15.6 Directional Derivatives and the Gradient Vector Definition (Directional derivative) Given a function f (x, y) and a unit vector u =< a, b >, we define the directional derivative of f at v = (xo , yo ) in the direction u as Du f (xo , yo ) = lim h→0 f (v + hu) − f (v) h f (xo + ha, yo + hb) − f (xo , yo ) h→0 h = lim provided that the limit exists. Theorem If f is differentiable at (xo , yo ), then f has the directional derivative in the direction of any unit vector u =< a, b > at (xo , yo ) and Du f (xo , yo ) = fx (xo , yo )a + fy (xo , yo )b = < fx (xo , yo ), fy (xo , yo ) > · < a, b > (Proof) Let φ(h) = f (xo + ha, yo + hb). Then Du f (xo , yo ) = lim h→0 φ(h) − φ(0) h = φ0 (0) = a = afx (xo , yo ) + bfy (xo , yo ) ∂f ∂f (xo + ha, yo + hb) + b (xo + ha, yo + hb) ∂x ∂y h=0 Definition (The gradient vector ∇f ) Given a function z = f (x, y), we define the gradient of f to be the vector ∇f =< fx (x, y), fy (x, y) >= fx i + fy j Property Du f (x, y) = ∇f (x, y) · u Definition (Directional derivative and gradient vector of a function with three variables) Consider a function f (x, y, z). (a) When u =< a, b, c > is a unit vector, we define the directional derivative of f at v =< xo , yo , zo > to be f (v + hu) − f (v) f (xo + ah, yo + bh, zo + ch) − f (xo , yo , zo ) = lim h→0 h→0 h h lim provided that the limit exists. (b) We also define the gradient vector of f by 1 ∇f =< fx , fy , fz >= fx i + fy j + fz k Property Du f (xo , yo , zo ) = ∇f (xo , yo , zo ) · u Example Find the directional derivative of f at the given point in the indicated direction. (a) f (x, y) = x sin(xy) ; (2, 0), in the direction θ = π/3 (b) f (x, y) = ln(x2 + xy + y 2 ) ; (2, 1), in the direction v =< −1, 2 > (Answers) (a) ∇f =< sin(xy) + xy cos(xy), x2 cos(xy) > : u =< cos π/3, sin π/3 > : ∇f = 2x + y x + 2y , 2 x2 + xy + y 2 + xy + y (b) √ √ u =< −1/ 5, 2/ 5 > x2 ∇f (2, 0) =< 0, 4 > √ 1 < 1, 3 > 2 √ Du f (2, 0) = ∇f (2, 0) · u = 2 3 u= : ∇f (2, 1) = h5/7, 4/7i Du f (2, 1) = ∇f (2, 1) · u = 3 √ 7 5 Theorem Suppose that f is a differentiable function with two(or three) variables. Let x be a point in the domain of f and u be a unit vector in R2 (or R3 ). (a) Then |Du f (x)| = |∇f (x) · u| = k∇f (x)k| cos θ| ≤ k∇f (x)k where θ is the angle between ∇f (x) and u. (b) Particularly, notice that Du f (x) is maximized when ∇f (x) and u are in the same direction and the maximum value is k∇f (x)k. Example Find the maximum rate of change of f and the direction in which it occurs at the given point. (a) f (p, q) = qe−p + pe−q , (0, 0) (b) f (x, y, z) = tan−1 (x + 2y + 3z), (1, 1, 1) (Answers) At (a,b), maximum rate occurs in the direction of ∇f (a, b) and the maximum rate is k∇f (a, b)k. (a) ∇f =< fp , fq >=< −qe−p + e−q , e−p − pe−q > ; ∇f (0, 0) =< 1, 1 >. √ Maximum rate is 2 and it occurs in the direction of ∇f (0, 0) =< 1, 1 > 1 2 3 , , 1 + (x + 2y + 3z)2 1 + (x + 2y + 3z)2 1 + (x + 2y + 3z)2 √ Maximum rate is 14/37 and it occurs in the direction of < 1, 2, 3 > (b) ∇f = ; ∇f (1, 1, 1) =< 1/37, 2/37, 3/37 >. Example Find all points at which the direction of the fastest change of the function f (x, y) = x2 + y 2 − 2x − 4y is < 1, 1 >. (Answer) Look for (x, y) such that < fx , fy > / / < 1, 1 >. From < 2x − 2, 2y − 4 >= k < 1, 1 >, 2x − 2 = k = 2y − 4, and the answer is {(x, y) : x − y = −1}. Theorem (a) Given a function y = f (x), let g(x, y) = f (x) − y. Then 2 ∇g(xo , yo ) ⊥ Tangent line to y = f (x) at (xo , yo ) (b) Given a function z = f (x, y), let g(x, y, z) = f (x, y) − y. Then ∇g(xo , yo , zo ) ⊥ Tangent plane to z = f (x, y) at (xo , yo , zo ) (Proofs) (a) The slope of the tangent line to y = f (x) at (xo , yo ) is f 0 (xo ). Hence the vector < 1, f 0 (xo ) > is parallel to the tangent line. On the other hand, ∇g(xo , yo ) =< gx (xo , yo ), gy (xo , yo ) >=< f 0 (xo ), −1 >. Because < 1, f 0 (xo ) > · < f 0 (xo ), −1 >= 0, (a) is proven. (b) It suffices to show that ∇g(xo , yo , zo ) is perpendicular to any curve on the surface z = f (x, y) through (xo , yo , zo ). Suppose < x(t), y(t), z(t) > is a curve on the surface z = f (x, y) such that (xo , yo , zo ) = (x(to ), y(to ), z(to )). ⇒ z(t) = f (x(t), y(t)) ⇒ g(x(t), y(t), z(t)) = f (x(t), y(t)) − z(t) = 0 dg = dt ∂g ⇒ , ∂x ∂g ⇒ , ∂x ⇒ ∂g dx ∂g dy ∂g dz + + =0 ∂x dt ∂y dt ∂z dt ∂g ∂g dx dy dz , · , , =0 ∂y ∂z dt dt dt ∂g ∂g dx dy dz , ⊥ , , That is, ∇g(xo , yo , zo ) ⊥ < x0 (to ), y 0 (to ), z 0 (to ) >. ∂y ∂z dt dt dt Example Use gradient vector to find (a)an equation of the tangent plane and (b)an equation of the normal line to the surface x − z = 4 arcsin(yz) at (x, y, z) = (1, 0, 1) (Answers) Let g(x, y, z) = 4 arcsin(yz) − x + z. Then the vector ∇g(1, 0, 1) =< −1, 4, 1 > is perpendicular to the tangent plane to the surface x − z = 4 arcsin(yz) at (1, 0, 1). Hence (a) tangent plane : −1(x − 1) + 4(y − 0) + (z − 1) = 0 x−1 y−0 z−1 (b) normal line : = = −1 4 1 3