Lagrange Multipliers

Lagrange Multipliers

Suppose that we have a function f(x,y) that we want to maximize in the restricted domain g(x,y) = c for some constant c.  Then we can look at the level curves of f and seek the largest level curve that intersects the curve g(x,y) = c.  It is not hard to see that these curves will be tangent.  Hence the gradient vectors will be parallel.

 Theorem Let f(x,y) be differentiable and g(x,y) = c define a smooth curve.  Then the maximum and minimum of f subject to the constraint g occur when           grad f  =  l grad g for some constant l.

Example

Find the extrema of

f(x,y)  =  x2 - y2

subject to the constraint

y - x2  =  0

Solution:

We have

<2x, -2y>  =  l<-2x, 1>

This gives us the three equations:

2x  =  -l 2x         -2y  =  l (1)          and         y - x2  =  0

the first equation gives us (for x nonzero)

l
=  -1

Hence the second equation becomes

-2y  =  -1

so that

y  =  1/2

the third equation gives us

1/2 - x2  =  0

Hence

x  = / 2

For x = 0, we see that y = 0.  Hence the two possible local extrema are

( / 2, 1/2)        and        (0,0)

Plugging into f(x,y), we see that

f ( /2, 1/2)  =  1/4

and

f(0,0)  =  0

Hence 1/4 is the local maximum and 0 is the local minimum.

Example 2

Find the distance from the origin to the surface

xyz  =  8

Solution

We minimize

D  =  x2 + y2 + z2

subject to the constraint

xyz  =  1

We have

<2x, 2y, 2z> = l <yz, xz, xy>

2x  =  l yz        2y  =  l xz,        2z  =  l xy

or

2x             2y               2z
l  =              =                 =
yz             xz                xy

Multiply all three by xyz to get

2x2  =  2y2  =  2z2

Hence

x  =  ± y  =  ± z

so that

±x3  =   8    or    x  =  ±2

We get the points

(2, 2, 2), (2, -2, -2), (-2, -2, 2), (-2, 2, -2)

these all have distance from the origin.

Two constraints

Example

Maximize

x2 + y2 + z2

on the intersection of the two surfaces:

xyz  =  1       and        x2 + y2 + 2z2  =  4

Solution

Now we set

which gives

<2x, 2y, 2z>  =  a <yz, xz, xy> + b <2x, 2y, 4z>

we have the five equations:

2x  =  ayz + 2bx,     2y  =  axz + 2by,     2z  =  axy + 4bz,

xyz  =  1,   and     x2 + z  =  1

Multiply the first equation by x, the second by y and the third by z gives

2x2  =  axyz + 2bx2,     2y2  =  axyz + 2by2,     2z2  =  axyz + 4bz2,

Solving each for axyz gives

axyz  =  2x2  - 2bx2  =  2y2  - 2by2  =  2z2  - 4bz2

This gives that

2x2(1 - b)  =  2y2(1 - b)  =  2z2 (1 - 2b)

This first equality gives

x  =  ±y

Using the last of the original equations to solve for x2 gives

x2   =  1 - z

The equation

xyz  =  1

becomes

(1 - z)z  =  1

or

z2 - z + 1  =  0

z  = 1/2 ± /2

Now we leave it to the reader to use

x2 + z  =  1      and      x  =  ±y

To find x and y.