Вы находитесь на странице: 1из 31

THE METHOD OF

LAGRANGE MULTIPLIERS
William F. Trench
Professor Emeritus
Department of Mathematics
Trinity University
San Antonio, Texas, USA
wtrench@trinity.edu
Copyright November 2012, WilliamF. Trench
This is a supplement to the authors
Introduction to Real Analysis
Reproduction and online posting are permitted for any valid noncommercial edu-
cational, mathematical, or scientic purpose. However, charges for prot beyond
reasonable printing costs are strictly prohibited. Acomplete instructors solutions
manual is available on request by email to the author, subject to verication of the
requestors faculty status.
THE METHOD OF LAGRANGE MULTIPLIERS
WilliamF. Trench
1 Foreword
This is a revised and extended version of Section 6.5 of my Advanced Calculus (Harper
& Row, 1978). It is a supplement to my textbook Introduction to Real Analysis, which
is referenced via hypertext links.
2 Introduction
To avoid repetition, it is to be understood throughout that and g
1
, g
2
,. . . , g
m
are
continuously differentiable on an open set D in R
n
.
Suppose that m < n and
g
1
(X) = g
2
(X) = = g
m
(X) = 0 (1)
on a nonempty subset D
1
of D. If X
0
D
1
and there is a neighborhood N of X
0
such
that
(X) _ (X
0
) (2)
for every X in ND
1
, then X
0
is a local maximumpoint of subject to the constraints
(1). However, we will usually say subject to rather than subject to the constraint(s).
If (2) is replaced by
(X) _ (X
0
). (3)
then maximum is replaced by minimum. A local maximum or minimum of
subject to (1) is also called a local extreme point of subject to (1). More briey, we
also speak of constrained local maximum, minimum, or extreme points. If (2) or (3)
holds for all X in D
1
, we omit local.
Recall that X
0
= (.
10
. .
20
. . . . . .
n0
) is a critical point of a differentiable function
1 = 1(.
1
. .
2
. . . . . .
n
) if
1
x
i
(.
10
. .
20
. . . . . .
n0
) = 0. 1 _ i _ n.
Therefore, every local extreme point of 1 is a critical point of 1; however, a critical
point of 1 is not necessarily a local extreme point of 1 (pp. 334-5).
Suppose that the system (1) of simultaneous equations can be solved for .
1
, . . . ,
.
m
in terms of the .
mC1
, . . . , .
n
; thus,
.
j
= h
j
(.
mC1
. . . . . .
n
). 1 _ _ m. (4)
Then a constrained extreme value of is an unconstrained extreme value of
(h
1
(.
mC1
. . . . . .
n
). . . . . h
m
(.
mC1
. . . . . .
n
). .
mC1
. . . . . .
n
). (5)
2
However, it may be difcult or impossible to nd explicit formulas for h
1
, h
2
, . . . , h
m
,
and, even if it is possible, the composite function (5) is almost always complicated.
Fortunately, there is a better way to to nd constrained extrema, which also requires
the solvability assumption, but does not require an explicit formulas as indicated in (4).
It is based on the following theorem. Since the proof is complicated, we consider two
special cases rst.
Theorem 1 Suppose that n > m. If X
0
is a local extreme point of subject to
g
1
(X) = g
2
(X) = = g
m
(X) = 0
and

dg
1
(X
0
)
d.
r
1
dg
1
(X
0
)
d.
r
2

dg
1
(X
0
)
d.
rm
dg
2
(X
0
)
d.
r
1
dg
2
(X
0
)
d.
r
2

dg
m
(X
0
)
d.
rm
.
.
.
.
.
.
.
.
.
.
.
.
dg
m
(X
0
)
d.
r
1
dg
m
(X
0
)
d.
r
2

dg
m
(X
0
)
d.
rm

= 0 (6)
for at least one choice of r
1
< r
2
< < r
m
in {1. 2. . . . . n]. then there are constants
z
1
. z
2
. . . . . z
m
such that X
0
is a critical point of
z
1
g
1
z
2
g
2
z
m
g
m
:
that is.
d(X
0
)
d.
i
z
1
dg
1
(X
0
)
d.
i
z
2
dg
2
(X
0
)
d.
i
z
m
dg
m
(X
0
)
d.
i
= 0.
1 _ i _ n.
The following implementation of this theorem is the method of Lagrange
multipliers.
(a) Find the critical points of
z
1
g
1
z
2
g
2
z
m
g
m
.
treating z
1
, z
2
, . . . z
m
as unspecied constants.
(b) Find z
1
, z
2
, . . . , z
m
so that the critical points obtained in (a) satisfy the con-
straints.
(c) Determine which of the critical points are constrained extreme points of . This
can usually be done by physical or intuitive arguments.
If a and b
1
, b
2
, . . . , b
m
are nonzero constants and c is an arbitrary constant, then the
local extreme points of subject to g
1
= g
2
= = g
m
= 0 are the same as the local
extreme points of a c subject to b
1
g
1
= b
2
g
2
= = b
m
g
m
= 0. Therefore, we
can replace z
1
g
1
z
2
g
2
z
m
g
m
by a z
1
b
1
g
1
z
2
b
2
g
2
z
m
b
m
g
m
c
to simplify computations. (Usually, the c indicates dropping additive constants.)
We will denote the nal form by 1 (for Lagrangian).
3
3 Extrema subject to one constraint
Here is Theorem 1 with m = 1.
Theorem 2 Suppose that n > 1. If X
0
is a local extreme point of subject to g(X) =
0 and g
xr
(X
0
) = 0 for some r {1. 2. . . . . n]. then there is a constant z such that

x
i
(X
0
) zg
x
i
(X
0
) = 0. (7)
1 _ i _ n: thus. X
0
is a critical point of zg.
Proof For notational convenience, let r = 1 and denote
U = (.
2
. .
3
. . . . .
n
) and U
0
= (.
20
. .
30
. . . . .
n0
).
Since g
x
1
(X
0
) = 0, the Implicit Function Theorem (Corollary 6.4.2, p. 423) implies
that there is a unique continuously differentiable function h = h(U). dened on a
neighborhood N R
n1
of U
0
. such that (h(U). U) D for all U N, h(U
0
) = .
10
,
and
g(h(U). U) = 0. U N. (8)
Now dene
z =

x
1
(X
0
)
g
x
1
(X
0
)
. (9)
which is permissible, since g
x
1
(X
0
) = 0. This implies (7) with i = 1. If i > 1,
differentiating (8) with respect to .
i
yields
dg(h(U). U)
d.
i

dg(h(U). U)
d.
1
dh(U)
d.
i
= 0. U N. (10)
Also,
d(h(U). U))
d.
i
=
d(h(U). U)
d.
i

d(h(U). U)
d.
1
dh(U)
d.
i
. U N. (11)
Since (h(U
0
). U
0
) = X
0
, (10) implies that
dg(X
0
)
d.
i

dg(X
0
)
d.
1
dh(U
0
)
d.
i
= 0. (12)
If X
0
is a local extreme point of subject to g(X) = 0, then U
0
is an unconstrained
local extreme point of (h(U). U); therefore, (11) implies that
d(X
0
)
d.
i

d(X
0
)
d.
1
dh(U
0
)
d.
i
= 0. (13)
Since a linear homogeneous system

a b
c J

0
0

4
has a nontrivial solution if and only if

a b
c J

= 0.
(Theorem 6.1.15, p. 376), (12) and (13) imply that

d(X
0
)
d.
i
d(X
0
)
d.
1
dg(X
0
)
d.
i
dg(X
0
)
d.
1

= 0. so

d(X
0
)
d.
i
dg(X
0
)
d.
i
d(X
0
)
d.
1
dg(X
0
)
d.
1

= 0.
since the determinants of a matrix and its transpose are equal. Therefore, the system
2
6
6
6
4
d(X
0
)
d.
i
dg(X
0
)
d.
i
d(X
0
)
d.
1
dg(X
0
)
d.
1
3
7
7
7
5

0
0

has a nontrivial solution (Theorem 6.1.15, p. 376). Since g


x
1
(X
0
) = 0, u must be
nonzero in a nontrivial solution. Hence, we may assume that u = 1, so
2
6
6
6
4
d(X
0
)
d.
i
dg(X
0
)
d.
i
d(X
0
)
d.
1
dg(X
0
)
d.
1
3
7
7
7
5

0
0

. (14)
In particular,
d(X
0
)
d.
1

dg(X
0
)
d.
1
= 0. so =

x
1
(X
0
)
g
x
1
(X
0
)
.
Now (9) implies that = z, and (14) becomes
2
6
6
6
4
d(X
0
)
d.
i
dg(X
0
)
d.
i
d(X
0
)
d.
1
dg(X
0
)
d.
1
3
7
7
7
5

1
z

0
0

.
Computing the topmost entry of the vector on the left yields (7).
Example 1 Find the point (.
0
. ,
0
) on the line
a. b, = J
closest to a given point (.
1
. ,
1
).
5
SolutionWe must minimize
p
(. .
1
)
2
(, ,
1
)
2
subject to the constraint. This
is equivalent to minimizing (. .
1
)
2
(, ,
1
)
2
subject to the constraint, which is
simpler. For, this we could let
1 = (. .
1
)
2
(, ,
1
)
2
z(a. b, J):
however,
1 =
(. .
1
)
2
(, ,
1
)
2
2
z(a. b,)
is better. Since
1
x
= . .
1
za and 1
y
= , ,
1
zb.
(.
0
. ,
0
) = (.
1
za. ,
1
zb), where we must choose z so that a.
0
b,
0
= J.
Therefore,
a.
0
b,
0
= a.
1
b,
1
z(a
2
b
2
) = J.
so
z =
J a.
1
b,
1
a
2
b
2
.
.
0
= .
1

(J a.
1
b,
1
)a
a
2
b
2
. and ,
0
= ,
1

(J a.
1
b,
1
)b
a
2
b
2
.
The distance from (.
1
. ,
1
) to the line is
p
(.
0
.
1
)
2
(,
0
,
1
)
2
=
[J a.
1
b,
1
[
_
a
2
b
2
.
Example 2 Find the extreme values of (.. ,) = 2. , subject to
.
2
,
2
= 4.
Solution Let
1 = 2. ,
z
2
(.
2
,
2
):
then
1
x
= 2 z. and 1
y
= 1 z,.
so (.
0
. ,
0
) = (2,z. 1,z). Since .
2
0
,
2
0
= 4, z =
_
5,2. Hence, the constrained
maximum is 2
_
5, attained at (4,
_
5. 2,
_
5), and the constrained minimum is 2
_
5,
attained at (4,
_
5. 2,
_
5).
Example 3 Find the point in the plane
3. 4, z = 1 (15)
closest to (1. 1. 1).
SolutionWe must minimize
(.. ,. z) = (. 1)
2
(, 1)
2
(z 1)
2
6
subject to (15). Let
1 =
(. 1)
2
(, 1)
2
(z 1)
2
2
z(3. 4, z):
then
1
x
= . 1 3z. 1
y
= , 1 4z. and 1

= z 1 z.
so
.
0
= 1 3z. ,
0
= 1 4z. z
0
= 1 z.
From (15),
3(1 3z) 4(1 4z) (1 z) 1 = 1 26z = 0.
so z = 1,26 and
(.
0
. ,
0
. z
0
) =

29
26
.
22
26
.
25
26

.
The distance from (.
0
. ,
0
. z
0
) to (1. 1. 1) is
p
(.
0
1)
2
(,
0
1)
2
(z
0
1)
2
=
1
_
26
.
Example 4 Assume that n _ 2 and .
i
_ 0, 1 _ i _ n.
(a) Find the extreme values of
n
X
i D1
.
i
subject to
n
X
i D1
.
2
i
= 1.
(b) Find the minimum value of
n
X
i D1
.
2
i
subject to
n
X
i D1
.
i
= 1.
Solution(a) Let
1 =
n
X
i D1
.
i

z
2
n
X
i D1
.
2
i
:
then
1
x
i
= 1 z.
i
. so .
i 0
=
1
z
. 1 _ i _ n.
Hence,
n
X
i D1
.
2
i 0
= n,z
2
, so z =
_
n and
(.
10
. .
20
. . . . . .
n0
) =

1
_
n
.
1
_
n
. . . . .
1
_
n

.
Therefore, the constrained maximum is
_
n and the constrained minimum is
_
n.
Solution(b) Let
1 =
1
2
n
X
i D1
.
2
i
z
n
X
i D1
.
i
:
7
then
1
x
i
= .
i
z. so .
i 0
= z. 1 _ i _ n.
Hence,
n
X
i D1
.
i 0
= nz = 1, so .
i 0
= z = 1,n and the constrained minimum is
n
X
i D1
.
2
i 0
=
1
n
There is no constrained maximum. (Why?)
Example 5 Show that
.
1=p
,
1=q
_
.


,
q
. .. , _ 0.
if
1

1
q
= 1. > 0. and q > 0. (16)
SolutionWe rst nd the maximum of
(.. ,) = .
1=p
,
1=q
subject to
.


,
q
= o. . _ 0. , _ 0. (17)
where o is a xed but arbitrary positive number. Since is continuous, it must assume
a maximum at some point (.
0
. ,
0
) on the line segment (17), and (.
0
. ,
0
) cannot be an
endpoint of the segment, since (o. 0) = (0. qo) = 0. Therefore, (.
0
. ,
0
) is in the
open rst quadrant.
Let
1 = .
1=p
,
1=q
z


,
q

.
Then
1
x
=
1
.
(.. ,)
z

and 1
y
=
1
q,
(.. ,)
z
q
= 0.
so .
0
= ,
0
= (.
0
. ,
0
),z. Now(16) and (17) imply that .
0
= ,
0
= o. Therefore,
(.. ,) _ (o. o) = o
1=p
o
1=q
= o =
.

,
q
.
This can be generalized (Exercise 53). It can also be used to generalize Schwarzs
inequality (Exercise 54).
8
4 Constrained Extrema of Quadratic Forms
In this section it is convenient to write
X =
2
6
6
6
4
.
1
.
2
.
.
.
.
n
3
7
7
7
5
.
An eigenvalue of a square matrix A = a
ij
|
n
i;jD1
is a number z such that the system
AX = zX.
or, equivalently,
(A zI)X = 0.
has a solution X = 0. Such a solution is called an eigenvector of A. You probably
know from linear algebra that z is an eigenvalue of A if and only if
det(A zI) = 0.
Henceforth we assume that A is symmetric (a
ij
= a
ji
. 1 _ i. _ n). In this case,
det(A zI) = (1)
n
(z z
1
)(z z
2
) (z z
n
).
where z
1
. z
2
. . . . . z
n
are real numbers.
The function
Q(X) =
n
X
i;jD1
a
ij
.
i
.
j
is a quadratic form. To nd its maximum or minimum subject to
n
X
i D1
.
2
i
= 1, we form
the Lagrangian
1 = Q(X) z
n
X
i D1
.
2
i
.
Then
1
x
i
= 2
n
X
jD1
a
ij
.
j
2z.
i
= 0. 1 _ i _ n.
so
n
X
jD1
a
ij
.
j0
= z.
i 0
. 1 _ i _ n.
Therefore, X
0
is a constrained critical point of Q subject to
n
X
i D1
.
2
i
= 1 if and only
if AX
0
= zX
0
for some z; that is, if and only if z is an eigenvalue and X
0
is an
9
associated unit eigenvector of A. If AX
0
= X
0
and
n
X
i
.
2
i 0
= 1, then
Q(X
0
) =
n
X
i D1
0
@
n
X
jD1
a
ij
.
j0
1
A
.
i 0
=
n
X
i D1
(z.
i 0
).
i 0
= z
n
X
i D1
.
2
i 0
= z:
therefore, the largest and smallest eigenvalues of A are the maximum and minimum
values of Q subject to
n
X
i D1
.
2
i
= 1.
Example 6 Find the maximum and minimum values
Q(X) = .
2
,
2
2z
2
2., 4.z 4,z
subject to the constraint
.
2
,
2
z
2
= 1. (18)
SolutionThe matrix of Q is
A =
2
4
1 1 2
1 1 2
2 2 2
3
5
and
det(A zI) =

1 z 1 2
1 1 z 2
2 2 2 z

= (z 2)(z 2)(z 4).


so
z
1
= 4. z
2
= 2. z
3
= 2
are the eigenvalues of A. Hence, z
1
= 4 and z
3
= 2 are the maximum and minimum
values of Q subject to (18).
To nd the points (.
1
. ,
1
. z
1
) where Q attains its constrained maximum, we rst
nd an eigenvector of A corresponding to z
1
= 4. To do this, we nd a nontrivial
solution of the system
(A 4I)
2
4
.
1
,
1
z
1
3
5
=
2
4
3 1 2
1 3 2
2 2 2
3
5
2
4
.
1
,
1
z
1
3
5
=
2
4
0
0
0
3
5
.
10
All such solutions are multiples of
2
4
1
1
2
3
5
. Normalizing this to satisfy (18) yields
X
1
=
1
_
6
2
4
.
1
,
1
z
1
3
5
=
2
4
1
1
1
3
5
.
To nd the points (.
3
. ,
3
. z
3
) where Q attains its constrained minimum, we rst
nd an eigenvector of A corresponding to z
3
= 2. To do this, we nd a nontrivial
solution of the system
(A 2I)
2
4
.
3
,
3
z
3
3
5
=
2
4
3 1 2
1 3 2
2 2 4
3
5
2
4
.
3
,
3
z
3
3
5
=
2
4
0
0
0
3
5
.
All such solutions are multiples of
2
4
1
1
1
3
5
. Normalizing this to satisfy (18) yields
X
3
=
2
4
.
2
,
2
z
2
3
5
=
1
_
3
2
4
1
1
1
3
5
.
As for the eigenvalue z
2
= 2, we leave it you to verify that the only unit vectors
that satisfy AX
2
= 2X
2
are
X
2
=
1
_
2
2
4
1
1
1
3
5
.
For more on this subject, see Theorem 4.
5 Extrema subject to two constraints
Here is Theorem 1 with m = 2.
Theorem 3 Suppose that n > 2. If X
0
is a local extreme point of subject to g
1
(X) =
g
2
(X) = 0 and

dg
1
(X
0
)
d.
r
dg
1
(X
0
)
d.
s
dg
2
(X
0
)
d.
r
dg
2
(X
0
)
d.
s

= 0 (19)
for some r and s in {1. 2. . . . . n]. then there are constants z and j such that
d(X
0
)
d.
i
z
dg
1
(X
0
)
d.
i
j
dg
2
(X
0
)
d.
i
= 0. (20)
1 _ i _ n.
11
Proof For notational convenience, let r = 1 and s = 2. Denote
U = (.
3
. .
4
. . . . .
n
) and U
0
= (.
30
. .
30
. . . . .
n0
).
Since

dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2

= 0. (21)
the Implicit Function Theorem (Theorem 6.4.1, p. 420) implies that there are unique
continuously differentiable functions
h
1
= h
1
(.
3
. .
4
. . . . . .
n
) and h
2
= h
1
(.
3
. .
4
. . . . . .
n
).
dened on a neighborhood N R
n2
of U
0
. such that (h
1
(U). h
2
(U). U) D for all
U N, h
1
(U
0
) = .
10
, h
2
(U
0
) = .
20
, and
g
1
(h
1
(U). h
2
(U). U) = g
2
(h
1
(U). h
2
(U). U) = 0. U N. (22)
From (21), the system
2
6
6
6
4
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2
3
7
7
7
5

z
j


x
1
(X
0
)

x
2
(X
0
)

(23)
has a unique solution (Theorem 6.1.13, p. 373). This implies (20) with i = 1 and
i = 2. If 3 _ i _ n, then differentiating (22) with respect to .
i
and recalling that
(h
1
(U
0
). h
2
(U
0
). U
0
) = X
0
yields
dg
1
(X
0
)
d.
i

dg
1
(X
0
)
d.
1
dh
1
(U
0
)
d.
i

dg
1
(X
0
)
d.
2
dh
2
(U
0
)
d.
i
= 0
and
dg
2
(X
0
)
d.
i

dg
2
(X
0
)
d.
1
dh
1
(U
0
)
d.
i

dg
2
(X
0
)
d.
2
dh
2
(U
0
)
d.
i
= 0.
If X
0
is a local extreme point of subject to g
1
(X) = g
2
(X) = 0, then U
0
is an
unconstrained local extreme point of (h
1
(U). h
2
(U). U); therefore,
d(X
0
)
d.
i

d(X
0
)
d.
1
dh
1
(U
0
)
d.
i

d(X
0
)
d.
2
dh
2
(U
0
)
d.
i
= 0.
The last three equations imply that

d(X
0
)
d.
i
d(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
i
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
i
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2

= 0.
12

d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2

= 0.
Therefore, there are constants c
1
, c
2
, c
3
, not all zero, such that
2
6
6
6
6
6
6
6
6
6
6
6
4
d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2
3
7
7
7
7
7
7
7
7
7
7
7
5
2
4
c
1
c
2
c
3
3
5
=
2
4
0
0
0
3
5
. (24)
If c
1
= 0, then
2
6
6
6
4
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2
3
7
7
7
5

c
2
c
3

0
0

.
so (19) implies that c
2
= c
3
= 0; hence, we may assume that c
1
= 1 in a nontrivial
solution of (24). Therefore,
2
6
6
6
6
6
6
6
6
6
6
6
4
d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2
3
7
7
7
7
7
7
7
7
7
7
7
5
2
4
1
c
2
c
3
3
5
=
2
4
0
0
0
3
5
. (25)
which implies that
2
6
6
6
4
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2
3
7
7
7
5

c
2
c
3


x
1
(X
0
)

x
2
(X
0
)

.
13
Since (23) has only one solution, this implies that c
2
= z and c
2
= j, so (25)
becomes
2
6
6
6
6
6
6
6
6
6
6
6
4
d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2
3
7
7
7
7
7
7
7
7
7
7
7
5
2
4
1
z
j
3
5
=
2
4
0
0
0
3
5
.
Computing the topmost entry of the vector on the left yields (20).
Example 7 Minimize
(.. ,. z. n) = .
2
,
2
z
2
n
2
subject to
. , z n = 10 and . , z 3n = 6. (26)
SolutionLet
1 =
.
2
,
2
z
2
n
2
2
z(. , z n) j(. , z 3n):
then
1
x
= . z j
1
y
= , z j
1

= z z j
1
w
= n z 3j.
so
.
0
= z j. ,
0
= z j. z
0
= z j. n
0
= z 3j. (27)
This and (26) imply that
(z j) (z j) (z j) (z 3j) = 10
(z j) (z j) (z j) (3z 9j) = 6.
Therefore,
4z 4j = 10
4z 12j = 6.
so z = 3 and j = 1,2. Now (27) implies that
(.
0
. ,
0
. z
0
. n
0
) =

5
2
.
7
2
.
5
2
3
2

.
14
Since (.. ,. z. n) is the square of the distance from(.. ,. z. n) to the origin, it attains
a minimum value (but not a maximum value) subject to the constraints; hence the
constrained minimum value is

5
2
.
7
2
.
5
2
.
3
2

= 27.
Example 8 The distance between two curves in R
2
is the minimum value of
p
(.
1
.
2
)
2
(,
1
,
2
)
2
.
where (.
1
. ,
1
) is on one curve and (.
2
. ,
2
) is on the other. Find the distance between
the ellipse
.
2
2,
2
= 1
and the line
. , = 4. (28)
SolutionWe must minimize
J
2
= (.
1
.
2
)
2
(,
1
,
2
)
2
subject to
.
2
1
2,
2
1
= 1 and .
2
,
2
= 4.
Let
1 =
(.
1
.
2
)
2
(,
1
,
2
)
2
z(.
2
1
2,
2
1
)
2
j(.
2
,
2
):
then
1
x
1
= .
1
.
2
z.
1
1
y
1
= ,
1
,
2
2z,
1
1
x
2
= .
2
.
1
j
1
y
2
= ,
2
,
1
j.
so
.
10
.
20
= z.
10
(i)
,
10
,
20
= 2z,
10
(ii)
.
20
.
10
= j (iii)
,
20
,
10
= j. (iv)
From (i) and (iii), j = z.
10
; from (ii) and (iv), j = 2z,
10
. Since the curves do
not intersect, z = 0, so .
10
= 2,
10
. Since .
2
10
2,
2
10
= 1 and (.
0
. ,
0
) is in the rst
quadrant,
(.
10
. ,
10
) =

2
_
6
.
1
_
6

. (29)
15
Now (iii), (iv), and (28) yield the simultaneous system
.
20
,
20
= .
10
,
10
=
1
_
6
. .
20
,
20
= 4.
so
(.
20
. ,
20
) =

2
1
2
_
6
. 2
1
2
_
6

.
From this and (29), the distance between the curves is
"

2
1
2
_
6

2
_
6

2
1
2
_
6

1
_
6

2
#
1=2
=
_
2

2
3
2
_
6

.
6 Proof of Theorem 1
Proof For notational convenience, let r
`
= , 1 _ _ m, so (6) becomes

dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2

dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2

dg
2
(X
0
)
d.
m
.
.
.
.
.
.
.
.
.
.
.
.
dg
m
(X
0
)
d.
1
dg
m
(X
0
)
d.
2

dg
m
(X
0
)
d.
m

= 0 (30)
Denote
U = (.
mC1
. .
mC2
. . . . .
n
) and U
0
= (.
mC1;0
. .
mC2;0
. . . . .
n0
).
From (30), the Implicit Function Theorem implies that there are unique continuously
differentiable functions h
`
= h
`
(U), 1 _ _ m, dened on a neighborhood N of U
0
,
such that
(h
1
(U). h
2
(U). . . . . h
m
(U). U) D. for all U N.
(h
1
(U
0
). h
2
(U
0
). . . . . h
m
(U
0
). U
0
) = X
0
. (31)
and
g
`
(h
1
(U). h
2
(U). . . . . h
m
(U). U) = 0. U N. 1 _ _ m. (32)
Again from (30), the system
2
6
6
6
6
6
6
6
6
6
6
6
6
4
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2

dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2

dg
2
(X
0
)
d.
m
.
.
.
.
.
.
.
.
.
.
.
.
dg
m
(X
0
)
d.
1
dg
m
(X
0
)
d.
2

dg
m
(X
0
)
d.
m
3
7
7
7
7
7
7
7
7
7
7
7
7
5
2
6
6
6
4
z
1
z
2
.
.
.
z
m
3
7
7
7
5
=
2
6
6
6
4

x
1
(X
0
)

x
2
(X
0
)
.
.
.

xm
(X
0
)
3
7
7
7
5
(33)
16
has a unique solution. This implies that
d(X
0
)
d.
i
z
1
dg
1
(X
0
)
d.
i
z
2
dg
2
(X
0
)
d.
i
z
m
dg
m
(X
0
)
d.
i
= 0 (34)
for 1 _ i _ m.
If m1 _ i _ n, differentiating (32) with respect to .
i
and recalling (31) yields
dg
`
(X
0
)
d.
i

m
X
jD1
dg
`
(X
0
)
d.
j
dh
j
(X
0
)
d.
i
= 0. 1 _ _ m.
If X
0
is local extreme point subject to g
1
(X) = g
2
(X) = = g
m
(X) = 0, then U
0
is an unconstrained local extreme point of (h
1
(U). h
2
(U). . . . h
m
(U). U); therefore,
d(X
0
)
d.
i

m
X
jD1
d(X
0
)
d.
j
dh
j
(X
0
)
d.
i
= 0.
The last two equations imply that

d(X
0
)
d.
i
d(X
0
)
d.
1
d(X
0
)
d.
2

d(X
0
)
d.
m
dg
1
(X
0
)
d.
i
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2

dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
i
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2

dg
2
(X
0
)
d.
m
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
dg
m
(X
0
)
d.
i
dg
m
(X
0
)
d.
1
dg
m
(X
0
)
d.
2

dg
m
(X
0
)
d.
m

= 0.
so

d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
. . .
dg
m
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
. . .
dg
m
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2
. . .
dg
m
(X
0
)
d.
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
d(X
0
)
d.
m
dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
m
. . .
dg
m
(X
0
)
d.
m

= 0.
17
Therefore, there are constant c
0
, c
1
, . . . c
m
, not all zero, such that
2
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
4
d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
. . .
dg
m
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
. . .
dg
m
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2
. . .
dg
m
(X
0
)
d.
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
d(X
0
)
d.
m
dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
m
. . .
dg
m
(X
0
)
d.
m
3
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5
2
6
6
6
6
6
4
c
0
c
1
c
3
.
.
.
c
m
3
7
7
7
7
7
5
=
2
6
6
6
6
6
4
0
0
0
.
.
.
0
3
7
7
7
7
7
5
. (35)
If c
0
= 0, then
2
6
6
6
6
6
6
6
6
6
6
6
6
4
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2

dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2

dg
2
(X
0
)
d.
m
.
.
.
.
.
.
.
.
.
.
.
.
dg
m
(X
0
)
d.
1
dg
m
(X
0
)
d.
2

dg
m
(X
0
)
d.
m
3
7
7
7
7
7
7
7
7
7
7
7
7
5
2
6
6
6
4
c
1
c
2
.
.
.
c
m
3
7
7
7
5
=
2
6
6
6
4
0
0
.
.
.
0
3
7
7
7
5
and (30) implies that c
1
= c
2
= = c
m
= 0; hence, we may assume that c
0
= 1 in
a nontrivial solution of (35). Therefore,
2
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
4
d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
. . .
dg
m
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
. . .
dg
m
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2
. . .
dg
m
(X
0
)
d.
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
d(X
0
)
d.
m
dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
m
. . .
dg
m
(X
0
)
d.
m
3
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5
2
6
6
6
6
6
4
1
c
1
c
2
.
.
.
c
m
3
7
7
7
7
7
5
=
2
6
6
6
6
6
4
0
0
0
.
.
.
0
3
7
7
7
7
7
5
. (36)
18
which implies that
2
6
6
6
6
6
6
6
6
6
6
6
6
4
dg
1
(X
0
)
d.
1
dg
1
(X
0
)
d.
2

dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
1
dg
2
(X
0
)
d.
2

dg
2
(X
0
)
d.
m
.
.
.
.
.
.
.
.
.
.
.
.
dg
m
(X
0
)
d.
1
dg
m
(X
0
)
d.
2

dg
m
(X
0
)
d.
m
3
7
7
7
7
7
7
7
7
7
7
7
7
5
2
6
6
6
4
c
1
c
2
.
.
.
c
m
3
7
7
7
5
=
2
6
6
6
4

x
1
(X
0
)

x
2
(X
0
)
.
.
.

xm
(X
0
)
3
7
7
7
5
Since (33) has only one solution, this implies that c
j
= z
j
, 1 _ _ n, so (36)
becomes
2
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
4
d(X
0
)
d.
i
dg
1
(X
0
)
d.
i
dg
2
(X
0
)
d.
i
. . .
dg
m
(X
0
)
d.
i
d(X
0
)
d.
1
dg
1
(X
0
)
d.
1
dg
2
(X
0
)
d.
1
. . .
dg
m
(X
0
)
d.
1
d(X
0
)
d.
2
dg
1
(X
0
)
d.
2
dg
2
(X
0
)
d.
2
. . .
dg
m
(X
0
)
d.
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
d(X
0
)
d.
m
dg
1
(X
0
)
d.
m
dg
2
(X
0
)
d.
m
. . .
dg
m
(X
0
)
d.
m
3
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5
2
6
6
6
6
6
4
1
z
1
z
2
.
.
.
z
m
3
7
7
7
7
7
5
=
2
6
6
6
6
6
4
0
0
0
.
.
.
0
3
7
7
7
7
7
5
.
Computing the topmost entry of the vector on the left yields yields (34), which com-
pletes the proof.
Example 9 Minimize
n
X
i D1
.
2
i
subject to
n
X
i D1
a
ri
.
i
= c
r
. 1 _ r _ m. (37)
where
n
X
i D1
a
ri
a
si
=
(
1 if r = s.
0 if r = s.
(38)
Solution Let
1 =
1
2
n
X
i D1
.
2
i

m
X
sD1
z
s
n
X
i D1
a
si
.
i
.
Then
1
x
i
= .
i

m
X
sD1
z
s
a
si
. 1 _ i _ n.
19
so
.
i 0
=
m
X
sD1
z
s
a
si
1 _ i _ n. (39)
and
a
ri
.
i 0
=
m
X
sD1
z
s
a
ri
a
si
.
Now (38) implies that
n
X
i D1
a
ri
.
i 0
=
m
X
sD1
z
s
n
X
i D1
a
ri
a
si
= z
r
.
From this and (37), z
r
= c
r
, 1 _ r _ m, and (39) implies that
.
i 0
=
m
X
sD1
c
s
a
si
. 1 _ i _ n.
Therefore,
.
2
i 0
=
m
X
r;sD1
c
r
c
s
a
ri
a
si
. 1 _ i _ n.
and (38) implies that
n
X
i D1
.
2
i 0
=
m
X
r;sD1
c
r
c
s
n
X
i D1
a
ri
a
si
=
m
X
rD1
c
2
r
.
The next theorem provides further information on the relationship between the
eigenvalues of a symmetric matrix and constrained extrema of its quadratic form. It
can be proved by successive applications of Theorem 1; however, we omit the proof.
Theorem 4 Suppose that A = a
rs
|
n
r;sD1
R
nn
is symmetric and let
Q(x) =
n
X
r;sD1
a
rs
.
r
.
s
.
Suppose also that
x
1
=
2
6
6
6
4
.
11
.
21
.
.
.
.
n1
3
7
7
7
5
minimizes Q subject to
P
n
i D1
.
2
i
. For 2 _ r _ n, suppose that
x
r
=
2
6
6
6
4
.
1r
.
2r
.
.
.
.
nr
3
7
7
7
5
.
20
minimizes Q subject to
n
X
i D1
.
2
i
= 1 and
n
X
i D1
.
i s
.
i
= 0. 1 _ s _ r 1.
Denote
z
r
=
n
X
i;jD1
a
ij
.
i r
.
jr
. 1 _ r _ n.
Then
z
1
_ z
2
_ _ z
n
and .
r
= z
r
.
r
. 1 _ r _ n.
21
7 Exercises
1. Find the point on the plane 2. 3, z = 7 closest to (1. 2. 3).
2. Find the extreme values of (.. ,) = 2. , subject to .
2
,
2
= 5.
3. Suppose that a. b > 0 and a
2
b
2
= 1. Find the extreme values of
(.. ,) = . , subject to a.
2
b,
2
= 1.
4. Find the points on the circle .
2
,
2
= 320 closest to and farthest from (2. 4).
5. Find the extreme values of
(.. ,. z) = 2. 3, z subject to .
2
2,
2
3z
2
= 1.
6. Find the maximum value of (.. ,) = ., on the line a. b, = 1, where
a. b > 0.
7. A rectangle has perimeter . Find its largest possible area.
8. A rectangle has area . Find its smallest possible perimeter.
9. A closed rectangular box has surface area . Find it largest possible volume.
10. The sides and bottom of a rectangular box have total area . Find its largest
possible volume.
11. A rectangular box with no top has volume V . Find its smallest possible surface
area.
12. Maximize (.. ,. z) = .,z subject to
.
a

,
b

z
c
= 1.
where a, b, c > 0.
13. Two vertices of a triangle are (a. 0) and (a. 0), and the third is on the ellipse
.
2
a
2

,
2
b
2
= 1.
Find its largest possible area.
14. Show that the triangle with the greatest possible area for a given perimeter is
equilateral, given that the area of a triangle with sides ., ,, z and perimeter s is
=
p
s(s .)(s ,)(s z).
22
15. Abox with sides parallel to the coordinate planes has its vertices on the ellipsoid
.
2
a
2

,
2
b
2

z
2
c
2
= 1.
Find its largest possible volume.
16. Derive a formula for the distance from (.
1
. ,
1
. z
1
) to the plane
a. b, cz = o.
17. Let X
i
= (.
i
. ,
i
. z
i
), 1 _ i _ n. Find the point in the plane
a. b, cz = o
for which
P
n
i D1
[XX
i
[
2
is a minimum. Assume that none of the X
i
are in the
plane.
18. Find the extreme values of (X) =
n
X
i D1
(.
i
c
i
)
2
subject to
n
X
i D1
.
2
i
= 1.
19. Find the extreme values of
(.. ,. z) = 2., 2.z 2,z subject to .
2
,
2
z
2
= 1.
20. Find the extreme values of
(.. ,. z) = 3.
2
2,
2
3z
2
2.z subject to .
2
,
2
z
2
= 1.
21. Find the extreme values of
(.. ,) = .
2
8., 4,
2
subject to .
2
2., 4,
2
= 1.
22. Find the extreme value of (.. ,) = ., subject to (a. b,)
2
= 1.
Assume that ab = 0.
23. Find the extreme values of (.. ,. z) = . ,
2
2z subject to
4.
2
9,
2
36z
2
= 36.
24. Find the extreme values of (.. ,. z. n) = (. z)(, n) subject to
.
2
,
2
z
2
n
2
= 1.
25. Find the extreme values of (.. ,. z. n) = (. z)(, n) subject to
.
2
,
2
= 1 and z
2
n
2
= 1.
23
26. Find the extreme values of (.. ,. z. n) = (. z)(, n) subject to
.
2
z
2
= 1 and ,
2
n
2
= 1.
27. Find the distance between the circle .
2
,
2
= 1 the hyperbola ., = 1.
28. Minimize (.. ,. .) =
.
2

2

,
2

2

z
2
;
2
subject to a. b, cz = J and .,
,, z > 0.
29. Find the distance from (c
1
. c
2
. . . . . c
n
) to the plane
a
1
.
1
a
2
.
2
a
n
.
n
= J.
30. Find the maximum value of (X) =
n
X
i D1
a
i
.
2
i
subject to
n
X
i D1
b
i
.
4
i
= 1, where
. q > 0 and a
i
, b
i
.
i
> 0, 1 _ i _ n.
31. Find the extreme value of (X) =
n
X
i D1
a
i
.
p
i
subject to
n
X
i D1
b
i
.
q
i
= 1, where ,
q>0 and a
i
, b
i
, .
i
> 0, 1 _ i _ n.
32. Find the minimum value of
(.. ,. z. n) = .
2
2,
2
z
2
n
2
subject to
. , z 3n = 1
. , 2z n = 2.
33. Find the minimum value of
(.. ,. z) =
.
2
a
2

,
2
b
2

z
2
c
2
subject to
1
.
2
,
3
z = J, assuming that at least one of
1
,
2
,
3
is
nonzero.
34. Find the extreme values of (.. ,. z) =
1
.
2
,
3
z subject to
.
2
a
2

,
2
b
2

z
2
c
2
= 1.
assuming that at least one of
1
,
2
,
3
is nonzero.
35. Find the distance from (1. 2. 3) to the intersection of the planes
. 2, 3z = 4 and 2. , 2z = 5.
24
36. Find the extreme values of (.. ,. z) = 2. , 2z subject to .
2
,
2
= 4
and . z = 2.
37. Find the distance between the parabola , = 1 .
2
and the line . , = 1.
38. Find the distance between the ellipsoid
3.
2
9,
2
6z
2
= 10
and the plane
3. 3, 6z = 70.
39. Show that the extreme values of (.. ,. z) = ., ,z z. subject to
.
2
a
2

,
2
b
2

z
2
c
2
= 1
are the largest and smallest eigenvalues of the matrix
2
4
0 a
2
a
2
b
2
0 b
2
c
2
c
2
0
3
5
.
40. Show that the extreme values of (.. ,. z) = ., 2,z 2z. subject to
.
2
a
2

,
2
b
2

z
2
c
2
= 1
are the largest and smallest eigenvalues of the matrix
2
4
0 a
2
,2 a
2
b
2
,2 0 b
2
c
2
c
2
0
3
5
.
41. Find the extreme values of .(, z) subject to
.
2
a
2

,
2
b
2

z
2
c
2
= 1.
42. Let a, b, c, , q, r, , , and ; be positive constants. Find the maximum value
of (.. ,. z) = .

subject to
a.
p
b,
q
cz
r
= 1 and .. ,. z > 0.
43. Find the extreme values of
(.. ,. z. n) = .n ,z subject to .
2
2,
2
= 4 and 2z
2
n
2
= 9.
25
44. Let a, b, c,and J be positive. Find the extreme values of
(.. ,. z. n) = .n ,z
subject to
a.
2
b,
2
= 1. cz
2
Jn
2
= 1.
if (a) aJ = bc; (b) aJ = bc.
45. Minimize (.. ,. z) = .
2
,
2
;z
2
subject to
a
1
. a
2
, a
3
z = c and b
1
. b
2
, b
3
z = J.
Assume that
. . ; > 0. a
2
1
a
2
2
a
2
3
= 0. and b
2
1
b
2
2
b
2
3
= 0.
Formulate and apply a required additional assumption.
46. Minimize (X. Y) =
n
X
i D1
(.
i

i
)
2
subject to
n
X
i D1
a
i
.
i
= c and
n
X
i D1
b
i
.
i
= J.
where
n
X
i D1
a
2
i
=
n
X
i D1
b
2
i
= 1 and
n
X
i D1
a
i
b
i
= 0.
47. Find (.
10;x
20
. . . . . .
n0
) to minimize
Q(X) =
n
X
i D1
.
2
i
subject to
n
X
i D1
.
i
= 1 and
n
X
i D1
i.
i
= 0.
Prove explicitly that if
n
X
jD1
,
i
= 1.
n
X
i D1
i,
i
= 0
and ,
i
= .
i 0
for some i {1. 2. . . . . n], then
n
X
i D1
,
2
i
>
n
X
i D1
.
2
i 0
.
26
48. Let
1
,
2
, . . . ,
n
and s be positive numbers. Maximize
(X) = (s .
1
)
p
1
(s .
2
)
p
2
(s .
n
)
pn
subject to .
1
.
2
.
n
= s.
49. Maximize (X) = .
p
1
1
.
p
2
2
.
pn
n
subject to .
i
> 0, 1 _ i _ n, and
n
X
i D1
.
i
o
i
= S.
where
1
,
2
,. . . ,
n
, o
1
, o
2
, . . . , o
n
, and V are given positive numbers.
50. Maximize
(X) =
n
X
i D1
.
i
o
i
subject to .
i
> 0, 1 _ i _ n, and
.
p
1
1
.
p
2
2
.
pn
n
= V.
where
1
,
2
,. . . ,
n
, o
1
, o
2
, . . . , o
n
, and S are given positive numbers.
51. Suppose that
1
,
2
, . . .
n
are positive and at least one of a
1
, a
2
, . . . , a
n
is
nonzero. Let (c
1
. c
2
. . . . . c
n
) be given. Minimize
Q(X) =
n
X
i D1
(.
i
c
i
)
2

i
subject to
a
1
.
1
a
2
.
2
a
n
.
n
= J.
52. Schwarzs inequality says that (a
1
. a
2
. . . . . a
n
) and (.
1
. .
2
. . . . . .
n
) are arbi-
trary n-tuples of real numbers, then
[a
1
.
1
a
2
.
2
a
n
.
n
[ _ (a
2
1
a
2
2
a
2
n
)
1=2
(.
2
1
.
2
2
.
2
n
)
1=2
.
Prove this by nding the extreme values of (X) =
n
X
i D1
a
i
.
i
subject to
n
X
i D1
.
2
i
= o
2
.
53. Let .
1
, .
2
, . . . , .
m
, r
1
, r
2
, . . . , r
m
be positive and
r
1
r
2
r
m
= r.
Show that

.
r
1
1
.
r
2
2
.
rm
m

1=r
_
r
1
.
1
r
2
.
2
r
m
.
m
r
.
and give necessary and sufcient conditions for equality. (Hint: Maximize
.
r
1
1
.
r
2
2
.
rm
m
subject to
P
m
jD1
r
j
.
j
= o > 0, .
1
> 0, .
2
> 0, . . . , .
m
> 0.)
27
54. Let A = a
ij
| be an m n matrix. Suppose that
1
,
2
, . . . ,
m
> 0 and
m
X
jD1
1

j
= 1.
and dene
o
i
=
n
X
jD1
[a
ij
[
p
i
. 1 _ i _ m.
Use Exercise 53 to show that

n
X
jD1
a
ij
a
2j
a
mj

_ o
1=p
1
1
o
1=p
2
2
o
1=pm
m
.
(With m = 2 this is Hlders inequality, which reduces to Schwarzs inequality
if
1
=
2
= 2.)
55. Let c
0
, c
1
, . . . , c
m
be given constants and n _ m1. Show that the minimum
value of
Q(X) =
n
X
rD0
.
2
r
subject to
n
X
rD0
.
r
r
s
= c
s
. 0 _ s _ m.
is attained when
.
r
=
m
X
sD0
z
s
r
s
. 0 _ r _ n.
where
m
X
`D0
o
sC`
z
`
= c
s
and o
s
=
n
X
rD0
r
s
. 0 _ s _ m.
Show that if {.
r
]
n
rD0
satises the constraints and .
r
= .
r0
for some r, then
n
X
rD0
.
2
r
>
n
X
rD0
.
2
r0
.
56. Suppose that n > 2k. Show that the minimum value of (W) =
n
X
i Dn
n
2
i
,
subject to the constraint
n
X
i Dn
n
i
1(r i ) = 1(r)
28
whenever r is an integer and 1 is a polynomial of degree _ 2k, is attained with
n
i 0
=
2k
X
rD0
z
r
i
r
. 1 _ i _ n.
where
2k
X
rD0
z
r
o
rCs
=
(
1 if s = 0.
0 if 1 _ s _ 2k.
and o
s
=
n
X
jDn

s
.
Show that if {n
i
]
n
i Dn
satises the constraint and n
i
= n
i 0
for some i , then
n
X
i Dn
n
2
i
>
n
X
i Dn
n
2
i 0
.
57. Suppose that n _ k. Show that the minimum value of
n
X
i D0
n
2
i
, subject to the
constraint
n
X
i D0
n
i
1(r i ) = 1(r 1)
whenever r is an integer and 1 is a polynomial of degree _ k, is attained with
n
i 0
=
k
X
rD0
z
r
i
r
. 0 _ i _ n.
where
k
X
rD0
o
rCs
z
r
= (1)
s
. 0 _ s _ k. and o
`
=
n
X
i D0
i
`
. 0 _ _ 2k.
Show that if
n
X
i D0
u
i
1(r i ) = 1(r 1)
whenever r is an integer and 1 is a polynomial of degree _ k, and u
i
= n
i 0
for some i , then
n
X
i D0
u
2
i
>
n
X
i D0
n
2
i 0
.
58. Minimize
(X) =
n
X
i D1
(.
i
c
i
)
2

i
subject to
n
X
i D1
a
i r
.
i
= J
r
. 1 _ r _ m
29
Assume that m > 1,
1
,
2
, . . .
m
> 0, and
n
X
i D1

i
a
i r
a
i s
=
(
1 if r = s.
0 if r = s.
30
8 Answers to selected exercises
1.

15
7

2
7
.
25
7

2. 5 3. 1,
_
ab, 1,
_
ab
4. (8. 16) is closest, (8. 16) is farthest. 5.
p
53,6 6. 1,4ab 7.
2
,4
8. 4
_
9.
3=2
,6
_
6 10.
3=2
,6
_
3 11. 3(2V )
2=3
12. abc,27
13. ab 15. 8abc,3
_
3
18. (1 j)
2
and (1 j)
2
, where j =
0
@
n
X
jD1
c
2
j
1
A
1=2
19. 1, 2 20. 2, 4
21. 2,3, 2 22. [[,4[ab[ 23.
_
5, 73,16 24. 1 25. 2
26. 2 27.
_
2 1 28.
J
2
(a)
2
(b
2
) (c;)
2
29.
[J a
1
c
1
a
2
c
2
a
n
c
n
)a
i
[
q
a
2
1
a
2
2
a
2
n
30.

n
X
i D1
a
2
i
b
i
!
1=2
31.

n
X
i D1
a
q=.qp/
i
b
p=.pq/
i
!
1p=q
is a constrained maximum if < q, a constrained
minimum if > q
32. 689,845 33.
J
2

2
1
a
2

2
2
b
2

2
3
c
2
34. (
2
1
a
2

2
2
b
2

2
3
c
2
)
1=2
35.
p
693,45 36. 2, 6 37. 7,4
_
2 38. 10
_
6,3 41. [c[
_
a
2
b
2
,2
42.
;
qr

;
r

3
43. 3 44. (a) 1,
_
bc (b) 1,
_
aJ = 1,
_
bc
46.

c
n
X
i D1
a
i

i
!
2

J
n
X
i D1
b
i

i
!
2
47. .
i 0
= (4n 2 6i ),n(n 1)
48.
h
.n1/s
P
i
P

p
1
1

p
2
2

pn
n
49.

2

n

p
1
Cp
2
CCpn
(
1
o
1
)
p
1
(
2
o
2
)
p
2
(
n
o
n
)
pn
50. (
1

2

n
)

V
(o
1

1
)
p
1
(o
2

2
)
p
2
(o
n

n
)
pn
1
p
1
Cp
2
CCpn
51.

J
n
X
i D1
a
i
c
i
!
)
2
,

n
X
i D1
a
2
i

i
!
52.

n
X
i D1
a
2
i
!
1=2

n
X
i D1
.
2
i 0
!
1=2
58.
m
X
rD1

J
r

n
X
i D1
a
i r
c
i
!
2
31

Вам также может понравиться