Scribd
Upload
85 views

Homework Problems: X X y y

- This document contains the third part of a weekly math homework assignment on extremizing functions subject to constraints using Lagrange multipliers. - It provides 4 multi-step problems as examples, covering finding extrema on circles and discs, minimizing distance from a point to a surface, and maximizing volume of a pyramid under a surface area constraint. - Hints are provided that the Lagrange equations for these types of constrained optimization problems involve setting partial derivatives of the function and constraint equal to each other, multiplied by a Lagrange multiplier.

Uploaded by

ss_nainamohammed
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
85 views

Homework Problems: X X y y

- This document contains the third part of a weekly math homework assignment on extremizing functions subject to constraints using Lagrange multipliers. - It provides 4 multi-step problems as examples, covering finding extrema on circles and discs, minimizing distance from a point to a surface, and maximizing volume of a pyramid under a surface area constraint. - Hints are provided that the Lagrange equations for these types of constrained optimization problems involve setting partial derivatives of the function and constraint equal to each other, multiplied by a Lagrange multiplier.

Uploaded by

ss_nainamohammed
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

7/22/2004 EXTREMA WITH CONSTRAINTS

Maths21a

This is part 3 (of 3) of the weekly homework. It is due July 27 at the beginning of class.
SUMMARY.
Extremize f (x, y) under the constraint g(x, y) = c: Solve g(x, y) = c, f (x, y) = g(x, y)
with Lagrange multiplier , These are 3 equations for 3 unknowns x, y, :
fx (x, y) = gx (x, y)
fy (x, y) = gy (x, y)
g(x, y) = c
In three dimensions, the Lagrange equations form 4 equations for 4 unknowns.
fx (x, y, z)
fy (x, y, z)
fz (x, y, z)
g(x, y, z)

=
=
=
=

gx (x, y, z)
gy (x, y, z)
gz (x, y, z)
c

Homework Problems
1) (4 points) Find the extrema of the function f (x, y) = e x
x2 + y 2 = 4 using the method of Lagrange multipliers.

2 y 2

(x2 + 2y 2 ) on the circle g(x, y) =

Solution:
The Lagrange equations are
(2x 2x3 4xy 2 )ex

2 y 2

= 2x

x2 y 2

(4y 4y 3 2x2 y)e


= 2y
2
2
x +y = 4.

Case 1: If x = 0, then the first equation is ok and we get from the third equation y = 2.
Case 2: If y = 0, then the second equation is ok and we get from the third edquation
x = 2.
Case 3: If x = 0 and y = 0, then the first two equations are ok, but clashes with the
third. Forget this case.
Case 4: If both x and y are not zero we can divide the first equation by 2x and the second
by 2y. We also replace x2 y 2 by 4
(1 x2 2y 2 )e4 =
(2 2y 2 x2 )e4 =
x2 + y 2 = 4 .
But setting the first two equations equal leads to a contradiction. Also this case 4) has no
solutions. We end up with the four solutions (2, 0), (2, 0), (0, 2), (0, 2). The minimal
values are f (2, 0) = 4e4 , the maximal values are f (0, 2) = 8e4 .

2) (4 points) Find the extrema of the same function f (x, y) = e x


problem but now on the entire disc {x2 + y 2 4} of radius 2.

2 y 2

(x2 + 2y 2 ) as in the previous

Solution:
2
2
In the last homework we have seen that the equation f = ((2x 2x 3 )ex y , (4y
2
2
4y 3 )ex y ) = (0, 0) has the solutions x = 0, y = 0, x = 0, y = 1, y = 0, x = 1. Together with the previous problem, we can now make a list of all the candidates for extrema
1. Extrema inside
2. Extrema on the boundary
point f =
point f =
(0,0) 0
(2,0) 4/e4
(1,0) 1/e
(-2,0) 4/e4
(-1,0) 1/e
(0,2) 8/e4
(0,1) 2/e
(0,-2) 8/e4
(0,-1) 2/e
We see that the origin is the minimum and the points (0, 1) are both the maxima.
3) (4 points) Find the points (x, y, z) on the surface g(x, y, z) = xy 2 z 3 2 = 0 that are closest
to the origin (0, 0, 0).

Solution:

Instead of extremizing the distance x2 + y 2 + z 2 we extremize the function f (x, y) =


x2 + y 2 + z 2 . We have the Lagrange equations
2x
2y
2z
xy 2

=
=
=
=

y 2
2xy
3z 2
z3 + 2

1. Case: z = 0. We can then not have y = 0 nor x = 0 and end up with


2x = y 2
2 = 2x
xy 2 = 2

which gives 2 = 2 y 2 , x2 = and so that x = 1, y = 2. 2. Case: x = 0. Gives y = 0


and z = 21/3 .
3. Case: y = 0. Gives x = 0 and z = 21/3 .
4. Case: all x, y, z are nonzero. Then
2x
1
2
xy 2

=
=
=
=

y 2
x
3z
z3 + 2

Eliminating = 1/x gives


2x2 = y 2
2 = 3z/x
xy 2 = z 3 + 2
Solving for y 2 = 2x2 and z = 2x/3 from the first two equations and
this into
plugging
3
3
(1/3)
(1/3)
the third gives 2x = 8x /27 + 2 gives x = 3/31
and so y = 2(3/31
).
(1/3)
(
The distance from the point
(0,
0,
2
)
to
the
origin
is
2
1/3)

1.25
the
distance
from

1/6
1/3
1.7723. The first one is the
the points (3/31 (1, 2, 2/3) to the origin is 31
minimum.
4) (4 points) Let a, b, c be non-negative constants and let F be the function F (x, y, z) = x log(x)
y log(y) z log(z) ax by cz. Find the maxima and minima of F on x > 0, y > 0, z > 0
under the constraint x + y + z = 1.

Solution:
The Lagrange equations are
log(x) 1 a
log(x) 1 b
log(x) 1 c
x+y+z

=
=
=
=

From the first three equations, we get


x = e(1+a+) = e1 ea
y = e(1+b+) = e1 eb
z = e(1+c+) = e1 ec
Plugging this into the fourth equation gives e1 (ea + eb + ec ) = 1 so that e1 =
(ea + eb + ec )1 and
x = ea /(ea + eb + ec )
y = eb /(ea + eb + ec )
z = ec /(ea + eb + ec )
5) (4 points) Which pyramid of height h over a square [a, a] [a, a] = {(x, y) | a x
a, a y a and surface area 4 has maximal volume?
P.S. Solving the Lagrange equations are a bit tricky here and most people have to make several
attempts to crack them. If you should be short on time to hack the equations, feel free to ask
Ben or Oliver.

Solution:

The area is 4a h2 + a2 + 4a2 = 4, the volume is V = 4ha2 /3. We have


the mathematical
problem to extremize f (x, y) = yx2 over the constraint g(x, y) = x y 2 + x2 + x2 = 1.
(We do not drag along the factor 4/3).
The Lagrange system is
q

2xy = ( y 2 + x2 + x2 / y 2 + x2 + 2x)
q

x2 = yx/ y 2 + x2
q

1 = x y 2 + x2 + x2
There are different possibilities to solve this a bit
system. Since x 6= 0, we can
tricky
2
2
divide the second equation by x. replace it by x x + y = y or its square. The third
equation becomes therefore 1 = y + x2 . The square roots of the first equation can then
also be replaced by y/x.
2xy = (y/x + x3 /(y) + 2x)
x2 (y 2 + x2 ) = 2 y 2
1 = y + x2 .
Until now, we were just trying to simplify, not yet eliminiating. To eliminate , we
multiply the first equation by and replace y by (1x2 ). We end up with two equations:
2xy 2 = (1 x2 )((1 x2 )/x + x3 /(1 x2 ) + 2x)
x2 (y 2 + x2 ) = (1 x2 )2
Te second equation is equivalent to x2 y 2 = 1 2x2 so that the second equation gives
y 2 = (1 2x2 )/x2 . This can be plugged into the first equation to obtain
2x(1 2x2 ) = x(1 x2 )2 + x5 + 2x3 (1 x2 )
which simplifies to x = 4x3 . The solution is therefore a = 1/2, h =

2. Whao!

Remark. This problem was on the edge what is possible to solve by hand. A two dimensional version of the
problem, where one asks for the equilateral triangle of height h and ground side 2a which has maximal area if the

circumfence 2a + 2 a2 + h2 = 2 is fixed. The solutions of the Lagrange equations are then a = 1/3, h = 1/ 3.

Remarks
(You dont need to read these remarks to do the problems.)
Remark to problem 4) This problem appears in thermodynamics and is relevant in biology
or chemistry. If x, y, z are the probabilities that a system is in state X, Y, Z and a, b, c are
the energies for these states. Then x log(x) y log(y) z log(z) is called the entropy of
the system and E = ax + by + cz is the energy. The number F (x, y, z) is called the free
energy. If energy is fixed, nature tries to maximize entropy. Otherwise it tries to minimize
the free energy F = S E. If we extremize F under the constraint of having total probability
G(x, y, z) = x + y + z = 1, we obtain the so called Gibbs distribution.

Challenge Problems
(Solutions to these problems are not turned in with the homework.)
1) What does it mean that the Lagrange multiplier is zero in a constrained optimization problem?
2) Extend the Lagrange method to arbitrary dimensions. Find the equations to find the extrema
of a function f (x1 , ..., xn ) under the constraint g(x1 , ..., xn ) = c.
P

3) Let I = ni=1 pi log(pi ) be the entropy of a probability distribution (p1 , . . . , pn ). Show that
among all probability distributions, the one where p i = 1/n is the one which maximizes entropy.
4) Find the box of dimensions x, y, z, w in four dimensional space which has maximal hyper
volume V = xyzw under the constraint that the surface volume A = 2xy + 2xz + 2xw +
2yz + 2yw + 2zw is fixed equal to 12.

You might also like