Directional Derivatives and Gradient Vectors (Sect. 14.5) .: Remark

Directional derivatives and gradient vectors (Sect. 14.5).
Directional derivative of functions of two variables. Partial derivatives and directional derivatives. Directional derivative of functions of three variables. The gradient vector and directional derivatives. Properties of the the gradient vector.
Directional derivative of functions of two variables.

Remark: The directional derivative generalizes the partial
derivatives to any direction.
Denition
The directional derivative of the function f : D R2 R at the point P0 = (x0 , y0 ) D in the direction of a unit vector u = ux , uy is given by Du f
P0
= lim
1 f (x0 + ux t , y0 + uy t ) f (x0 , y0 ) , t 0 t
if the limit exists.
Notation: The directional derivative is also denoted as

df dt
u,P 0
Directional derivatives generalize partial derivatives.

Example
The partial derivatives fx and fy are particular cases of directional derivatives Du f P = limt 0 1 t f (x0 + ux t , y0 + uy t ) f (x0 , y0 ) :
0
u = 1, 0 = i, then Di f u = 0, 1 = j , then Dj f
P0
= fx (x0 , y0 ). = fy (x0 , y0 ).
P0
f ( x, y )
f ( x, y0 )
fx (x0, y0 )
f (x,y)
f y (x0, y0 )
( Du f )P
f ( x0, y )
11111111111111 00000000000000 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 1111111 0000000 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 1 0 1 0
( x0, y0 )
00000000000000 11111111111111 11111111111111 00000000000000 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111
0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 1 0 0 1
P0 = (x0 , y 0)
|u|=1
Directional derivative of functions of two variables.

Remark: The condition |u| = 1 implies that the parameter t in the
line r(t ) = x0 , y0 + u t is the distance between the points (x (t ), y (t )) = (x0 + ux t , y0 + uy t ) and (x0 , y0 ).
Proof.
d = | x x0 , y y0 |, = | ux t , uy t |, = |t | |u|, that is, d = |t |.
Remark: The directional derivative of f (x , y ) at P0 = (x0 , y0 )

along u, denoted as Du f P , is the pointwise rate of change of f 0 with respect to the distance along the line parallel to u passing through (x0 , y0 ).
Directional derivative and partial derivatives.

z f (x,y)
Remark: The directional

derivative Du f P is the 0 derivative of f along the line r(t ) = x0 , y0 + u t .
( Du f )P
0
00000000000000 11111111111111 11111111111111 00000000000000 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111 00000000000000 11111111111111
0000000 1111111 1111111 0000000 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 0000000 1111111 1 0 1 0
P0 = (x0 , y 0)
|u|=1
Theorem
If the function f : D R2 R is dierentiable at P0 = (x0 , y0 ) and u = ux , uy is a unit vector, then Du f
P0
= fx (x0 , y0 ) ux + fy (x0 , y0 ) uy .

Proof.
The line r(t ) = x0 , y0 + ux , uy t has parametric equations: x (t ) = x0 + ux t and y (t ) = y0 + uy t ; (t ) = f (x (t ), y (t )). Denote f evaluated along the line as f (0) = Du f , since Now, on the one hand, f P
0
(t ) f (0) , (0) = lim 1 f f t 0 t 1 = lim f (x0 + ux t , y0 + uy t ) f (x0 , y0 ) , t 0 t = Du f (x0 , y0 ). On the other hand, the chain rule implies: (0) = fx (x0 , y0 ) x (0) + fy (x0 , y0 ) y (0). f Therefore, Du f
P0
= fx (x0 , y0 ) ux + fy (x0 , y0 ) uy .

Example
Compute the directional derivative of f (x , y ) = sin(x + 3y ) at the point P0 = (4, 3) in the direction of vector v = 1, 2 . Solution: We need to nd a unit vector in the direction of v. v 1 Such vector is u = u = 1, 2 . |v| 5 We now use the formula Du f = fx (x0 , y0 ) ux + fy (x0 , y0 ) uy . That is, Du f P = cos(x0 + 3y0 )(1/ 5) + 3 cos(x0 + 3y0 )(2/ 5). 0 Equivalently, Du f P = (7/ 5) cos(x0 + 3y0 ). 0 Then , Du f P = (7/ 5) cos(10).
P0 0
Directional derivative of functions of three variables.

Denition
The directional derivative of the function f : D R3 R at the point P0 = (x0 , y0 , z0 ) D in the direction of a unit vector u = ux , uy , uz is given by Du f
P0
= lim
1 f (x0 + ux t , y0 + uy t , z0 + uz t ) f (x0 , y0 , z0 ) , t 0 t
if the limit exists.
Theorem
If the function f : D R3 R is dierentiable at P0 = (x0 , y0 , z0 ) and u = ux , uy , uz is a unit vector, then Du f
P0
= fx (x0 , y0 , z0 ) ux + fy (x0 , y0 , z0 ) uy + fz (x0 , y0 , z0 ) uz .
Directional derivative of functions of three variables.

Example
Find Du f P for f (x , y , z ) = x 2 + 2y 2 + 3z 2 at the point 0 P0 = (3, 2, 1) along the direction given by v = 2, 1, 1 . Solution: We rst nd a unit vector along v, u= v |v| 1 u = 2, 1, 1 . 6
Then, Du f
is given by Du f = (2x )ux + (4y )uy + (6z )uz .

P0
We conclude, Du f that is, Du f
2 1 1 = (6) + (8) + (6) , 6 6 6
P0
26 = . 6
The gradient vector and directional derivatives.
Remark: The directional derivative of a function can be written in

terms of a dot product. In the case of 2 variable functions: Du f = fx ux + fy uy Du f = (f ) u, with f = fx , fy .
In the case of 3 variable functions: Du f = fx ux + fy uy + fz uz , Du f = (f ) u, with f = fx , fy , fz .

Denition
The gradient vector of a dierentiable function f : D R2 R at any point (x , y ) D is the vector f = fx , fy . The gradient vector of a dierentiable function f : D R3 R at any point (x , y , z ) D is the vector f = fx , fy , fz .
Notation:
For two variable functions: f = fx i + fy j . For two variable functions: f = fx i + fy j + fz k.
Theorem
If f : D Rn R, with n = 2, 3, is a dierentiable function and u is a unit vector, then, Du f = (f ) u.

Example
Find the gradient vector at any point in the domain of the function f (x , y ) = x 2 + y 2 . Solution: The gradient is f = fx , fy , that is, f = 2x , 2y .
f(x,y) = x 2 + y 2
y f
Remark:
f = 2r, with r = x, y .
y x
Duf
Properties of the the gradient vector.

Remark: If is the angle between f and u, then holds
Du f = f u = |f | cos().
The formula above implies: The function f increases the most rapidly when u is in the direction of f , that is, = 0. The maximum increase rate of f is |f |. The function f decreases the most rapidly when u is in the direction of f , that is, = . The maximum decrease rate of f is |f |. The function f does not change along level curve or surfaces, that is, Du f = 0. Therefore, f is perpendicular to the level curves or level surfaces.

Example
Find the direction of maximum increase of the function f (x , y ) = x 2 /4 + y 2 /9 at an arbitrary point (x , y ), and also at the points (1, 0) and (0, 1). Solution: The direction of maximum increase of f is the direction of its gradient vector: f = x 2y , . 2 9
At the points (1, 0) and (0, 1) we obtain, respectively, f = 1 ,0 . 2 f = 0, 2 . 9

Example
Given the function f (x , y ) = x 2 /4 + y 2 /9, nd the equation of a line tangent to a level curve f (x , y ) = 1 at the point P0 = (1, 3 3/2). Solution: We rst verify that P0 belongs to the level curve f (x , y ) = 1. This is the case, since 1 (9)(3) 1 + = 1. 4 4 9 The equation of the line we look for is 3 3 + t vx , vy , r(t ) = 1, 2 where v = vx , vy is tangent to the level curve f (x , y ) = 1 at P0 .

Example
Given the function f (x , y ) = x 2 /4 + y 2 /9, nd the equation of a line tangent to a level curve f (x , y ) = 1 at the point P0 = (1, 3 3/2). Solution: Therefore, v f at P0 . Since, x 2y f = , 2 9 Therefore, 0 = v f
P0
P0
1 23 3 1 1 = , = , . 2 9 2 2 3
1 1 vx = vy 2 3
v = 2, 3 .
3 3 The line is r(t ) = 1, + t 2, 3 . 2

y
f
3
3
1 f = ,0 , 2
2 f = 0, , 9
3 3 r(t ) = 1, + t 2, 3 . 2
Further properties of the the gradient vector.
Theorem
If f , g are dierentiable scalar valued vector functions, g = 0, and k R any constant, then holds, 1. (kf ) = k (f ); 2. (f g ) = f g ; 3. (fg ) = (f ) g + f (g ); f (f ) g f (g ) 4. = . g g2
Tangent planes and linear approximations (Sect. 14.6).
Review: Dierentiable functions of two variables. The tangent plane to the graph of a function. The linear approximation of a dierentiable function. Bounds for the error of a linear approximation. The dierential of a function.
Review: Scalar functions of one variable. Scalar functions of more than one variable.
Review: Dierentiable functions of two variables.

Denition
Given a function f : D R2 R and an interior point (x0 , y0 ) D , let L be the linear function L(x , y ) = fx (x0 , y0 ) (x x0 ) + fy (x0 , y0 ) (y y0 ) + f (x0 , y0 ). The function f is called dierentiable at (x0 , y0 ) i the function f is approximated by the linear function L near (x0 , y0 ), that is, f (x , y ) = L(x , y ) + where the functions
1 1 (x
x0 ) +
2 (y
y0 )
and
0 as (x , y ) (x0 , y0 ).
Theorem
If the partial derivatives fx and fy of a function f : D R2 R are continuous in an open region R D, then f is dierentiable in R.
Review: Dierentiable functions of two variables.

Example
Show that the function f (x , y ) = x 2 + y 2 is dierentiable for all (x , y ) R2 . Furthermore, nd the linear function L, mentioned in the denition of a dierentiable function, at the point (1, 2). Solution: The partial derivatives of f are given by fx (x , y ) = 2x and fy (x , y ) = 2y , which are continuous functions. Therefore, the function f is dierentiable. The linear function L at (1, 2) is L(x , y ) = fx (1, 2) (x 1) + fy (1, 2) (y 2) + f (1, 2). That is, we need three numbers to nd the linear function L: fx (1, 2), fy (1, 2), and f (1, 2). These numbers are: fx (1, 2) = 2, fy (1, 2) = 4, f (1, 2) = 5.
Therefore, L(x , y ) = 2(x 1) + 4(y 2) + 5.
The tangent plane to the graph of a function.

Remark:
The function L(x , y ) = 2(x 1) + 4(y 2) + 5 is a plane in R3 . We usually write down the equation of a plane using the notation z = L(x , y ), that is, z = 2(x 1) + 4(y 2) + 5, or equivalently 2(x 1) + 4(y 2) (z 5) = 0. 0 = (1, 2, 5) with normal vector This is a plane passing through P n = 2, 4, 1 . Analogously, the function L(x , y ) = fx (x0 , y0 )(x x0 ) + fy (x0 , y0 )(y y0 ) + f (x0 , y0 ) is a plane in R3 . Using the notation z = L(x , y ) we obtain fx (x0 , y0 )(x x0 ) + fy (x0 , y0 )(y y0 ) (z f (x0 , y0 )) = 0. 0 = (x0 , y0 , f (x0 , y0 )) with normal This is a plane passing through P vector n = fx (x0 , y0 ), fy (x0 , y0 ), 1 .

Theorem
The plane tangent to the graph of a dierentiable function f : D R2 R at the point (x0 , y0 ) is given by L(x , y ) = fx (x0 , y0 ) (x x0 ) + fy (x0 , y0 ) (y y0 ) + f (x0 , y0 ).
Proof
f ( x, y ) = z n P0 L(x,y)
The plane contains the point 0 = (x0 , y0 , f (x0 , y0 )). We only P need to nd its normal vector n.
x
Plane that approximates f ( x, y ) at P0
y P
0

f ( x, y ) = z n P0 L(x,y)
The vector n normal to the plane L(x , y ) is a vector perpendicular to the surface z = f (x , y ) at P0 = (x0 , y0 ).
x
Plane that approximates f ( x, y ) at P0
y P
0
This surface is the level surface F (x , y , z ) = 0 of the function F (x , y , z ) = f (x , y ) z . A vector normal to this level surface is its gradient F . That is, F = Fx , Fy , Fz = fx , fy , 1 . Therefore, the normal to the tangent plane L(x , y ) at the point P0 is n = fx (x0 , y0 ), fy (x0 , y0 ), 1 . Recall that the plane contains 0 = (x0 , y0 , f (x0 , y0 )). The equation for the plane is the point P fx (x0 , y0 )(x x0 ) + fy (x0 , y0 )(y y0 ) (z f (x0 , y0 )) = 0.

Summary: We have shown that the linear L given by
L(x , y ) = fx (x0 , y0 ) (x x0 ) + fy (x0 , y0 ) (y y0 ) + f (x0 , y0 ) is the plane tangent to the graph of f at (x0 , y0 ).
f ( x, y ) = z L(x,y) Plane that approximates f ( x, y ) at ( x0 , y0 )
Plane that does not approximate f (x,y) near ( 0 , 0 ).
L(x,y)
f(x,y)
1
y ( x 0 , y0 ) x Function f is differentiable at (x0 , y 0 ).
Function f is not differentiable at ( 0 , 0 ).
Remark: The graph of a dierentiable function f : D R2 R is

approximated by the tangent plane L at every point in D .

Example
Show that f (x , y ) = arctan(x + 2y ) is dierentiable and nd the plane tangent to f (x , y ) at (1, 0). Solution: The partial derivatives of f are given by fx (x , y ) = 1 , 1 + (x + 2y )2 fy (x , y ) = 2 . 1 + (x + 2y )2
These functions are continuous in R2 , so f (x , y ) is dierentiable at every point in R2 . The plane L(x , y ) at (1, 0) is given by L(x , y ) = fx (1, 0)(x 1) + fy (1, 0)(y 0) + f (1, 0), where f (1, 0) = arctan(1) = /4, fx (1, 0) = 1/2, fy (1, 0) = 1. 1 Then, L(x , y ) = (x 1) + y + . 2 4
The linear approximation of a dierentiable function.

Denition
The linear approximation of a dierentiable function f : D R2 R at the point (x0 , y0 ) D is the plane L(x , y ) = fx (x0 , y0 ) (x x0 ) + fy (x0 , y0 ) (y y0 ) + f (x0 , y0 ).
Example
Find the linear approximation of f = 17 x 2 4y 2 at (2, 1). Solution: L(x , y ) = fx (2, 1)(x 2) + fy (2, 1)(y 1) + f (2, 1). We need three numbers: f (2, 1), fx (2, 1), and fy (2, 1). These are: f (2, 1) = 3, fx (2, 1) = 2/3, and fy (2, 1) = 4/3. 2 4 Then the plane is given by L(x , y ) = (x 2) (y 1) + 3. 3 3
Bounds for the error of a linear approximation.
Theorem
Assume that the function f : D R2 R has rst and second partial derivatives continuous on an open set containing a rectangular region R D centered at the point (x0 , y0 ). If M R is the upper bound for |fxx |, |fyy |, and |fxy | in R, then the error E (x , y ) = f (x , y ) L(x , y ) satises the inequality |E (x , y )| 1 2 M |x x0 | + |y y0 | , 2
where L(x , y ) is the linearization of f at (x0 , y0 ), that is, L(x , y ) = fx (x0 , y0 ) (x x0 ) + fy (x0 , y0 ) (y y0 ) + f (x0 , y0 ).
Bounds for the error of a linear approximation.

Example
Find an upper bound for the error in the linear approximation of f (x , y ) = x 2 + y 2 at the point (1, 2) over the rectangle R = {(x , y ) R2 : |x 1| < 0.1, |y 2| < 0.1}
Solution: The second derivatives of f are fxx = 2, fyy = 2, fxy = 0. Therefore, we can take M = 2. 1 2 Then the formula |E (x , y )| M |x x0 | + |y y0 | , implies 2 |E (x , y )| |x 1| + |y 2|
2
< (0.1 + 0.1)2 = 0.04,
that is |E (x , y )| < 0.04. Since f (1, 2) = 5, the percentage relative error 100 E (x , y )/f (1, 2) is bounded by 0.8%
Review: Dierential of functions of one variable.

Denition
The dierential at x0 D of a dierentiable function f : D R R is the linear function df (x ) = L(x ) f (x0 ).
Remark: The linear approximation of f (x ) at x0 is the line given

by L(x ) = f (x0 ) (x x0 ) + f (x0 ). Therefore
y f(x) L(x) f
df (x ) = f (x0 ) (x x0 ). Denoting dx = x x0 , df = f (x0 ) dx .

df f(x0)
x0 dx = x
Dierential of functions of more than one variable.

Denition
The dierential at (x0 , y0 ) D of a dierentiable function f : D R2 R is the linear function df (x , y ) = L(x , y ) f (x0 , y0 ).
Remark: The linear approximation of f (x , y ) at (x0 , y0 ) is the

plane L(x , y ) = fx (x0 , y0 ) (x x0 ) + fy (x0 , y0 ) (y y0 ) + f (x0 , y0 ). Therefore df (x , y ) = fx (x0 , y0 ) (x x0 ) + fy (x0 , y0 ) (y y0 ). Denoting dx = x x0 and dy = (y y0 ) we obtain the usual expression df = fx (x0 , y0 ) dx + fy (x0 , y0 ) dy . Therefore, df and L are similar concepts: The linear approximation of a dierentiable function f .

Example
Compute the df of the function f (x , y ) = ln(1 + x 2 + y 2 ) at the point (1, 1). Evaluate this df for dx = 0.1, dy = 0.2. Solution: The dierential of f at (x0 , y0 ) is given by df = fx (x0 , y0 )dx + fy (x0 , y0 )dy . The partial derivatives fx and fy are given by fx (x , y ) = 2x , 1 + x2 + y2 fy (x , y ) = 2y . 1 + x2 + y2
2 Therefore, fx (1, 1) = 2/3 = fy (1, 1). Then df = 2 3 dx + 3 dy . Evaluating this dierential at dx = 0.1 and dy = 0.2 we obtain
df =
2 1 2 2 2 3 + = 3 10 3 10 3 10
1 df = . 5

Example
Use dierentials to estimate the amount of aluminum needed to build a closed cylindrical can with internal diameter of 8cm and height of 12cm if the aluminum is 0.04cm thick. Solution: The data of the problem is: h0 = 12cm, r0 = 4cm, dr = 0.04cm and dh = 0.08cm. The function to consider is the mass of the cylinder, M = V , where = 2.7gr /cm3 is the aluminum density and V is the volume of the cylinder, V (r , h) = r 2 h. The metal to build the can is given by M = V (r + dr , h + dh) V (r , h) , recall dh = 2dr .
h 0 = 12 dr = 0.04
r0 = 4

Example
Use dierentials to estimate the amount of aluminum needed to build a closed cylindrical can with internal diameter of 8cm and height of 12cm if the aluminum is 0.04cm thick. Solution: The metal to build the can is given by M = V (r + dr , h + dh) V (r , h) , A linear approximation to V = V (r + dr , h + dh) V (r , h) is dV = Vr dr + Vh dh, that is, dV = Vr (r0 , h0 )dr + Vh (r0 , h0 )dh.
2 dh. Since V (r , h) = r 2 h, we obtain dV = 2 r0 h0 dr + r0
Therefore, dV = 16.1 cm3 . Since dM = dV , a linear estimate for the aluminum needed to build the can is dM = 43.47 gr .
Local and absolute extrema, saddle points (Sect. 14.7).
Review: Local extrema for functions of one variable. Denition of local extrema. Characterization of local extrema.
First derivative test. Second derivative test.
Absolute extrema of a function in a domain.
Review: Local extrema for functions of one variable.

Recall: Main results on local extrema for f (x ):
f(x)
at a b c d
a b c d x
f max. in. min. in.
f 0 =0 0 =0
f <0 0 >0 0
Remarks: Assume that f is twice continuously dierentiable.

If x0 is local maximum or minimum of f , then f (x0 ) = 0. If f (x0 ) = 0 then x0 is a critical point of f , that is, x0 is a maximum or a minimum or an inection point. The second derivative test determines whether a critical point is a maximum, minimum or an inection point.
Denition of local extrema for functions of two variables.

Denition
A function f : D R2 R has a local maximum at the point (a, b ) D i holds that f (x , y ) f (a, b ) for every point (x , y ) in a neighborhood of (a, b ). A function f : D R2 R has a local minimum at the point (a, b ) D i holds that f (x , y ) f (a, b ) for every point (x , y ) in a neighborhood of (a, b ).
11 00 00 11 00 11
00 11 11 00 00 11
Denition of local extrema for functions of two variables.

Denition
A dierentiable function f : D R2 R has a saddle point at an interior point (a, b ) D i in every open disk in D centered at (a, b ) there always exist points (x , y ) where f (x , y ) > f (a, b ) and other points (x , y ) where f (x , y ) < f (a, b ).
11 00 00 11 00 11
Characterization of local extrema.

First derivative test.
Theorem
If a dierentiable function f has a local maximum or minimum at (a, b ) then holds f (a,b) = 0, 0 .
Remark: The tangent plane at a local extremum is horizontal,

since its normal vector is n = fx , fy , 1 = 0, 0, 1 .
Denition
The interior point (a, b ) D of a dierentiable function f : D R2 R is a critical point of f i f (a,b) = 0, 0 .
Remark:
Critical points include local maxima, local minima, and saddle points.
11 00 11 00 11 00
11 00 11 00 11 00
11 00 00 11 00 11

First derivative test.
Example
Find the critical points of the function f (x , y ) = x 2 y 2 Solution: The critical points are the points where f vanishes. Since f = 2x , 2y , the only solution to f = 0, 0 is x = 0, y = 0. That is, (a, b ) = (0, 0). 0 for all (x , y ) R2 and f (0, 0) = 0, then the point (0, 0) must be a local maximum of f .
Remark: Since f (x , y )
Example
Find the critical points of the function f (x , y ) = x 2 y 2 Solution: Since f = 2x , 2y , the only solution to f = 0, 0 is x = 0, y = 0. That is, we again obtain (a, b ) = (0, 0).

Second derivative test.
Theorem
Let (a, b ) be a critical point of f : D R2 R, that is, f (a,b) = 0, 0 . Assume that f has continuous second derivatives in an open disk in D with center in (a, b ) and denote D = fxx (a, b ) fyy (a, b ) fxy (a, b ) . Then, the following statements hold: If D > 0 and fxx (a, b ) > 0, then f (a, b ) is a local minimum. If D > 0 and fxx (a, b ) < 0, then f (a, b ) is a local maximum. If D < 0, then f (a, b ) is a saddle point. If D = 0 the test is inconclusive.
2
Notation: The number D is called the discriminant of f at (a, b).

Example
Find the local extrema of f (x , y ) = y 2 x 2 and determine whether they are local maximum, minimum, or saddle points. Solution: We rst nd the critical points: f = 2x , 2y f
(a ,b )
= 0, 0
i (a, b ) = (0, 0).
The only critical point is (a, b ) = (0, 0). We need to compute D = fxx (a, b ) fyy (a, b ) fxy (a, b ) . Since fxx (0, 0) = 2, fyy (0, 0) = 2, and fxy (0, 0) = 0, we get D = (2)(2) = 4 < 0 saddle point at (0, 0).
2

Example
Is the point (a, b ) = (0, 0) a local extrema of f (x , y ) = y 2 x 2 ? Solution: We rst verify that (0, 0) is a critical point of f : f (x , y ) = 2xy 2 , 2yx 2 , therefore, (0, 0) is a critical point. f
(0,0)
= 0, 0 ,
Remark: The whole axes x = 0 and y = 0 are critical points of f .

We need to compute D = fxx (a, b ) fyy (a, b ) fxy (a, b ) . Since fxx (x , y ) = 2y 2 , fyy (x , y ) = 2x 2 , and fxy (x , y ) = 4xy , we obtain fxx (0, 0) = 0, fyy (0, 0) = 0, and fxy (0, 0) = 0, hence D = 0 and the test is inconclusive.
2

Example
Is the point (a, b ) = (0, 0) a local extrema of f (x , y ) = y 2 x 2 ? Solution: From the graph of f = x 2 y 2 is simple to see that (0, 0) is a local minimum: (also a global minimum.)
z

Denition
A function f : D R2 R has an absolute maximum at the point (a, b ) D i f (x , y ) f (a, b ) for all (x , y ) D . A function f : D R2 R has an absolute minimum at the point (a, b ) D i f (x , y ) f (a, b ) for all (x , y ) D .
f(x) Absolute maximum Local maximum
Remark: Local extrema need not

be the absolute extrema.
a c
Local and absolute minimum
f(x)
No absolute extrema Local minimum Local maximum
Remark: Absolute extrema may

not be dened on open intervals.
a
Review: Functions of one variable.

Theorem
Every continuous functions f : [a, b ] R R, with a < b R always has absolute extrema.
f(x) Absolute maximum Local maximum Local and absolute minimum
Recall:
Intervals [a, b ] are bounded and closed sets in R. The set [a, b ] is closed, since the boundary points belong to the set, and it is bounded, since it does not extend to innity.
Recall: On open and closed sets in Rn .

Denition
A set S Rn , with n N, is called open i every point in S is an interior point. The set S is called closed i S contains its boundary. A set S is called bounded i S is contained in ball, otherwise S is called unbounded.
y
000000000000 111111111111 111111111111 000000000000 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111 000000000000 111111111111
y closed and unbounded
open and bounded
111111111111111 000000000000000 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111 000000000000000 111111111111111
bounded x
1111111111111111111111111111111 0000000000000000000000000000000 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111 0000000000000000000000000000000 1111111111111111111111111111111
closed and bounded
Theorem
If f : D R2 R is continuous in a closed and bounded set D, then f has an absolute maximum and an absolute minimum in D.
Absolute extrema on closed and bounded sets.

Problem:
Find the absolute extrema of a function f : D R2 R in a closed and bounded set D . Solution: (1) Find every critical point of f in the interior of D and evaluate f at these points. (2) Find the boundary points of D where f has local extrema, and evaluate f at these points. (3) Look at the list of values for f found in the previous two steps. If f (x0 , y0 ) is the biggest (smallest) value of f in the list above, then (x0 , y0 ) is the absolute maximum (minimum) of f in D .

Example
Find the absolute extrema of the function f (x , y ) = 3 + xy x + 2y on the closed domain given in the Figure.
y 4
II
III
Solution: (1) We nd all critical points in the interior of the domain: f = (y 1), (x + 2) = 0, 0 (x0 , y0 ) = (2, 1).
Since (2, 1) does not belong to the domain, we discard it. (2) Three segments form the boundary of D : Boundary I: The segment y = 0, x [1, 5]. We select the end points (1, 0), (5, 0), and we record: f (1, 0) = 2 and f (5, 0) = 2. We look for critical point on the interior of Boundary I: Since g (x ) = f (x , 0) = 3 x , so g = 1 = 0. No critical points in the interior of Boundary I.

Example
Find the absolute extrema of the function f (x , y ) = 3 + xy x + 2y on the closed domain given in the Figure.
y 4
II
III
Solution: Boundary II: The segment x = 1, y [0, 4]. We select the end point (1, 4) and we record: f (1, 4) = 14. We look for critical point on the interior of Boundary II: Since g (y ) = f (1, y ) = 3 + y 1 + 2y = 2 + 3y , so g = 3 = 0. No critical points in the interior of Boundary II. Boundary III: The segment y = x + 5, x [1, 5]. We look for critical point on the interior of Boundary III: Since g (x ) = f (x , x + 5) = 3 + x (x + 5) x + 2(x + 5). We obtain g (x ) = x 2 + 2x + 13, hence g (x ) = 2x + 2 = 0 implies x = 1. So, y = 4, and we selected the point (1, 4), which was already in our list. No critical points in the interior of Boundary III.

Example
Find the absolute extrema of the function f (x , y ) = 3 + xy x + 2y on the closed domain given in the Figure. Solution: (3) Our list of values is: f (1, 0) = 2 We conclude: Absolute maximum at (1, 4), Absolute minimum at (5, 0). f (1, 4) = 14 f (5, 0) = 2.
y
II
III
A maximization problem with a constraint.

Example
Find the maximum volume of a closed rectangular box with a given surface area A0 . Solution: This problem can be solved by nding the local maximum of an appropriate function f . The function f is obtained as follows: Recall the functions volume and area of a rectangular box with vertex at (0, 0, 0) and sides x , y and z : V (x , y , z ) = xyz , A(x , y , z ) = 2xy + 2xz + 2yz . A0 2xy , that is 2(x + y )
Since A(x , y , z ) = A0 , we obtain z =
A0 xy 2x 2 y 2 f (x , y ) = . 2(x + y )
A maximization problem with a constraint.

Example
Find the maximum volume of a closed rectangular box with a given surface area A0 . Solution: We must nd the critical points of f (x , y ) = 2A0 y 2 4x 2 y 2 8xy 3 fx = , 4(x + y )2 A0 xy 2x 2 y 2 . 2(x + y )
2A0 x 2 4x 2 y 2 8yx 3 fy = . 4(x + y )2
The conditions fx = 0 and fy = 0 and x = 0, y = 0 imply A0 = 2x 2 + 4xy , A0 = 2y 2 + 4xy , x = y . Recall z= A0 2xy , 2(x + y ) A0 /6.
A0 2x 2 so, z = = y . Therefore, x0 = y0 = z0 = 4x

Directional Derivatives and Gradient Vectors (Sect. 14.5) .: Remark

Uploaded by

Copyright:

Available Formats

Directional Derivatives and Gradient Vectors (Sect. 14.5) .: Remark

Uploaded by

Document Information

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

Directional Derivatives and Gradient Vectors (Sect. 14.5) .: Remark

Uploaded by

Copyright:

Available Formats

Directional derivatives and gradient vectors (Sect. 14.5).

Directional derivative of functions of two variables.

if the limit exists.

Notation: The directional derivative is also denoted as

Directional derivatives generalize partial derivatives.

Directional derivative of functions of two variables.

Remark: The directional derivative of f (x , y ) at P0 = (x0 , y0 )