Marsden-3620111 VC September 27, 2011 9:27 76

76 Differentiation

2.1 The Geometry of Real-Valued Functions

We launch our investigation of real-valued functions by developing methods for visual-
izing them. In particular, we introduce the notions of a graph, a level curve, and a level
surface of such functions.

Functions and Mappings

Let f be a function whose domain is a subset A of Rn and with a range contained
in Rm . By this we mean that to each x = (x1 , . . . , xn ) ∈ A, f assigns a value f (x),
an m-tuple in Rm . Such functions f are called vector-valued functions1 if m > 1, and
scalar-valued functions if m = 1. For example, the scalar-valued function f (x, y, z) =
(x 2 + y 2 + z 2 ) −3/2 maps the set A of (x, y, z) =
 (0, 0, 0) in R3 (n = 3, in this case) to
R (m = 1). To denote f we sometimes write
f : (x, y, z) → (x 2 + y 2 + z 2 ) −3/2 .
Note that in R3 we often use the notation (x, y, z) instead of (x1 , x2 , x3 ). In general, the
notation x → f (x) is useful for indicating the value to which a point x ∈ Rn is sent.
We write f : A ⊂ Rn → Rm to signify that A is the domain of f (a subset of Rn ) and the
range is contained in Rm . We also use the expression f maps A into Rm . Such functions
f are called functions of several variables if A ⊂ Rn , n > 1.
As another example we can take the vector-valued function g: R6 → R2 defined by
the rule
  
g(x) = g(x1 , x2 , x3 , x4 , x5 , x6 ) = x1 x2 x3 x4 x5 x6 , x12 + x62 .

The first coordinate of the value of g at x is the product of the coordinates of x.

Functions from Rn to Rm are not just mathematical abstractions, they arise natu-
rally in problems studied in all the sciences. For example, to specify the temperature
T in a region A of space requires a function T : A ⊂ R3 → R (n = 3, m = 1);
thus, T (x, y, z) is the temperature at the point (x, y, z). To specify the velocity of
a fluid moving in space requires a map V: R4 → R3 , where V(x, y, z, t) is the ve-
locity vector of the fluid at the point (x, y, z) in space at time t (see Figure 2.1.1).
To specify the reaction rate of a solution consisting of six reacting chemicals A, B,
C, D, E, F in proportions x, y, z, w, u, v requires a map σ : U ⊂ R6 → R, where
σ (x, y, z, w, u, v) gives the rate when the chemicals are in the indicated proportions.

(x, y, z)
( , y, z, t ) = Fluid velocity
V(x
V
figure 2.1.1 A fluid in motion defines
a vector field V by specifying the
velocity of the fluid particles at each
point in space and time.

1
Some mathematicians would write such an f in boldface, using the notation f(x), because the function
is vector-valued. We did not do so, as a matter of personal taste. We use boldface primarily for mappings
that are vector fields, introduced later. The notion of function was developed over many centuries, with
the definition extended to cover more cases as they arose. For example, in 1667 James Gregory defined
a function as “a quantity obtained from other quantities by a succession of algebraic operations or
by any other operation imaginable.” In 1755 Euler gave the following definition: “If some quantities
depend on others in such a way as to undergo variation when the latter are varied then the former are
called functions of the latter.”
Marsden-3620111 VC September 27, 2011 9:27 77

2.1 The Geometry of Real-Valued Functions 77

To specify the cardiac vector (the vector giving the magnitude and direction of electric
current flow in the heart) at time t requires a map c: R → R3 , t → c(t).
When f : U ⊂ Rn → R, we say that f is a real-valued function of n variables with
domain U . The reason we say “n variables” is simply that we regard the coordinates
of a point x = (x1 , . . . , xn ) ∈ U as n variables, and f (x) = f (x 1 , . . . , xn ) depends on
these variables. We say “real-valued” because f (x1 , . . . , xn ) is a real number. A good
deal of our work will be with real-valued functions, so we give them special attention.

Graphs of Functions
For f : U ⊂ R → R (n = 1), the graph of f is the subset of R2 consisting of all points
(x, f (x)) in the plane, for x in U . This subset can be thought of as a curve in R2 . In
symbols, we write this as
graph f = {(x, f (x)) ∈ R2 | x ∈ U },
where the curly braces mean “the set of all” and the vertical bar is read “such that.”
Drawing the graph of a function of one variable is a useful device to help visualize how
the function actually behaves (see Figure 2.1.2). It will be helpful to generalize the idea
of a graph to functions of several variables. This leads to the following definition:

Definition Graph of a Function Let f : U ⊂ Rn → R. Define the graph

of f to be the subset of Rn+1 consisting of all the points

(x1 , . . . , xn , f (x1 , . . . , xn ))

in Rn+1 for (x1 , . . . , xn ) in U . In symbols,

graph f = {(x1 , . . . , xn , f (x1 , . . . , xn )) ∈ Rn+1 | (x1 , . . . , xn ) ∈ U }.

For the case n = 1, the graph is a curve in R2 , while for n = 2, it is a surface in R3

(see Figure 2.1.2). For n = 3, it is difficult to visualize the graph, because, since we are
humans living in a three-dimensional world, it is hard for us to envisage sets in R4 . To
help overcome this handicap, we introduce the idea of a level set.

y z

Graph of f

Graph of f

x y
U
U

x
(a) (b)

figure 2.1.2 The graphs of (a) a function of one variable, and (b) a function of two variables.
 Marsden-3620111 VC September 27, 2011 9:27 78

78 Differentiation

h = 50
h = 100
h = 150
figure 2.1.3 Level contours of a
function are defined in the same h h = 200
manner as contour lines for a
topographical map.

(a) (b)

Level Sets, Curves, and Surfaces

Suppose f (x, y, z) = x 2 + y 2 + z 2 . A level set is a subset of R3 on which f is
constant; for instance, the set where x 2 + y 2 + z 2 = 1 is a level set for f . This we
can visualize: It is just a sphere of radius 1 in R3 . Formally, a level set is the set of
(x, y, z) such that f (x, y, z) = c, where c is a constant. The behavior or structure of a
function is determined in part by the shape of its level sets; consequently, understanding
these sets aids us in understanding the function in question. Level sets are also useful
for understanding functions of two variables f (x, y), in which case we speak of level
curves or level contours.
The idea is similar to that used to prepare contour maps, where one draws lines to
represent constant altitudes; walking along such a line would mean walking on a level
path. In the case of a hill rising from the x y plane, a graph of all the level curves gives
us a good idea of the function h(x, y), which represents the height of the hill at point
(x, y) (see Figure 2.1.3).

example 1 The constant function f : R2 → R, (x, y) → 2—that is, the function f (x, y) = 2—has
as its graph the horizontal plane z = 2 in R3 . The level curve of value c is empty if
c= 2, and is the whole x y plane if c = 2. ▲

example 2 The function f : R2 → R, defined by f (x, y) = x + y + 2, has as its graph the inclined
plane z = x + y + 2. This plane intersects the x y plane (z = 0) in the line y = −x − 2
and the z axis at the point (0, 0, 2). For any value c ∈ R, the level curve of value c is
the straight line y = −x + (c − 2); or in symbols, the set
L c = {(x, y) | y = −x + (c − 2)} ⊂ R2 .

We indicate a few of the level curves of the function in Figure 2.1.4. This is a contour
map of the function f .

f(x
, y)
f(x
, =
y) x+
= figure 2.1.4 The level curves of f (x, y) = x + y + 2
f(x x+ y+
,y 2 show the sets on which f takes a given value.
)= y+ =
2 4 x
Line of x+ =
y+ 2
intersection of
2
plane z = x + y + 2 =
0
and the x y plane
Marsden-3620111 VC September 27, 2011 9:27 79

2.1 The Geometry of Real-Valued Functions 79

x+y+2= 4 Level curves lifted

to surface
x+y+2= 2 2

Level curve x + y + 2 = 0 −2
−2
x + y + 2 = 2 in xy plane
2
x + y + 2 = 4 in xy plane 2 y

figure 2.1.5 The relationship of level curves of Figure 2.1.4 to the graph of
the function f (x, y) = x + y + 2, which is the plane z = x + y + 2.

From level curves labeled with the value or “height” of the function, the shape of the
graph may be inferred by mentally elevating each level curve to the appropriate height,
without stretching, tilting, or sliding it. If this procedure is visualized for all level curves,
L c —that is, for all values c ∈ R, they will assemble to give the entire graph of f , as
indicated by the shaded plane in Figure 2.1.5. If the graph is visualized using a finite
number of level curves, a contour model is produced. If f is a smooth function, its graph
will be a smooth surface, and so the contour model, mentally smoothed over, gives a
good impression of the graph. ▲

Definition Level Curves and Surfaces Let f : U ⊂ Rn → R and let c ∈

R. Then the level set of value c is defined to be the set of those points x ∈
U at which f (x) = c. If n = 2, we speak of a level curve (of value c);
and if n = 3, we speak of a level surface. In symbols, the level set of value
c is written

{x ∈ U | f (x) = c} ⊂ Rn .

Note that the level set is always in the domain space.

example 3 Describe the graph of the quadratic function

f : R2 → R, (x, y) → x 2 + y 2 .

solution The graph is the paraboloid of revolution z = x 2 + y 2 , oriented upward from the origin,
around the z axis. The level curve of value c is empty for c < 0; for √ c > 0 the level
curve of value c is the set {(x, y) | x 2 + y 2 = c}, a circle of radius c centered at√ the
origin. Thus, raised to height c above the x y plane, the level set is a circle of radius c,
indicating a parabolic shape (see Figures 2.1.6 and 2.1.7).
 Marsden-3620111 VC September 27, 2011 9:27 80

80 Differentiation

figure 2.1.6 Some level curves for the function

1 2 3 4
x f (x, y) = x 2 + y2 .

x2 + y2 = 12
x2 + y2 = 22
x2 + y2 = 32
x2 + y2 = 42

16 x2 + y2 = 42

figure 2.1.7 Level curves in Figure 2.1.6 raised to

9 x2 + y2 = 32
the graph.

4 x2 + y2 = 22

1 x2 + y2 = 12
y

x ▲

The Method of Sections

By a section of the graph of f we mean the intersection of the graph and a (vertical)
plane. For example, if P1 is the x z plane in R3 , defined by y = 0, then the section of f
in Example 3 is the set

P1 ∩ graph f = {(x, y, z) | y = 0, z = x 2 },

which is a parabola in the x z plane. Similarly, if P2 denotes the yz plane, defined by

x = 0, then the section

P2 ∩ graph f = {(x, y, z) | x = 0, z = y 2 }

is a parabola in the yz plane (see Figure 2.1.8). It is usually helpful to compute at least
one section to complement the information given by the level sets.

example 4 The graph of the quadratic function

f : R2 → R, (x, y) → x 2 − y 2

is called a hyperbolic paraboloid, or saddle, centered at the origin. Sketch the graph.
Marsden-3620111 VC September 27, 2011 9:27 81

2.1 The Geometry of Real-Valued Functions 81

S 1: z = x 2, y = 0

S 2 : z = y 2, x = 0

figure 2.1.8 Two sections of the graph

of f (x, y) = x 2 + y 2 .

solution To visualize this surface, we first draw the level curves. To determine the level curves, we
solve the equation x 2 − y 2 = c. Consider the values c = 0, ±1, ±4. For c = 0, we have
y 2 = x 2 , or y = ±x, so that this level set consists of two
√ straight lines through the origin.
For c = 1, the level curve is x 2 − y 2 = 1, or y = ± x 2 − 1, which is a hyperbola that
passes vertically through the x axis at the points√ (±1, 0) (see Figure 2.1.9). Similarly,
for c = 4, the level curve is defined by y = ± x 2 − 4, the hyperbola passing vertically
throughthe x axis at (±2, 0). For c = −1, we obtain the curve x 2 − y 2 = −1—that is,
x = ± y 2 − 1—the hyperbola passing horizontally through the y axis at (0, ±1). And
for c = −4, the hyperbola through (0, ±2) is obtained. These level curves are shown
in Figure 2.1.9. Because it is not easy to visualize the graph of f from these data alone,
we shall compute two sections, as in the previous example. For the section in the x z
plane, we have

P1 ∩ graph of f = {(x, y, z) | y = 0, z = x 2 },

which is a parabola opening upward; and for the yz plane,

P2 ∩ graph f = {(x, y, z) | x = 0, z = −y 2 },

2
2 −1
2 =
y
−
x2

x2

x2
= 22
x2

12

0
−y

−y

=
−

1 2
=
2

− y2
2

y2

y
y2

−
=2

=1

2
−
0

x
x2
2

x2

x
−2 −1 1 2

−1

−2

figure 2.1.9 Level curves for the function f (x, y) = x 2 − y2 .

 Marsden-3620111 VC September 27, 2011 9:27 82

82 Differentiation

z
z = x 2, y = 0

x 2 − y 2 = 12; z = 1

x x 2 − y 2 = − (12 ); z = −1

y
z = − y 2, x = 0

figure 2.1.10 Some level curves on the graph of

f (x, y) = x 2 − y 2 .

z
2

y axis
0

x
−1
y
−2
−2 −1 0 1 2
x axis

figure 2.1.11 The graph of z = x 2 − y 2 and its level curves.

which is a parabola opening downward. The graph may now be visualized by lifting the
level curves to the appropriate heights and smoothing out the resulting surface. Their
placement is aided by computing the parabolic sections. This procedure generates the
hyperbolic saddle indicated in Figure 2.1.10. Compare this with the computer-generated
graphs in Figure 2.1.11 (note that the orientation of the axes has been changed). ▲

example 5 Describe the level sets of the function

f : R3 → R, (x, y, z) → x 2 + y 2 + z 2 .

solution This is the three-dimensional analogue of Example 3. In this context, level sets are
surfaces in the three-dimensional domain R3 . The graph, in R4 , cannot be visualized
directly, but sections can nevertheless be computed.
The level set with value c is the set

L c = {(x, y, z) | x 2 + y 2 + z 2 = c},
√
which is the sphere centered at the origin with radius c for c > 0, is a single point at
the origin for c = 0, and is empty for c < 0. The level sets for c = 0, 1, 4, and 9 are
indicated in Figure 2.1.12.
Marsden-3620111 VC September 27, 2011 9:27 83

2.1 The Geometry of Real-Valued Functions 83

x 2 + y 2 + z 2 = 12
x2 + y2 + z2 x 2 + y 2 + z 2 = 22
x 2 + y 2 + z 2 = 32

figure 2.1.12 Some level surfaces for

f (x, y, z) = x 2 + y 2 + z 2 . ▲

example 6 Describe the graph of the function f : R3 → R defined by f (x, y, z) = x 2 + y 2 − z 2 ,

which is the three-dimensional analogue of Example 4, and is also called a saddle.

solution Formally, the graph of f is a subset of four-dimensional space. If we denote points in

this space by (x, y, z, t), then the graph is given by
{(x, y, z, t) | t = x 2 + y 2 − z 2 }.
The level surfaces of f are defined by
L c = {(x, y, z) | x 2 + y 2 − z 2 = c}.

For c = 0, this is the cone z =± x 2 + y 2 centered on the z axis. For c negative,
say, c = −a 2 , we obtain z = ± x 2 + y 2 + a 2 , which is a hyperboloid of two sheets
around the z axis, passing through the z axis at the points (0, 0, ±a). For c positive, say,
c = b2 , the level surface is the single-sheeted hyperboloid of revolution around the

z axis defined by z = ± x 2 + y 2 − b2 , which intersects the x y plane in the circle of
radius |b|. These level surfaces are sketched in Figure 2.1.13.
z

x 2 + y 2 − z 2 = −2 2
2 x 2 + y 2 − z 2 = −12
1 x 2 + y 2 − z 2 = 02
x 2 + y 2 − z 2 = 12
1
2
x2 + y2 − z2 = 22
x
y

figure 2.1.13 Some level surfaces of the function f (x, y, z) = x 2 + y 2 − z 2 .

 Marsden-3620111 VC September 27, 2011 9:27 84

84 Differentiation

x2 − y2 = 1

x
x 2 − y 2 = −1
z

t = −z 2, x = 0

figure 2.1.14 The y = 0 section of the graph of f (x, y, z) = x 2 + y 2 − z 2 .

Another view of the graph may be obtained from a section. For example, the subspace
Sy=0 = {(x, y, z, t) | y = 0} intersects the graph in the section
Sy=0 ∩ graph f = {(x, y, z, t) | y = 0, t = x 2 − z 2 },

that is, the set of points of the form (x, 0, z, x 2 − z 2 ), which may be considered to be a
surface in x zt space (see Figure 2.1.14). ▲

3 3
2 2
1 1
0 0
−2 −2 −2 −2
−1 −1 −1 −1
0 0 0 0
y axi 1 1 is y ax 1 1 is
s 2 2 x ax is 2 2 x ax
(a) (b)

y axis 0

−1

−2

−2 −1 0 1 2
x axis
(c)

figure 2.1.15 Computer-generated graph of z = (x 2 + 3y 2 ) exp (1 − x 2 − y 2 ) represented in three

ways: (a) by sections, (b) by level curves on a graph, and (c) by level curves in the xy plane.