Lecture 31: Lagrange Multiplier Method: P DF DT T DR DT P P P P

1
Lecture 31 : Lagrange Multiplier Method

Let f : S R, S R3 and X0 S. If X0 is an interior point of the constrained set S, then
we can use the necessary and sufficient conditions (first and second derivative tests) studied in the
previous lecture in order to determine whether the point is a local maximum or minimum (i.e.,
local extremum) of f on S. If X0 is not an interior point then one cannot apply these tests. For
example, one cannot apply these tests at a point on a sphere x2 + y 2 + z 2 = c2 , because no point
is an interior point in this constrained set.
In general, constrained extremum problems are very difficult to solve and there is no general
method for solving such problems. In case the constrained set is a level surface, for example a
sphere, there is a special method called Lagrange multiplier method for solving such problems. So,
we will be dealing with the following type of problem.
Problem : Find the local or absolute maxima and minima of a function f (x, y, z) on the (level)
surface S := {(x, y, z) : g(x, y, z) = 0} where g : R3 R.
Let us illustrate the problem with an example.
Example : Find a point on the plane {(x, y, z) : 2x + 3y z = p
5} which is nearest to the origin of
3
R . Note that here we are minimizing the function f (x, y, z) = x2 + y 2 + z 2 or x2 + y 2 + z 2 over
the constrained set S defined by g(x, y, z) = 2x + 3y z 5.
Necessary condition: Let P0 = (x0 , y0 , z0 ) be a point on S := {(x, y, z) : g(x, y, z) = 0}. Suppose
P0 is a local extremum of f over S. Let us try to find a necessary condition. Our argument is going
to be geometric and so we will not question certain assumptions made in the following argument.
Let us first assume that f and g have continuous partial derivatives and g|P0 6= 0. Let C =
R(t) = x(t)i + y(t)j + z(t)k be a curve on S passing through P0 and let P0 = R(t0 ). Since f has
a local extremum at P0 on S, it has a local extremum on C as well. Therefore df
dt |t0 = 0. By the
dR
chain rule, f dt = 0. Since the curve C is arbitrary, we conclude that f |P0 is perpendicular to
the tangent plane of S at P0 . But we already know that g|P0 is also perpendicular to the tangent
plane of S at P0 . Therefore, there exists R such that f |P0 = g|P0 .
So the following method is anticipated.
Lagrange Multiplier Method:
Suppose f and g have continuous partial derivatives. Let
(x0 , y0 , z0 ) S := {(x, y, z) : g(x, y, z) = 0} and g(x0 , y0 , z0 ) 6= 0. If f has a local maximum or
minimum at (x0 , y0 , z0 ) then there exists R such that
f (x0 , y0 , z0 ) = g(x0 , y0 , z0 ).
To find the extremum points, in practice, we consider the following equations
f (x, y, z) = g(x, y, z) and g(x, y, z) = 0.
(1)
These equations are solved for the unknowns x, y, z and . Then the local extremum points are
found among the solutions of these equations.
Let us illustrate the method with a few examples.
Examples: 1. Let us find a point on the plane 2x + 3y z = 5 in R3 which is nearest to
the origin. We have to minimize the function f (x, y, z) = x2 + y 2 + z 2 subject to the constraint
g(x, y, z) = 2x + 3y z 5 = 0. Here note that g 6= 0 at all points. The equations given in (1)
imply that
2x = 2, 2y = 3, 2z = and 2x + 3y z 5 = 0.
Substituting x = , y = 3
2 and z = 2 in the equation 2x + 3y z 5 = 0, we obtain that
5
= 75 and hence = 57 and (x, y, z) = ( 57 , 15
14 , 14 ) satisfy the equation (1). Since f attains its
15
5
minimum on the plane, by the Lagrange multipliers method, the point ( 57 , 14
, 14
) has to be the
nearest point.
2. Consider the problem of minimizing the function f (x, y) = x2 + y 2 subject to the condition
g(x, y) = (x 1)3 y 2 = 0. The problem is to find a point on the curve y 2 = (x 1)3 which is
nearest to the origin of R2 . Geometrically, it is clear that the point (1, 0) is the nearest point. But
g(1, 0) = 0 while f (1, 0) = (2, 0). Therefore f (1, 0) 6= g(1, 0) for any . This explains that
the condition g(x0 , y0 , z0 ) 6= 0 cannot be dropped from the Lagrange multiplier method and a
point at which g is (0, 0) could be an extremum point.
3. Let us evaluate the minimum and maximum value of the function f (x, y) = 2 x2 2y 2 subject
to the condition g(x, y) = x2 + y 2 1. If we use the Lagrange multiplier method, the equations in
(1) imply that 2x + 2x = 0, 4y + 2y = 0 and x2 + y 2 1 = 0. From the first two equations, we
must have either = 1 or = 2. If = 1, then y = 0, x = 1 and f (x, y) = 1. Similarly,
if = 2, then y = 1, x = 0 and f (x, y) = 0. Since the continuous function f (x, y) achieves its
maximum and minimum over the closed and bounded set x2 + y 2 = 1, the points (0, 1) are the
minima and (1, 0) are the maxima, and the maximum value and the minimum value of f are 1
and 0 respectively.
This problem can also be solved using substitution as follows. Since x2 + y 2 1 = 0, x2 = 1 y 2 .
Substituting this into f , we get f (x, y) = 1 y 2 . We are back to a one variable problem, which
has a maximum at y = 0, where x = 1 and f (x, y) = 1. Since y [1, 1], f has a minimum at
y = 1 where x = 0 and f (x, y) = 0. Although we solved this problem easily using substitution, it
is usually very hard to solve such constrained problems using substitution.
4. Given n positive numbers a1 , a2 , .., an , let us find the maximum value of the expression a1 x1 +
a2 x2 + ... + an xn where x21 + ... + x2n = 1. Note that here f (x1 , x2 , .., xn ) = a1 x1 + a2 x2 + ... + an xn
and g(x1 , x2 , ..., xn ) = x21 + ... + x2n 1. Although we stated the Lagrange multiplier method
in R3 , it works in Rn as well. The equations in (1) imply that a1 = 2x1 , ...,
an = 2xn and
a2 +..+a2
x21 + ... + x2n 1 = 0. Therefore, a21 + a22 + .. + a2n = 42 . This implies that = 1 2 n . Since
the continuous function
its minimum and maximum on the closed and bounded subset
f2 achieves
p
2
a
+..+a
a1 a2
n
1
leads to the maximum value f ( 2
, 2 , ..., a2n ) = a21 + .. + a2n and
x21 + ... + x2n = 1, =
2
2
a +..+a2
= 1 2 n leads to the minimum value of f .
Problem 1: Assume that among all rectangular boxes with fixed surface area of 20 square meters,
there is a box of largest possible volume. Find its dimensions.
Solution: Let the box have sides of length x, y, z > 0. Then V (x, y, z) = xyz and xy+yz+xz = 10.
Using the method of Lagrange multipliers, we see that yz = (y+z), xz = (x+z) and xy = (x+y).
r
10
It is easy to see that x, y, z > 0. Now, we can see that x = y = z and therefore, x = y = z =
.
3
Problem 2: A company produces steel boxes at three different plants in amounts x, y and z,
respectively, producing an annual revenue of f (x, y, z) = 8xyz 2 200(x + y + z). The company is
to produce 100 units annually. How should the production be distributed to maximize revenue?
Solution: Here, g(x, y, z) = x + y + z 100. The Lagrange multiplier method implies that
8yz 2 200 = , 8xz 2 200 = , 16xyz 200 = and x + y + z 100 = 0. These imply that
x = y, z = 2x and x = 25.

Lecture 31: Lagrange Multiplier Method: P DF DT T DR DT P P P P

Uploaded by

Copyright:

Available Formats

Lecture 31: Lagrange Multiplier Method: P DF DT T DR DT P P P P

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture 31: Lagrange Multiplier Method: P DF DT T DR DT P P P P

Uploaded by

Copyright:

Available Formats

1

Lecture 31 : Lagrange Multiplier Method

You might also like