Math 254 - Vector Calculus Notes

12.1 Planes and Surfaces

What does a consistent equation in one variable represent in the number line? ... a point!

What does a consistent equation in two variables represent in the xy-plane? ... a curve!

What does a consistent equation in three variables represent in the xyz-space? ... a surface!

What does a consistent equation in n variables represent in n-dimensional space? ... some-
thing (n − 1)-dimensional!

A plane in R3 can be determined by a point in the plane and a direction normal (perpendic-
ular) to the plane.

Suppose A(x0 , y0 , z0 ) is a point in the plane P . Suppose n =< a, b, c > is a non-zero normal
vector to the plane. [DRAW IT!]
−→
Let B(x, y, z) represent a variable point in the plane. Then the vector AB is perpendicular
to n.

Thus:

< a, b, c > · < x − x0 , y − y0 , z − z0 >= 0 →

a(x − x0 ) + b(y − y0 ) + c(z − z0 ) = 0 →

ax + by + cz + d = 0,
where d = −ax0 − by0 − cz0 .
(Every such equation is called linear. Just like how a linear equation in two variables gives
you a line in the plane, a linear equation in three variables gives a plane in space.)

Conversely, given a plane’s equation, one can pick out the variable coefficients to determine
a normal (not unique) and plug-in for two free-variables to solve for a point in the plane.
Examples:

(1) Find an equation of the plane passing through (1, 2, 3) and normal to < 1, 0, 1 >.
(2) Find an equation of the plane passing through (2, 0, −5), (0, 1, 3) and (−3, −1, 7).
(3) Find an equation of the line which is the intersection of the planes 4x − z = 8 and
x + 3y − 2z = 6.
(4) Find an equation of the plane that contains (0, 1, −1) and the line (1+2t, −t, 4(1+t)).

(ELFY - ”Exercise Left For You”) Derive a formula for the distance from a plane to a point
not in the plane. Hint: Find a vector from a point in the plane to the point not in the plane
and project it onto a normal vector. The length of this vector will be the distance formula!

Two planes are parallel if their respective normal vectors are parallel. Two planes are
orthogonal if their respective normal vectors are orthogonal.

Indeed, one can calculate the angle between planes by calculating the angle between two
normals (and taking the complement if that angle is more than π).

Find the angle between x + 2y + 3z = 0 and 2y = z + 3.

A coordinate plane goes through the origin and is normal to i (yz-plane), j (xz-plane), or j
(xy-plane).

Given a curve C in a plane P and a line ` passing through C, but not in P , a cylinder
consists of all lines passing through C that parallel to `.

The classic (right, circular) cylinder radius R is the collection of all lines passing through
x2 + y 2 = R2 parallel to k̂

When one variable is missing from an equation representing a surface in xyz-space, it is a

cylinder given by taking all lines parallel to the axis of the missing variable going through
the curve in the coordinate plane of the variables present.

For example, the equation z = y 2 + 1 describes the cylinder given by the standard upward
parabola in the yz-plane (x = 0) moved along the x-axis.

What does 4x2 + y 2 = 16 correspond to as a cylinder?

A trace of a surface is the set of points on the surface at which the surface intersects a plane
parallel to a coordinate plane (Like x = 1 or y = 4). Hence, traces are found by plugging in
a constant for one of the variables.

Quadratic equations in three variables describe quadratic surfaces. We study a few types
and their standard equations:

(1) Ellipsoid:
x2 y 2 z 2
+ 2 + 2 = 1.
a2 b c
All traces are either empty, points, or ellipses. The extreme points that are furthest
from the origin are (±a, 0, 0), (0, ±b, 0), and (0, 0, ±c).

Here is the graph of x2 + 4y 2 + 16z 2 = 16.

4 y
2
0
-2
-4

0 z

-2

-4 -2 0 2 4

x
(2) Elliptic paraboloid:
x2 y 2
z= + 2 or any permutation of x, y, z.
a2 b

Traces with z set to a number are either empty, a point, or ellipses. Traces with
either x or y set to a number are always parabolas.

Here is the graph of z = x2 + y 2 .

z 4

2
0

-2
0
y
0

x -2
2
(3) Hyperbolic paraboloid:
x2 y 2
z= − 2 or any permutation of x, y, z.
a2 b

Traces with z set to a number are either hyperbolas or a pair of lines. Traces with
either x or y set to a number are always parabolas (but open in opposite z-directions.

Here is the graph of z = x2 − y 2 .

y
2
0
-2

z
0

-2

-4

x
-2
(4) Elliptic Cone:
z2 x2 y 2
= + 2 or any permutation of x, y, z.
c2 a2 b

Traces with z set to a number are ellipses or a point. Traces with x or y set to a
number are hyperbolas or a pair of lines.

Here is the graph of z 2 = 4x2 + y 2 .

-10

z 0

10
0 x

-10
-5
0 -5
5
y
10
(5) Hyperboloid of one sheet:
x2 y 2 z 2
+ 2 − 2 = 1 or any permutation of x, y, z.
a2 b c

Traces with z set to a number are ellipses. Traces with x or y set to a number are
hyperbolas or a pair of lines. The surface is made of one connected piece.

Here is the graph of z 2 + 4 = 4x2 + y 2 .

10

z 0

10

0 y
-10

-5

-5
0 -10
5
x
(6) Hyperboloid of two sheets:
x2 y 2 z 2
− − 2 + 2 = 1 or any permutation of x, y, z.
a2 b c

Traces with z set to a number are empty, points or ellipses. Traces with x or y set
to a number are hyperbolas or a pair of lines. The surface is made of two disconnected
pieces.

Here is the graph of z 2 − 4 = 4x2 + y 2 .

10

z 0

10

0 y
-10
-5

-5
0 -10
5
x
Of course replacing any variable by that variable plus/minus a positive constant shifts the
graph in the variable’s positive/negative direction by the constant. For example, replacement
of x by x − 2 moves the graph by 2î (2 units in the pos. x direction).

Classify and sketch:

(1)
x2 + 4x2 = 64 − 16z 2 .

(2)
9z 2 + 4x2 = 36y 2 .

(3)
−x2 = y 2 − z 2 + 4.

(4)
z + x2 = 2y 2 .

(5)
z 2 + 4 + 100x2 = y 2 .
12.2 - Graphs and Level Curves

A function f of two variables x and y often expressed with z set to the output.

z = f (x, y).

The domain is a subset D of the xy-plane at which f is defined. The range is R =

{f (x, y)|(x, y) ∈ D}, a subset of the real numbers.

Find the domain of the function:

f (x, y) = ln (9 − x2 − y 2 ).

The graph of z = f (x, y) is the set of points {(x, y, f (x, y))|(x, y) ∈ D}.

A surface represents a function whenever it passes the vertical line test: Any line parallel
to the z-axis intersects the surface at most once. So, z = 16 − x2 − y 2 (downward circular
2
paraboloid) is a function, but x2 + y 2 + z4 = 4 (ellipsoid) is not.

p
Graph the function f (x, y) = 1 + x2 + y 2 . Note the domain is all of R2 (xy-plane) and
the range is [1, ∞). How does it relate to a standard quadratic surface? It’s the top sheet of
a hyperboloid of two sheets...

6 -5

0
-5 0 5
Consider a surface defined by z = f (x, y). Imagine the intersection of the surface with a
plane parallel to the xy-plane. Algebraically this is done by setting z to a constant. This
curve in xyz-space is called a contour. Projected into the xy-plane, these curves are graphed
together and are called level curves.

Note that two level curves cannot intersect. If they do, then there is a point in the domain
D of the function f which corresponds to more than one z-value, which contradicts that f
is a function.

2 −y 2
Graph the level curves of f (x, y) = −x2 − y. Graph the level curves of g(x, y) = e−x
noting that the range of g is (0, 1].

Of course, one can talk about functions of n-variables. Consider y = f (x1 , x2 , ..., xn ). Then
the domain is in n-dimensional space, with a graph in n + 1 dimensional space.

“Level curves” would become level (n − 1)-dimensional subsets of n-dimensional space. For
example, describe the level curves of the function f (x, y, z) = x2 + y 2 − z.
12.3 - Limits and Continuity

A function f of two variables has a limit L as (x, y) approaches (x0 , y0 ) if |f (x, y) − L| can
be made arbitrarily small for all points sufficiently close to (x0 , y0 ).

More precisely: f (x, y) has a limit L as (x, y) approaches (x0 , y0 ) if for every ² > 0, there
exists δ > 0 such that 0 < | < x − x0 , y − y0 > | < δ implies |f (x, y) − L| < ².

Notation:
lim f (x, y) = L.
(x,y)→(0,0)

All the standard limit laws apply. The limit of a constant is the constant, the limit of a
sum/diff is the sum/diff of limits, the limit of a product is the product of the limits, etc...

Find the following limit:

x3 y − 5x2 − 25xy + 125

lim .
(x,y)→(5,0) x2 − 25

Of course, a limit might not exist (dne).

If a limit L does exist then approaching the point on any path (curve) in the domain will
yield the same limit. So, a common way to show a limit does not exist (dne) is to show that
different paths give different limits:

Let’s show that the following limit does not exist:

x2 y
lim .
(x,y)→(0,0) x4 + y 2
Let’s show that the following limit does not exist:

y2
lim .
(x,y)→(1,0) (x − 1)3 + y 2

Let’s show that the following limit does exist by using polar coordinates, where x = r cos θ,
y = r sin θ, and r → 0:

x2 y 2
lim .
(x,y)→(0,0) x2 + y 2

Let R be a region in R2 . An point P in R is an interior point of R if there exists a disk

centered at P that is contained inside R. A point Q is a boundary point of R if every disk
centered at the point Q contains at least 1 point from within R and from outside of R.

A region R is open if it contains none of the boundary points. A region R is closed if it

contains all the boundary points. A region that contains some, but not all boundary points
is neither.

A function f is continuous at (x0 , y0 ) if

f (x0 , y0 ) = lim f (x, y).

(x,y)→(x0 ,y0 )

So for f to be continuous at a point, f is defined at the point, the limit of f exists at the
point, and that limit equals the value of f at the point.

All of these definitions naturally extend to 3 and higher dimensions.

Find points where f (x, y, z) = ln (x2 + y 2 + z 2 − 1) is continuous.

This function will be continuous on its domain: x2 + y 2 + z 2 > 1.

12.4 Partial Derivatives

The Partial derivative of a function f (x1 , x2 , ..., xn ) with respect to xm is given by:

∂f f (x1 , ..., xm + h, ..., xn ) − f (x1 , ..., xm , ..., xn )

= lim .
∂xm h→0 h

The above partial derivative is also given the subscript notation fxm .

It is the derivative of f where all variables except xm are held fixed during the differentiation.
Consequently, all the rules of differential calculus apply with all other variables treated as
constants.

Let f (x, y, z) = x2 yeyz . Find fx , fy , and fz .

Higher-order partial derivatives can also be taken with respect to multiple variables. The
order corresponds to how many derivatives have been taken.

For example:

µ ¶
∂ ∂f ∂ 2f
= = fxx ,
∂x ∂x ∂x2

and

µ ¶
∂ ∂f ∂2f
= = fyx ,
∂x ∂y ∂x∂y

are second-order, and

µ µ ¶¶
∂ ∂ ∂f ∂ 3f
= = fyxx ,
∂x ∂x ∂y ∂x2 ∂y

is third-order.

Higher-order partial derivatives involving more than one variable are called mixed partial derivatives.

Find all first-order and second-order derivatives of f (x, y) = x sin (xy).

Theorem (Equality of Mixed Partials): Let f (x, y) be a function and assume the mixed
partials fxy and fyx are continuous. Then fxy = fyx .

The proof is skipped.

Careful: The existence of partial derivatives are not sufficient to declare that a function is
differentiable!

The function z = f (x, y) is differentiable at (x0 , y0 ) provided fx (x0 , y0 ) and fy (x0 , y0 ) exist
and that
∆z = fx (x0 , y0 )∆x + fy (x0 , y0 )∆y + ²1 ∆x + ²2 ∆y,
where ²1 and ²2 both approach 0 as ∆x and ∆y approach 0.

This theorem makes checking differentiability easier most of the time.

Theorem: If a function f (x, y) has partial derivatives defined on an open region containing
(x0 , y0 ) with both fx and fy continuous at (x0 , y0 ), then f is differentiable at (x0 , y0 ).

We do not devote the time to prove it, but the proof is given in Appendix B of the textbook.

Theorem: (Differentiability Implies Continuity) If a function f (x, y) is differentiable at

(x0 , y0 ) then it is continuous at (x0 , y0 ).

Proof: Assume f is differentiable at (x0 , y0 ). Then fx (x0 , y0 ) and fy (x0 , y0 ) exist, and
∆z = fx (x0 , y0 )∆x + fy (x0 , y0 )∆y + ²1 ∆x + ²2 ∆y,
where ²1 and ²2 both approach 0 as ∆x and ∆y approach 0.

On one hand,

lim ∆z = lim (fx (x0 , y0 )∆x + fy (x0 , y0 )∆y + ²1 ∆x + ²2 ∆y) = 0.

(x,y)→(x0 ,y0 ) (∆x,∆y)→(0,0)

On the other hand,

lim ∆z = lim (f (x0 + ∆x, y0 + ∆y) − f (x0 , y0 )) →

(x,y)→(x0 ,y0 ) (∆x,∆y)→(0,0)

lim f (x, y) = f (x0 , y0 ) and so we are done ¤

(x,y)→(x0 ,y0 )
Let’s consider the function below:

½ 2xy
x2 +y 2
if (x, y) 6= (0, 0)
f (x, y) =
0 if (x, y) = (0, 0)

Consider the following partial derivatives:

f (0 + h, 0) − f (0, 0) f (0, 0 + h) − f (0, 0)
fx (0, 0) = lim = 0 and fy (0, 0) = lim = 0.
h→0 h h→0 h

However, this function is not differentiable at (0, 0) as it isn’t even continuous at (0, 0). Just
consider the limit as (x, y) → (0, 0).

If the path (x, x) is taken, the function approaches 1 as x → 0. However, if the path (x, 0)
is taken, the function approaches 0 as x → 0.

Hence the limit doesn’t exists and the function is not continuous at (0, 0).
12.5 Chain Rule

Case 1: z is a function of x and y, which are each functions of t. Then we have

dz ∂z dx ∂z dy
= + .
dt ∂x dt ∂y dt

One can draw a tree diagram: z is at the top, two edges go down to x and y, from each an
additional edge goes down for t.

One can also apply this to functions of 3 (or more) intermediate variables that are all
functions of a single common variable. For example, suppose w is a function of x, y and z,
which are each functions of t. Then we have
dw ∂w dx ∂w dy ∂w dz
= + + .
dt ∂x dt ∂y dt ∂z dt

Example: Let w = xy tan−1 (z), x = t2 , y = 2t4 and z = 1t . Calculate dw

dt
in two ways:

First use the chain rule (by substitution express everything in terms of t at the end).

Second use direct substitution to determine w as a function of t explicitly. Then find the
derivative of w with respect to t directly.

Case 2: z is a function of x and y, which are each functions of s and t. Then we have
∂z ∂z ∂x ∂z ∂y ∂z ∂z ∂x ∂z ∂y
= + and = + .
∂s ∂x ∂s ∂y ∂s ∂t ∂x ∂t ∂y ∂t

One can draw a tree diagram: z is at the top, two edges go down to x and y, from each an
additional two edges go down for s and t.

One can also apply this to functions of 3 (or more) intermediate variables that are all
functions of two variables. For example, suppose w is a function of x, y and z, which are
each functions of s and t. Then we have
∂w ∂w ∂x ∂w ∂y ∂w ∂z ∂w ∂w ∂x ∂w ∂y ∂w ∂z
= + + and = + + .
∂s ∂x ∂s ∂y ∂s ∂z ∂s ∂t ∂x ∂t ∂y ∂t ∂z ∂t

p
Example: Let z = x x2 + y 2 , x = st, y = s + t. Using the chain rule (two intermediate
variables), find zs and zt .
One can of course generalize. By using a tree diagram, a chain rule can be figured out for
any situation!

Implicit differentiation: Consider the situation, when a function y = f (x) is implicitly defined
by F (x, y) = C, where C is a constant.

By differentiating both sides with respect to x, one arrives at

dy dy Fx
Fx + Fy =0→ =− .
dx dx Fy

dy
Consider the equation x2 sin y = y. Find dx
. Hint: Set F (x, y) = x2 sin y − y and use the
formula!

What if F (x, y, z) = k implicitly defines z = f (x, y)? In that case, fx = − FFxz and fy = − FFyz .

Consider the equation x2 y + xz + yz 2 = 1. Find zx , zy . Hint: Set F (x, y, z) = x2 y + xz + yz 2

and use the formulas!
12.6 Directional Derivatives and the Gradient

Suppose z = f (x, y) and you want to know how z changes at the point (x0 , y0 ) with respect
to a direction (within the domain) given by a unit vector û = (a, b) (so a2 + b2 = 1). Then
the derivative of f with respect to the direction given by û is given by

f (x0 + ha, y0 + hb) − f (x0 , y0 )

Dû f (x0 , y0 ) = lim .
h→0 h

A nice formula for directional derivatives:

Consider the function g(s) = f (x0 + as, y0 + bs). Then,

g(0 + h) − g(0)
g 0 (0) = lim = Dû f (x0 , y0 ).
h→0 h

And by the chain rule:

g 0 (0) = fx (x0 , y0 ) · a + fy (x0 , y0 ) · b.

Definition: The gradient of f (x1 , x2 , ..., xn ) is given by

∇f =< fx1 , fx2 , ..., fxn > .

So, using that notation,

Dû f (x0 , y0 ) = ∇f (x0 , y0 ) · (a, b).

So the directional derivative is just the dot product of the gradient and a unit vector!

All of this generalizes. The directional derivative of f (x1 , x2 , ..., xn ) at (a1 , a2 , ..., an ) in the
direction of the n-dimensional unit vector û is given by
Dû f = ∇f (a1 , a2 , ..., an ) · û.
Let f (x, y) = x3 y 2 + tan−1 (xy). Find the directional derivative of f (x, y) at the point (1, 3)
in the direction of the vector < 3, −4 >.

For what direction is Dû f (x0 , y0 ) at a maximum? By the dot product identity,
Dû f (x0 , y0 ) = |∇f (x0 , y0 )||û| cos θ = |∇f (x0 , y0 )| cos θ.

So it is at a maximum when θ = 0. That occurs when û is in the direction of the gradient

∇f (x0 , y0 ). Moreover, this maximum directional derivative is equal to the magnitude of the
gradient.

So the gradient points in the direction of maximum increase (negative gradient points in the
direction of maximum decrease) and its length is that rate of increase (or decrease).

Find the direction of maximum increase for the function f (x, y, z) = xy 2 z 3 at the point
(2, −2, 1). What is the maximum rate of increase?

Also important: In a direction orthogonal to the gradient, the directional derivative is zero.
Such a direction is parallel to a level set of the function (because the function does not
change in that direction).

Find a direction from the point (1, 2) in which the function f (x, y) = 4x2 − 9y 2 does not
change.
12.7 Tangent Planes and Linear Approximation

Consider the surface in 3-dimensions defined implicitly by F (x, y, z) = C (a constant), which

is a level set of F .

Note an explicit function z = f (x, y) can always be defined implicitly by having F (x, y, z) =
z − f (x, y).

Since the gradient is orthogonal to the level set, it can be used as a normal vector to give
the tangent plane to the surface at the point.

The equation of tangent plane to the surface F (x, y, z) = C at the point (x0 , y0 , z0 ) is given
by

∇F (x0 , y0 , z0 )· < x − x0 , y − y0 , z − z0 >= 0 →

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

In the case where z = f (x, y) and z0 = f (x0 , y0 ) this becomes

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

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

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

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

This last form also gives what’s called the linear approximation to z = f (x, y) at (x0 , y0 ).

In the language of differentials, the differential of z is dz = fx (x0 , y0 )dx + fy (x0 , y0 )dy.

Below is the graph of f (x, y) = 100 − x2 − y 4 and its tangent plane at (3, 1, 90). Try to find
an equation for the plane!

-5

y 0

5
100

z 50

0
10 5 0 -5 -10
x
EXAMPLES:
(1) Find an equation for the tangent plane to graph of the function f (x, y) = x2 y when
x = −2 and y = 3.
(2) Find an equation for the tangent line to the surface given by x2 + y 2 + 1 = z 2 at the
point (2, −2, 3).
p
(3) Let fp(x, y) = x x2 + y 2 . Use the linear approximation of f at (5, 12) to estimate
(5.1) (5.1)2 + (11.9)2 .
0
(4) Recall: ff represents the percent rate of change in f . Use differentials to find a
formula the percent rate of change in the area of an ellipse with major and minor
axis of lengths 2a and 2b respectively. Note: A = πab.

dA πbda + πadb da db
= = + .
A πab a b
If a increases by 5 percent and b decreases by 2 percent, by what percent does the
area increase or decrease?
12.8 Max/Min Problems

f (x, y) has a local maximum/minimum at (x0 , y0 ) if f (x0 , y0 ) is the largest/smallest value

in an open disk about (x0 , y0 ).

(x0 , y0 ) is a critical point of f (x, y) if fx (x0 , y0 ) = fy (x0 , y0 ) = 0 or one of the partial

derivatives does not exist.

Find the critical points of f (x, y) = x4 + y 4 − 16xy.

(x0 , y0 ) is a saddle point of f (x, y) if every disk about (x0 , y0 ) contains values smaller and
larger than f (x0 , y0 ).

Second Derivative Test: (proof omitted)

Suppose the second partial derivatives of f (x, y) are continuous throughout an open disk
about a critical point (x0 , y0 ). Let D(x, y) = fxx fyy − (fxy )2 . Then

(1) If D(x0 , y0 ) > 0 and fxx (x0 , y0 ) > 0, then f has a local minimum at (x0 , y0 ).
(2) If D(x0 , y0 ) > 0 and fxx (x0 , y0 ) < 0, then f has a local maximum at (x0 , y0 ).
(3) If D(x0 , y0 ) < 0, then f has a saddle at (x0 , y0 ).
(4) If D(x0 , y0 ) = 0, then the test is inconclusive.

Note: D above is sometimes called the discriminant, and can also be defined as the deter-
minant of the second derivative matrix.

Classify all the critical points of f (x, y) = x2 y 2 − xy 2 + 2x2 y − 2xy + 5.

Classify all the critical points of f (x, y) = tan−1 (xy).

Classify all the critical points of f (x, y) = x2 + xy 2 − 2x + 1.

x
Classify all the critical points of f (x, y) = x2 +y 2 +1
.
f (x, y) has an absolute maximum/minimum at (x0 , y0 ) if f (x0 , y0 ) is the largest/smallest
value in the range.

If f is defined on a closed and bounded set:

Find the values of f at all critical points in the interior. Find the maximum and mini-
mum value of f on the boundary. The largest/smallest values correspond to the absolute
maximum/minimum.

Open and unbounded domains don’t allow for a systematic method...

2 −y 2
Find the absolute extremes of f (x, y) = e−x on {(x, y)|x2 + y 2 ≤ 1}. What if the
inequality was strict?

A lidless box is to be made out of 4 square meters of cardboard. Find the dimensions of the
box with the largest possible volume.

Find the point on the plane 2x + y − z = 1 that is nearest to the origin.

12.9 Lagrange Multipliers

Let f be a differentiable function in a region of R2 (xy-plane) that contains the “smooth”

curve C given by g(x, y) = k where k is a constant and ∇g(x, y) 6= 0 for all points on C.

Let’s try to derive a method to find the local extremes values of f restricted to C, called
a constraint. Let C be parameterized by r(t) =< x(t), y(t) > where x(t) and y(t) are
differentiable. Then let h(t) = f (x(t), y(t)), which is differentiable. We want the extreme
values of h. By the chain rule:

h0 (t) = fx (x(t), y(t))x0 (t) + fy (x(t), y(t))y 0 (t).

h has critical points where 0 = fx x0 (t) + fy y 0 (t). This could be rewritten as

0 = ∇f (x(t), y(t)) · r0 (t).

So the extreme values of f restricted to C occur where the gradient of f is orthogonal to the
curve C (as r0 (t) is parallel to C).

So from this we get the method of Lagrange multipliers:

To find the min./max. value of a differentiable function f (x, y) restricted to a “smooth”

curve C given by g(x, y) = k where ∇g(x, y) 6= 0 for all points on C, simply solve the system
given by the equations:

∇f (x, y) = λ∇g(x, y) and g(x, y) = k.

This generalizes in an obvious way to a differentiable function f (x, y, z) restricted to a “nice”

surface S given by g(x, y, z) = k where ∇g(x, y, z) 6= 0 on S. In general, one can apply the
method to functions of n variables restricted to (n − 1)-dimensional “hypersurfaces.”

Recall: ∇g(x, y) evaluated at a point on C will be orthogonal to C as well, so at an extreme

of f restricted to C, there is a scalar λ such that ∇f (x, y) = λ∇g(x, y).
This picture gives the general idea. The curve C, defined by g(x, y) = k with ∇g(x, y) 6= 0,
is drawn among several level curves of a differentiable function f (x, y). From the picture we
can visually verify the method of Lagrange multipliers yields a min. value of 4 and a max.
value of 8 at points where the gradients of f and g are parallel:
EXAMPLES:

• Find the min. and max. values of f (x, y) = xy 2 subject to x2 + y 2 = 1.

• Find the min. and max. values of f (x, y) = xy + x + y subject to x2 y 4 = 16.

• Find the min. and max. values of f (x, y) = x2 + y 2 − z subject to z = 2x2 y 2 + 1.

• Find the dimensions (radius, height) of a right circular cylinder of max. volume that can be
inscribed in a sphere of radius 4. Hint: The radius of the sphere, the radius of the cylinder,
and half the height of the cylinder form a right triangle...

• Find the point(s) on the cone z 2 = x2 +y 2 that are closest to (1, 2, 0). What is that minimum
distance? Hint: Minimize the square distance subject to the cone as the constraint.

If we are looking to maximize f (x, y, z) subject to two constraints, g(x, y, z) = k1 and

h(x, y, z) = k2 , we solve the following Lagrange multiplier system (skipping the justification):

∇f (x, y, z) = λ∇g(x, y, z) + µ∇h(x, y, z), g(x, y, z) = k1 , and h(x, y, z) = k2 .

Use the method of Lagrange multipliers to find the point closest to the origin that lies on L,
the line of intersection of the planes x + 2z = 12 and x + y = 6. Can you verify the result
using methods described in chapter 11?