Math 241: Multivariable calculus, Lecture 15 Lagrange Multipliers.
go.illinois.edu/math241fa17
Wednesday, October 4th, 2017
go.illinois.edu/math241fa17. Math 241: Problem of the day
Problem: Find the absolute maximum and minimum value of the function f (x, y) = x3 − x + y 2x + y + 2 on the unit disk D = {(x, y) | x2 + y 2 ≤ 1}. What guarantees that the max and min exist?
go.illinois.edu/math241fa17. ∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
−1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x . (Note that x nor y can be zero, so dividing by x or y is no problem.) Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
2 2 1 Put z = x , get 3z − z + 4 = 0. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points. So we need to look at the boundary!
Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
go.illinois.edu/math241fa17. −1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x . (Note that x nor y can be zero, so dividing by x or y is no problem.) Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
2 2 1 Put z = x , get 3z − z + 4 = 0. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points. So we need to look at the boundary!
Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
go.illinois.edu/math241fa17. (Note that x nor y can be zero, so dividing by x or y is no problem.) Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
2 2 1 Put z = x , get 3z − z + 4 = 0. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points. So we need to look at the boundary!
Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
−1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x .
go.illinois.edu/math241fa17. Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
2 2 1 Put z = x , get 3z − z + 4 = 0. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points. So we need to look at the boundary!
Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
−1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x . (Note that x nor y can be zero, so dividing by x or y is no problem.)
go.illinois.edu/math241fa17. 2 2 1 Put z = x , get 3z − z + 4 = 0. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points. So we need to look at the boundary!
Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
−1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x . (Note that x nor y can be zero, so dividing by x or y is no problem.) Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
go.illinois.edu/math241fa17. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points. So we need to look at the boundary!
Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
−1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x . (Note that x nor y can be zero, so dividing by x or y is no problem.) Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
2 2 1 Put z = x , get 3z − z + 4 = 0.
go.illinois.edu/math241fa17. So we need to look at the boundary!
Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
−1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x . (Note that x nor y can be zero, so dividing by x or y is no problem.) Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
2 2 1 Put z = x , get 3z − z + 4 = 0. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points.
go.illinois.edu/math241fa17. Solution Part 1: Critical points
If f (x, y) = x3 − x + y 2x + y + 2 the gradient vector is
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
−1 If this is the zero vector we get 2xy + 1 = 0, or y = 2x . (Note that x nor y can be zero, so dividing by x or y is no problem.) Substitute in the first component to get 1 1 3x2 − 1 + = 0 =⇒ 3x4 − x2 + = 0. 4x2 4
2 2 1 Put z = x , get 3z − z + 4 = 0. The discriminant b2 − 4ac = 1 − 3 = −2, so there are NO critical points. So we need to look at the boundary!
go.illinois.edu/math241fa17. Look on the boundary for points where Du~f (a, b) = 0, for u~ tangent to the boundary. In other words we want ∇f (a, b) · u~ = 0. But we know that u~ satisfies ∇g(a, b) · u~ = 0. So we want
∇f (a, b) = λ∇g(a, b) (*)
for some constant λ, called the Lagrange multiplier. At such points where (∗) holds, the level curves of f and of g are tangent.
Lagrange multipliers Idea
IDEA: Function f (x, y) is defined on a domain with boundary given by g(x, y) = 0. Want: absolute max/min of f on boundary.
go.illinois.edu/math241fa17. In other words we want ∇f (a, b) · u~ = 0. But we know that u~ satisfies ∇g(a, b) · u~ = 0. So we want
∇f (a, b) = λ∇g(a, b) (*)
for some constant λ, called the Lagrange multiplier. At such points where (∗) holds, the level curves of f and of g are tangent.
Lagrange multipliers Idea
IDEA: Function f (x, y) is defined on a domain with boundary given by g(x, y) = 0. Want: absolute max/min of f on boundary. Look on the boundary for points where Du~f (a, b) = 0, for u~ tangent to the boundary.
go.illinois.edu/math241fa17. But we know that u~ satisfies ∇g(a, b) · u~ = 0. So we want
∇f (a, b) = λ∇g(a, b) (*)
for some constant λ, called the Lagrange multiplier. At such points where (∗) holds, the level curves of f and of g are tangent.
Lagrange multipliers Idea
IDEA: Function f (x, y) is defined on a domain with boundary given by g(x, y) = 0. Want: absolute max/min of f on boundary. Look on the boundary for points where Du~f (a, b) = 0, for u~ tangent to the boundary. In other words we want ∇f (a, b) · u~ = 0.
go.illinois.edu/math241fa17. So we want
∇f (a, b) = λ∇g(a, b) (*)
for some constant λ, called the Lagrange multiplier. At such points where (∗) holds, the level curves of f and of g are tangent.
Lagrange multipliers Idea
IDEA: Function f (x, y) is defined on a domain with boundary given by g(x, y) = 0. Want: absolute max/min of f on boundary. Look on the boundary for points where Du~f (a, b) = 0, for u~ tangent to the boundary. In other words we want ∇f (a, b) · u~ = 0. But we know that u~ satisfies ∇g(a, b) · u~ = 0.
go.illinois.edu/math241fa17. At such points where (∗) holds, the level curves of f and of g are tangent.
Lagrange multipliers Idea
IDEA: Function f (x, y) is defined on a domain with boundary given by g(x, y) = 0. Want: absolute max/min of f on boundary. Look on the boundary for points where Du~f (a, b) = 0, for u~ tangent to the boundary. In other words we want ∇f (a, b) · u~ = 0. But we know that u~ satisfies ∇g(a, b) · u~ = 0. So we want
∇f (a, b) = λ∇g(a, b) (*) for some constant λ, called the Lagrange multiplier.
go.illinois.edu/math241fa17. Lagrange multipliers Idea
IDEA: Function f (x, y) is defined on a domain with boundary given by g(x, y) = 0. Want: absolute max/min of f on boundary. Look on the boundary for points where Du~f (a, b) = 0, for u~ tangent to the boundary. In other words we want ∇f (a, b) · u~ = 0. But we know that u~ satisfies ∇g(a, b) · u~ = 0. So we want
∇f (a, b) = λ∇g(a, b) (*) for some constant λ, called the Lagrange multiplier. At such points where (∗) holds, the level curves of f and of g are tangent.
go.illinois.edu/math241fa17. ∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
Boundary g(x, y) = x2 + y 2 = 1
∇g = h2x, 2yi
Then ∇f = λ∇g gives
3x2 − 1 + y 2 = λ2x, 2yx + 1 = λ2y
Solution (0, 1) or (0, −1).Max: f (0, 1) = 3, Min: f (0, −1) = 1.
Solution Part 2: Boundary
f (x, y) = x3 − x + y 2x + y + 2 with gradient vector
go.illinois.edu/math241fa17. Boundary g(x, y) = x2 + y 2 = 1
∇g = h2x, 2yi
Then ∇f = λ∇g gives
3x2 − 1 + y 2 = λ2x, 2yx + 1 = λ2y
Solution (0, 1) or (0, −1).Max: f (0, 1) = 3, Min: f (0, −1) = 1.
Solution Part 2: Boundary
f (x, y) = x3 − x + y 2x + y + 2 with gradient vector
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
go.illinois.edu/math241fa17. Then ∇f = λ∇g gives
3x2 − 1 + y 2 = λ2x, 2yx + 1 = λ2y
Solution (0, 1) or (0, −1).Max: f (0, 1) = 3, Min: f (0, −1) = 1.
Solution Part 2: Boundary
f (x, y) = x3 − x + y 2x + y + 2 with gradient vector
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
Boundary g(x, y) = x2 + y 2 = 1
∇g = h2x, 2yi
go.illinois.edu/math241fa17. Solution (0, 1) or (0, −1).Max: f (0, 1) = 3, Min: f (0, −1) = 1.
Solution Part 2: Boundary
f (x, y) = x3 − x + y 2x + y + 2 with gradient vector
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
Boundary g(x, y) = x2 + y 2 = 1
∇g = h2x, 2yi
Then ∇f = λ∇g gives
3x2 − 1 + y 2 = λ2x, 2yx + 1 = λ2y
go.illinois.edu/math241fa17. Max: f (0, 1) = 3, Min: f (0, −1) = 1.
Solution Part 2: Boundary
f (x, y) = x3 − x + y 2x + y + 2 with gradient vector
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
Boundary g(x, y) = x2 + y 2 = 1
∇g = h2x, 2yi
Then ∇f = λ∇g gives
3x2 − 1 + y 2 = λ2x, 2yx + 1 = λ2y
Solution (0, 1) or (0, −1).
go.illinois.edu/math241fa17. Solution Part 2: Boundary
f (x, y) = x3 − x + y 2x + y + 2 with gradient vector
∇f (x, y) = h3x2 − 1 + y 2, 2yx + 1i
Boundary g(x, y) = x2 + y 2 = 1
∇g = h2x, 2yi
Then ∇f = λ∇g gives
3x2 − 1 + y 2 = λ2x, 2yx + 1 = λ2y
Solution (0, 1) or (0, −1).Max: f (0, 1) = 3, Min: f (0, −1) = 1.
go.illinois.edu/math241fa17. Find absolute max/min of f (x, y) = x2 − y 2 on D = {(x, y) | x2 + y 2 ≤ 1}. Critical points on D: ∇f = h2x, −2yi = h0, 0i ⇒ (x, y) = (0, 0), and f (0, 0) = 0.
We can parameterize the boundary, and solve... in general this can be hard....
Finding max/min on boundary.
go.illinois.edu/math241fa17. Critical points on D: ∇f = h2x, −2yi = h0, 0i ⇒ (x, y) = (0, 0), and f (0, 0) = 0.
We can parameterize the boundary, and solve... in general this can be hard....
Finding max/min on boundary.
Find absolute max/min of f (x, y) = x2 − y 2 on D = {(x, y) | x2 + y 2 ≤ 1}.
go.illinois.edu/math241fa17. ⇒ (x, y) = (0, 0), and f (0, 0) = 0.
We can parameterize the boundary, and solve... in general this can be hard....
Finding max/min on boundary.
Find absolute max/min of f (x, y) = x2 − y 2 on D = {(x, y) | x2 + y 2 ≤ 1}. Critical points on D: ∇f = h2x, −2yi = h0, 0i
go.illinois.edu/math241fa17. We can parameterize the boundary, and solve... in general this can be hard....
Finding max/min on boundary.
Find absolute max/min of f (x, y) = x2 − y 2 on D = {(x, y) | x2 + y 2 ≤ 1}. Critical points on D: ∇f = h2x, −2yi = h0, 0i ⇒ (x, y) = (0, 0), and f (0, 0) = 0.
go.illinois.edu/math241fa17. Finding max/min on boundary.
Find absolute max/min of f (x, y) = x2 − y 2 on D = {(x, y) | x2 + y 2 ≤ 1}. Critical points on D: ∇f = h2x, −2yi = h0, 0i ⇒ (x, y) = (0, 0), and f (0, 0) = 0.
We can parameterize the boundary, and solve... in general this can be hard....
go.illinois.edu/math241fa17. Level curves of f (x, y) = x2 − y 2 and boundary circle of disk.
Look for points of tangency between circle and level curves:
Tangent level curves
go.illinois.edu/math241fa17. Look for points of tangency between circle and level curves:
Tangent level curves
Level curves of f (x, y) = x2 − y 2 and boundary circle of disk.
go.illinois.edu/math241fa17. Tangent level curves
Level curves of f (x, y) = x2 − y 2 and boundary circle of disk.
Look for points of tangency between circle and level curves:
go.illinois.edu/math241fa17. Max/min of f (x, y) constrained to g(x, y) = 0.
Question: How do we find tangencies without picture?
Answer: Find solutions to system:
∇f (x, y) = λ∇g(x, y) g(x, y) = 0
Call a solution (x, y) of this system a critical point of f constrained by g = 0. (we don’t typically care what λ is...)
Finding tangencies
go.illinois.edu/math241fa17. Question: How do we find tangencies without picture?
Answer: Find solutions to system:
∇f (x, y) = λ∇g(x, y) g(x, y) = 0
Call a solution (x, y) of this system a critical point of f constrained by g = 0. (we don’t typically care what λ is...)
Finding tangencies
Max/min of f (x, y) constrained to g(x, y) = 0.
go.illinois.edu/math241fa17. Answer: Find solutions to system:
∇f (x, y) = λ∇g(x, y) g(x, y) = 0
Call a solution (x, y) of this system a critical point of f constrained by g = 0. (we don’t typically care what λ is...)
Finding tangencies
Max/min of f (x, y) constrained to g(x, y) = 0.
Question: How do we find tangencies without picture?
go.illinois.edu/math241fa17. Call a solution (x, y) of this system a critical point of f constrained by g = 0. (we don’t typically care what λ is...)
Finding tangencies
Max/min of f (x, y) constrained to g(x, y) = 0.
Question: How do we find tangencies without picture?
Answer: Find solutions to system:
∇f (x, y) = λ∇g(x, y) g(x, y) = 0
go.illinois.edu/math241fa17. Finding tangencies
Max/min of f (x, y) constrained to g(x, y) = 0.
Question: How do we find tangencies without picture?
Answer: Find solutions to system:
∇f (x, y) = λ∇g(x, y) g(x, y) = 0
Call a solution (x, y) of this system a critical point of f constrained by g = 0. (we don’t typically care what λ is...)
go.illinois.edu/math241fa17. Method of Lagrange multipliers
Theorem. Suppose ∇g 6= 0 on the set
n C = {(x1,..., xn) ∈ R | g(x1,..., xn) = 0} and suppose f (x1,..., xn) is differentiable. Then if f attains a max/min value on C, these must occur at critical points of f constrained to C; that is, points (x1,..., xn) such that ∇f (x1,..., xn) = λ∇g(x1,..., xn) g(x1,..., xn) = 0 for some real number λ.
go.illinois.edu/math241fa17.