z dt

27
In particular, when t = t0 , ∇F (x0 , y0 , z0 ) ⋅ r ′ (t0 ) = 0. This means that at P , the gradient vector
∇F (x0 , y0 , z0 ) is perpendicular to the tangent vector r ′ (t0 ). Since the curve C is arbitrarily
defined, then, ∇F (x0 , y0 , z0 ) is orthogonal to all the tangent vectors to any curve on the surface
that passes through P . These tangent vectors line in the same plane, which is the tangent
plane to the surface S at P with a normal vector ∇F (x0 , y0 , z0 ). We can write the equation
of this tangent plane as

∇F (x0 , y0 , z0 ) ⋅ ⟨x − x0 , y − y0 , z − z0 ⟩ = 0

or

Fx (x0 , y0 , z0 )(x − x0 ) + Fy (x0 , y0 , z0 )(y − y0 ) + Fz (x0 , y0 , z0 )(z − z0 ) = 0

x2 y 2
EXAMPLE 1 F (x, y, z) = + + z 2 − 1 = 0 is an ellipsoid.
9 25
3
(a) Find the tangent plane at (0, 4, )
5
(b) At which points the tangent planes are horizontal?

3 2x 2y 8 6
Solution. (a) ∇F (0, 4, ) = ⟨ , , 2z⟩∣ = ⟨0, , ⟩
5 9 25 (0,4, 35 )
25 5
8 6 3
Therefore, the tangent plane is given by 0(x − 0) + (y − 4) + (z − ) = 0 or 4y + 15z = 25
25 5 5
(b) A horizontal tangent plane has a normal vector parallel to k, i.e., ∇F = ⟨0, 0, 2z⟩. Since
we are finding points on the surface, we have z = ±1. Therefore, at (0, 0, ±1), the tangent planes
are horizontal.

28
Tangent Plane to Explicitly Defined Surface z = f (x, y)
When a surface is explicitly given as z = f (x, y), the equation of a tangent plane is obtained by
defining a new function F (x, y, z) = f (x, y) − z = 0, which has the gradient vector ∇F (x, y, z) =
⟨fx (x, y), fy (x, y), −1⟩. Therefore the equation for the tangent plane at (x0 , y0 , z0 ), where z0 =
f (x0 , y0 ) is written as

fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ) − (z − z0 ) = 0

or

z = fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ) + f (x0 , y0 )

EXAMPLE 2 Find the tangent plane to a elliptical paraboloid z = 32 − 3x2 − 4y 2 at (2, 1, 16)
Solution. fx = −6x and fy = −8y.
The tangent plane is z = (−12)(x − 2) + (−8)(y − 1) + 16 = 48 − 12x − 8y

Linear Approximation
The linear approximation to the surface z = f (x, y) at the point (x0 , y0 , f (x0 , y0 )) is the tangent
plane given by

L(x, y) = fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ) + f (x0 , y0 )

Note that “approximation” is in the sense: Lx = fx (x0 , y0 ), Ly = fy (x0 , y0 ), L(x0 , y0 ) = f (x0 , y0 ).

29
 5
EXAMPLE 3 Find the linear approximation to z = f (x, y) = at (−1, 2, 1)
x2 + y2
10x 10y
Solution. fx (x, y) = − , f y (x, y) = −
(x2 + y 2 )2 (x2 + y 2 )2
10(−1) 2 10(2) 4
fx (−1, 2) = − = , fy (−1, 2) = − =−
25 5 25 5
2 4 2 4
L(x, y) = (x + 1) − (y − 2) + 1 = x − y + 3
5 5 5 5
L(−1.05, 2.1) = 0.9 → good approximation for f (−1.05, 2.1) ≃ 0.9070295

Difference & Differential

If z = f (x, y) is differentiable, at (x0 , y0 , z0 ), where z0 = f (x0 , y0 ), f (x, y) is approximated by
L(x, y) = fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ) + f (x0 , y0 ).

DEFINITION Difference is the exact change in z between the points (x, y, z) and (x0 , y0 , z0 )

∆z = f (x, y) − f (x0 , y0 )

DEFINITION Differential is the difference between L(x, y) and f (x0 , y0 )

dz = L(x, y) − f (x0 , y0 ) = fx (x0 , y0 )dx + fy (x0 , y0 )dy, with dx = x − x0 , dy = y − y0

Therefore, the difference in z can be approximated by the differential ∆z ≃ dz = fx (x0 , y0 )dx+

fy (x0 , y0 )dy, and the approximation improves as dx and dy approach 0. Linear approxima-
tion is equivalent to differentiability: ∆z = fx (x0 , y0 )∆x + fy (x0 , y0 )∆y + ε1 ∆x + ε2 ∆y with
(ε1 , ε2 ) → (0, 0) as (∆x, ∆y) → (0, 0)
5
EXAMPLE 4 Calculate the difference and differential for z = f (x, y) = when (x, y)
x2 + y2
changes from (−1, 2) to (−0.93, 1.94).
2 4
Solution. fx (−1, 2) = , fy (−1, 2) = − , dx = −0.93 − (−1) = 0.07, dy = 1.94 − 2 = −0.06
5 5
∆z = f (−0.93, 1.94) − f (−1, 2) ≃ 0.0802636
2 4
dz = fx (−1, 2)dx + fy (−1, 2)dy = (0.07) − (−0.06) = 0.076 → good approximation to ∆z
5 5

8 Maximum & Minimum Problems

The following figure illustrates various extreme of a two-variable function, including local max-
imum, local minimum, absolute maximum and absolute minimum.

30
DEFINITION Local Maximum/Minimum Values
A function of two variables has a local maximum at (a, b) if f (x, y) ⩽ f (a, b) when (x, y)
is near (a, b). This means f (x, y) ⩽ f (a, b) for all points (x, y) in some disk with the center at
(a, b) in the domain of f . Then, f (a, b) is a local maximum value.
A function of two variables has a local minimum at (a, b) if f (x, y) ⩾ f (a, b) when (x, y)
is near (a, b). This means f (x, y) ⩾ f (a, b) for all points (x, y) in some disk with the center at
(a, b) in the domain of f . Then, f (a, b) is a local minimum value.

REMARK When finding local extrema, comparison between f (x, y) and f (a, b) is made in a
small neighborhood around (a, b).

THEOREM If f has a local maximum or minimum at (a, b), and fx and fy exist at (a, b),
then fx (a, b) = fy (a, b) = 0
PROOF Let g(x) = f (x, b) with y = b fixed. Since f has a local extrema at (a, b), we have
g ′ (a) = fx (a, b) = 0. Similarly, let h(y) = f (a, y) with x = a fixed, then, h′ (b) = fy (a, b) = 0.

This theorem provides candidates for local extrema.

DEFINITION Critical Points

An interior point (a, b) within the domain of f is a critical points if either

1. fx (a, b) = fy (a, b) = 0

2. at least one of fx and fy does not exist at (a, b)

EXAMPLE 1 Find all the critical points of f (x, y) = xy(x − 2)(y + 3)

Solution. f is a polynomial. Partial derivatives exist everywhere. Critical points are the ones
at which fx = fy = 0.
fx = 2(x − 1)y(y + 3), fy = x(x − 2)(2y + 3)
fx = 0 → x = 1, y = 0, or y = −3
3
when x = 1, fy = −(2y + 3) = 0 → y = −
2
when y = 0, fy = 3x(x − 2) = 0 → x = 0 or x = 2
when y = −3, fy = −3x(x − 2) = 0 → x = 0 or x = 2
3
Five critical points: (1, − ), (0, 0), (2, 0), (0, −3) and (2, −3). Some of them may correspond
2
to local extrema.

31
THEOREM Second Derivative Test
Suppose that the second partial derivatives of f are continuous throughout an open disk
centered at the point (a, b), where fx (a, b) = fy (a, b) = 0.
2
Let D(x, y) = fxx (x, y)fyy (x, y) − [fxy (x, y)]

1. If D(a, b) > 0 and fxx (a, b) < 0, then f has a local maximum value at (a, b)

2. If D(a, b) > 0 and fxx (a, b) > 0, then f has a local minimum value at (a, b)

3. If D(a, b) < 0, then f has a saddle point at (a, b)

4. If D(a, b) = 0, then the test is inconclusive

DEFINITION Saddle Point

A critical point (a, b) is a saddle point if in every open disk centered at (a, b), there are
points (x, y) for which f (x, y) > f (a, b) and points for which f (x, y) < f (a, b).

EXAMPLE 2 Show that (0, 0) is a saddle point for f (x, y) = xy(x − 2)(y + 3)
Solution. (0, 0) is a critical point for f (see EXAMPLE 1)
fxx = 2y(y + 3), fyy = 2x(x − 2), fxy = fyx = 2(x − 1)(2y + 3)
Therefore fxx (0, 0) = fyy (0, 0) = 0, fxy (0, 0) = fyx (0, 0) = −6
D(0, 0) = −36 < 0, then (0, 0) is a saddle point

EXAMPLE 3 Find the local extrema of f (x, y) = x2 + 2y 2 − 4x + 4y + 6.

Solution. fx = 2x − 4, fy = 4y + 4
f has one critical point (2, −1) where fx (2, −1) = fy (2, −1) = 0
fxx = 2, fyy = 4, and fxy = fyx = 0, then, D(2, −1) = fxx fyy − (fxy )2 = 8 > 0 and fxx > 0,
therefore f has a local minimum at (2, −1), which is f (2, −1) = 0.

32
EXAMPLE 4 Show that (0, 0) is a saddle point of f (x, y) = x4 + y 4 − 4xy + 1. Sketch the level
curves near the saddle point.
Solution. fx = 4x3 − 4y, fy = 4y 3 − 4x, fxx = 12x2 , fyy = 12y 2 , fxy = fyx = −4
Since fx (0, 0) = fy (0, 0) = 0, (0, 0) is a critical point. Moreover, D(0, 0) = −16 < 0, therefore
(0, 0) is a saddle point.
The graph and level curves of f are shown below. Level curves are crossed at the saddle
point.

DEFINITION Absolute Maximum/Minimum Values

If f (x, y) ⩽ f (a, b) for all (x, y) in the domain of f , then f (a, b) is the absolute maximum
value of f . If f (x, y) ⩾ f (a, b) for all (x, y) in the domain of f , then f (a, b) is the absolute
minimum value of f .

PROCEDURE Find Absolute Maximum/Minimum Values

Let f be continuous on a closed and bounded set R in R2 . We find the absolute maximum
and minimum values of f on R by

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 step 1 and 2 is the absolute maximum value of f ,
and the least function value found in step 1 and 2 is the minimum value of f

EXAMPLE 5 Find the absolute maximum and minimum values of f (x, y) = x2 +y 2 −2x+2y +5
on R = {(x, y) ∣ x2 + y 2 ⩽ 4}
Solution. We first find the critical points from fx = 2x − 2 = 0 and fy = 2y + 2 = 0. Therefore,
(1, −1) is the only critical point in R, at which f (1, −1) = 3.
Second, we find the extrema on the boundary of R where x2 + y 2 = 4. The boundary
can be parametrized by x = 2 cos t, y = 2 sin t for t ∈ [−π, π]. Then, on the boundary of R,
g(t) = f (x(t), y(t)) = 4 cos2 t + 4 sin2 t − 4 cos t + 4 sin t + 5 = 9 − 4 cos t + 4 sin t.
d π
Extrema on the boundary of R occur when g(t) = 4 sin t + 4 cos t = 0, i.e., t = − and
dt 4
3π √ √ √
t= . Then, the extrema are obtained as f (x(−π/4), y(−π/4)) = f ( 2, − 2) = 9 − 4 2 and
4 √ √ √
f (x(3π/4), y(3π/4)) = f (− 2, 2) = 9 + 4 2
Finally, we compare

33
f (1,
√ −1)√ =3 √ → absolute minimum
f( √ 2, −√2) = 9 − 4√2
f (− 2, 2) = 9 + 4 2 → absolute maximum

34