Summary of §11 Partial Derivatives Math 1321 (Qinghai Zhang) 2013-FEB-25 Definition 1. The open ball centered at P0 ∈ Rn with Theorem 12 (Clairaut’s). If f : Rn → R has continuous radius r > 0 is the point set second partial derivatives at P , then ∀i, j = 1, 2, · · · , n, B(P0 , r) = P |P − P0 | < r . (1) ∂ 2 f (P ) ∂ 2 f (P ) = . (8) ∂xi ∂xj ∂xj ∂xi It is an open interval in 1D and an open disk in 2D. The open ball without the center is denoted by Definition 13. The linear approximation of f (x, y) at B0 (P0 , r) = B(P0 , r) \ {P0 }. (2) P0 = (x0 , y0 ) is Definition 2. P0 is a boundary point of an point set U iff ∀r > 0, ∃P ∈ B(P0 , r) s.t. P 6∈ U. ∂f ∂f (x−x0 )+ (y−y0 ). (9) f (x, y) ≈ f (x0 , y0 )+ ∂x P0 ∂y P0 Definition 3. A set is an open set if it contains none of n its boundary points. A set is an closed set if it contains Definition 14. The total differential of f : R → R is defined in terms of the differential s dxi ’s: all of its boundary points. Definition 4. A is a subset of B, written as A ⊆ B, iff x ∈ A ⇒ x ∈ B. df = n X ∂f dxi ∂x i i=1 (10) Definition 5. A point set U ⊆ Rn is bounded iff U ⊆ B(P0 , r) for some P0 ∈ Rn and r > 0. Definition 15. The increment of a function z = f (x, y) Definition 6. The limit of a function f : B (P , r) → R at a point P0 = (x0 , y0 ) is 0 0 exists as P approaches P0 , written as limP →P0 f (P ) = L, iff ∆z = f (x, y) − f (x0 , y0 ). (11) ∀ > 0, ∀ paths P → P0 , ∃δ > 0, s.t. (3) Definition 16. f : B(P0 , r) → R is differentiable at Pn0 if there exists a linear function L(r) = c · r where c ∈ R such that for all paths P → P0 , Formula 7. The limit of f does not exist at P0 if there exists two different paths P → P0 and P P0 s.t. f (P ) − f (P0 ) − L(P − P0 ) = 0. (12) lim P →P0 |P − P0 | (4) lim f (P ) = L 6= L = lim f (P ) . ∀P ∈ B(P0 , δ), 1 P →P0 |f (P ) − L| < . 2 P P0 Theorem 17. f : Rn → R is differentiable at P iff ∂f ∀i ∈ 1, 2, · · · , n, ∂x exists and is continuous at P . i Theorem 8 (The squeeze theorem). If ∃r > 0 s.t. ∀P ∈ B0 (P0 , r), f (P ) ≤ g(P ) ≤ h(P ), then lim f (P ) = lim h(P ) = L ⇒ lim g(P ) = L. P →P0 P →P0 Definition 18. A function f (x) is C 1 or continuously differentiable if f 0 (x) exists and is itself continuous. P →P0 (5) Definition 19. A C 1 function f : R2 → R admits a pair of tangent vectors along x- and y- directions for the surface z = f (x, y): n Definition 9. f : R → R is continuous at Q iff lim f (P ) = f (Q). P →Q (6) u= f is continuous on a point set U if (6) holds ∀Q ∈ U. Theorem 10. A polynomial is continuous everywhere. ∂z 1, 0, ∂x , v= ∂z 0, 1, ∂y . (13) The tangent plane of the surface at P0 is the plane that contains P0 and the tangent vectors u(P0 ) and v(P0 ). Definition 11. The partial derivative of f : Rn → R with respect to the ith dimension at P0 is ∂f (P0 ) ∂f f (P0 + hei ) − f (P0 ) = = lim , (7) ∂xi ∂xi P0 h→0 h Lemma 20. If f : R2 → R is C 1 at z0 = f (x0 , y0 ), the scalar equation of the tangent plane of the surface σ(x, y) = hx, y, f (x, y)i at (x0 , y0 , z0 ) is where h ∈ R and {ei | i = 1, 2, · · · , n} is the set of standard basis vectors of the Euclidean n-space, i.e. the ith component of ei is 1 and all other components are 0. z − z0 = 1 ∂z ∂z (x − x ) + (y − y0 ). 0 ∂x P0 ∂y P0 (14) Summary of §11 Partial Derivatives Math 1321 (Qinghai Zhang) 2013-FEB-25 Theorem 21 (The chain rule). If f : Rn → R is dif- Definition 30. f : Rn → R has a local maximum at ferentiable and each variable xi is a differentiable func- P0 ∈ Rn iff tion of m variables t1 , t2 , · · · , tm , then f is a function of ∃r > 0, s.t. ∀P ∈ B(P0 , r), f (P ) ≤ f (P0 ). (23) t1 , t2 , · · · , tm and ∀i = 1, 2, · · · , m, n Changing ≤ to ≥ in (23) yields a local minimum. (15) An extremum is either a maximum or minimum. X ∂f ∂xj ∂f = . ∂ti ∂xj ∂ti j=1 Definition 31. A critical point of f : Rn → R is a point Definition 22. A level curve of F : R2 → R is the curve P0 ∈ Rn satisfying ∇f | = 0. P0 with equation F (x, y) = c where c is a constant. A level surface of F : R3 → R is the surface with Theorem 32 (Fermat’s). Every local extremum of f : Rn → R is a critical point. equation F (x, y, z) = c where c is a constant. Theorem 23 (Implicit function theorem: 2D). Consider Formula 33 (Second derivatives test for critical points). a critical point P0 = (a, b) of f : R2 → R. If ∂F 2 F : R → R at P0 = (x0 , y0 ). If ∂y 6= 0 and F is C 1 Consider ∂2f ∂2f ∂2f P0 2 , ∂y 2 , and ∂x∂y are all continuous on B(P0 , r) for ∂x on the open disk B(P0 , r) for some r > 0, then the level some r > 0, then the discriminant curve F (x, y) = c defines a C 1 function y = y(x) on the 2 ∂ f open interval B(x0 , r), and ∂2f ∂x2 ∂x∂y (24) D = det ∂F 2 2 dy ∂ f ∂ f ∂x ∂y∂x = − ∂F . (16) 2 ∂y P0 dx ∂y might determine the D < 0, 2 D > 0, ∂∂xf2 > 0 P 0 2 D > 0, ∂∂xf2 < 0 Theorem 24 (Implicit function theorem: 3D). Con sider F : R3 → R at P0 = (x0 , y0 , z0 ). If ∂F ∂z P0 6= 0 and F is C 1 on the open ball B(P0 , r) for some r > 0, then the level surface F (x, y, z) = c defines a C 1 function z = z(x, y) on the open disk B((x0 , y0 ), r), and ∂z =− ∂x ∂F ∂x ∂F ∂z , ∂z =− ∂y ∂F ∂y ∂F ∂z P0 . type of the critical point: ⇒ f (P0 ) is not a local extremum, ⇒ f (P0 ) is a local minimum, ⇒ f (P0 ) is a local maximum. (25) (17) Theorem 34. If f : Rn → R is continuous on a closed, n Definition 25. The gradient vector of a function f : bounded set D ⊆ R , then f attains an absolute maximum value and an absolute minimum value on D. Rn → R is a vector in Rn : X n Theorem 35. Consider determining extremums of a C 1 ∂f ∂f ∂f ∂f ∇f = , ,··· , = ei , (18) function f : Rn → R under the constraint F (P ) = c ∂x1 ∂x2 ∂xn ∂xi i=1 where F : Rn → R, F ∈ C 1 , c is a constant. If f attains Definition 26. The directional derivative of a function an extremum at P0 and ∂F∂x(P0 ) 6= 0, then ∃λ ∈ R s.t. n f : Rn → R at P0 ∈ Rn in the direction of a vector n = λ ∇F |P0 . (26) ∇f | u ∈ R is the scalar P0 f (P0 + hu) − f (P0 ) 36. Consider determining extremums of a . (19) Definition h C 1 function f (x, y, z) under the constraint F (x, y, z) = c 1 Theorem 27. If f : Rn → R is differentiable at P , then where F ∈ C and c is a constant. By constructing the Lagrangian function Du f (P ) = u · ∇f (P ). (20) Λ(x, y, z, λ) = f (x, y, z) + λ c − F (x, y, z) , (27) n Theorem 28. If f : R → R is differentiable at P , then max Du f (P ) = |∇f (P )|. (21) the method of Lagrangian multipliers first find the point u∈Rn , |u|=1 set U = {(x, y, z) | ∇Λ(x, y, z, λ) = 0} by solving Lemma 29. Let z = z(x, y) be the implicit function ∇f = λ∇F, (28) defined by the level surface F (x, y, z) = c. The unit F (x, y, z) = c. normal vector of the tangent plane to the level surface and then evaluate f over U to obtain F (x, y, z) = c is ∇F n= . (22) max f = max f (P ), min f = min f (P ). (29) |∇F | P ∈U P ∈U Du f (P0 ) = lim h→0 2