MATH 321: Real Variables II Lecture #32 University of British Columbia Lecture #32: Instructor: Scribe: March 31, 2008 Dr. Joel Feldman Peter Wong Theorem. (Uniform Convergence of Fourier Series) If f : R → C is 2π-periodic and is piecewise C 1 , then the Fourier series for f converges uniformly to f on any closed interval that does not contain a point of discontinuity of f . Proof. Step 1: If, in addition, f is continuous, then the Fourier series of f converges uniformly to f . Proof. Done. Step 2: Define y y = F (x) 1 2 b −3π b −π −2π b π 2π 3π x The Fourier series converges uniformly on any closed interval that does not contain a point in 2πZ = { 2πk | k ∈ Z } Proof. Done. Step 3: The general case: − Let the points of discontinuity of f on [−π, π) be p1 , . . . , pm . Let the jump heights be δj = f (p+ j ) − f (pj ). Pm Pm Write f (x) = g(x) + j=1 δj F (x − pj ) with g(x) = f (x) − j=1 δj F (x − pj ). The Fourier series for δj F (x − pj ) converges uniformly on any closed interval not containg a point of pj + 2πZ. It suffices to show that g(x) is continuous. g(x) is obviously continuous except possibly at the pk ’s, which are in [−π, π). To check continuity at pk ’s − + − g(p+ k ) − g(pk ) = f (pk ) − f (pk ) − = δk − n X j=1 =0 δj ( n X δj [F (( pk − pj )+ ) − F ((pk − pj )− )] | {z } j=1 ∈(−2π,2π] 0, j 6= k, 1, j = k Manifolds (See web notes and references therein) (1) Definition (2) Examples (3) Integration on Manifolds & Stoke’s theorem 2 MATH 321: Lecture #32 Rough Definition • an n-dimensional manifold is something that looks locally like Rn • it is a union of subsets with each subset having coordinates that run over any open subset of Rn Definition. (An n-dimensional manifold) Let M be a metric space. (a) A chart on M is a pair { U, ϕ } with U ⊂ M open and ϕ : U → Rn a homeomorphism between U and ϕU. Think of ϕ as assigning coordinates in Rn to each m ∈ U. (b) Two charts { U, ϕ } and { V, ψ } are compatible if ψ ◦ ϕ−1 : ϕ(U ∩ V) ⊂ Rn → ψ(U ∩ V) ⊂ Rn ϕ ◦ ψ −1 : ψ(U ∩ V) ⊂ Rn → ϕ(U ∩ V) ⊂ Rn are C ∞ (i.e., all partial derivatives are defined and continuous.) MATH 321: Real Variables II Lecture #33 University of British Columbia Lecture #33: Instructor: Scribe: April 2, 2008 Dr. Joel Feldman Peter Wong Definition. (Manifold of dimension n) Let M be a metric space. (a) A chart is a pair { U, ϕ } with • U an open subset of M • ϕ a homeomorphism from U to an open subset of Rn (b) { U, ϕ } and { V, ψ } are compatible if ψ ◦ ϕ−1 and ϕ ◦ ψ −1 are C ∞ (c) An atlas A for the M is a set A = { { Ui , ϕi } | i ∈ I }, with I being a completely arbitrary index set, of charts obeying (a) { Ui } covers M (i.e., S i∈I Ui = M) (b) Every pair of charts in A is compatible. A maximal atlas is an atlas A with the property that { U, ϕ } being any chart that is compatible with every chart in A implies { U, ϕ } ∈ A. (d) A manifold of dimension n is a metric space M together with a maximal atlas A. Lemma. Let { U, ϕ } and { V, ψ } be • U, V ∈ M with U ∪ V = 6 ∅ • ϕ : U → Rn be a homeomorphism from U to ϕ(U) ⊂ Rn open • ψ : V → Rm be a homeomorphism from V to ϕ(V) ⊂ Rm open • ψ ◦ ϕ−1 and ϕ ◦ ψ −1 are C ∞ Proof. Homework problem. Lemma. If A is any atlas, then there exists a unique maximal atlas containing A Proof. Homework problem. (Hint: Try to guess what the maximal atlas is.) Example 1. Let M be an open subset of Rn (e.g., M = Rn ) and A = { { U = M, ϕ = identity map } }. Then (U, ϕ) is an identity map. Example 2. Let M = S 1 = { (x, y) ∈ Rn | x2 + y 2 = 1 } and U1 = S 1 \ { (−1, 0) } , ϕ1 (x, y) = arctan xy ∈ (−π, π) (−1, 0) bc = a unique angle with (x, y) = (cos ϕ1 , sin ϕ1 ) ϕ 1 U2 = S \ { (1, 0) } , ϕ2 (x, y) = arctan xy ∈ (0, 2π) = a unique angle with (x, y) = (cos ϕ2 , sin ϕ2 ) 2 MATH 321: Lecture #33 Then { { U1 , ϕ1 } , { U2 , ϕ2 } } is an atlas for S 1 verification of compatibility: ϕ1 (U1 ∩ U2 ) = (−π, 0) ∪ (0, π) ϕ2 (U1 ∩ U2 ) = (0, π) ∪ (π, 2π) θ , if − π < θ < 0, ϕ2 (cos θ, sin θ) −1 ϕ2 ◦ ϕ1 (θ) = (cos θ, sin θ) ϕ2 , if 0 < θ < π, θ ( θ + 2π, if − π < θ < 0, = U1 ∩ U2 = S 1 \ { (−1, 0), (1, 0) } θ, if 0 < θ < π, bc b bc b bc b bc bc b bc ∈ C∞ Similarly for ϕ1 ◦ ϕ2−1 . Example. (Any n-dimensional surface in Rn+m ) Handwavy definition of an n-dimensional surface in Rn+m • a subset of Rn+m such that locally – m coordinates of points on the surface are determined by the other n coordinates in a C ∞ way. b x=− p 1 − y2 b √ y = − 1 − x2 MATH 321: Real Variables II Lecture #34 University of British Columbia Lecture #34: Instructor: Scribe: April 4, 2008 Dr. Joel Feldman Peter Wong Example. (n-dimensional surface in Rm+n ) Rough definition: A subset of Rn+m is an n-dimensional surface if locally, m coordinates of points on the surface are determined by the n other coordinates (i.e. are functions of the other n coordinates) in a C ∞ way. Example. S 1 = { (x, y) ∈ R2 | x2 + y 2 = 1 } p For y > 0, y = 1 − x2 p For x < 0, x = − 1 − y 2 Formal definition: M ⊂ Rm+n is an n-dimensional surface if, for each ~z ∈ M, there are • a neighborhood U~z of ~z in Rm+n • n natural numbers 1 ≤ j1 < j2 < · · · < jn ≤ m + n (indices of the independent coordinates) Rm M U~z • m C ∞ functions fk (xj1 , . . . , xjn ) with k ∈ { 1, . . . , m + n } \ { 1 ≤ j1 < · · · < jn } such that the point ~x ∈ U~z is in M if and only if xk = fk (xj1 , . . . , xjn ) for all k ∈ { 1, . . . , m + n } \ { 1 ≤ j1 < · · · < jn } b b ~ x ~z b Rn (xj1 , . . . , xjn ) This is a manifold with the atlas A = { { M ∩ U~z , ϕ~z } | ~z ∈ M } where ϕ~z (x1 , . . . , xn+m ) = (xj1 , . . . , xjn ). Equivalent definition: M ⊂ Rm+n is an n-dimensional surface if, for each ~z ∈ M, there are • a neighborhood U~z of ~z in Rm+n • m C ∞ functions Rm U~z gk : U~z → R, k = 1, . . . , m The two definitions are equivalent because of the . . . 0 b b ~ k (~z) | 1 ≤ k ≤ m } are such that { ∇g linearly independent such that the point ~x ∈ U~z is in M if and only if gk (~x) = 0 for all 1 ≤ k ≤ m. M = x) g(~ z ~ x ~ b (xj1 , . . . , xjn ) Rn 2 MATH 321: Lecture #34 ~y ∈ Rm Implicit Function Theorem U Let g(~x) = 0 V (~ x, ~ y) • m, n ∈ N • U ⊂R m+n b b (~ x0 , ~ y0 ) be open • ~g : U → Rm be C ∞ b b W ~x ~x0 obey ~x ∈ Rn • ~g (~x0 , ~y0 ) = 0 for some ~x0 ∈ Rn , ~y0 ∈ Rm with (~x0 , ~y0 ) ∈ U ∂gi • det (~x0 , ~y0 ) 6= 0 1≤i≤m ∂yi 1≤j≤m Then there exist open sets V ⊂ Rm+n and W ⊂ Rn with ~x0 ∈ W and (~x0 , ~y0 ) ∈ V such that for each ~x ∈ W , there is a unique (~x, ~y) ∈ V with ~g(~x, ~y) = ~0. Call ~y ∈ f~(~x). Furthermore, f~ : W → Rm is C ∞ and f~(~x0 ) = ~y0 and ~g (~x, f~(~x)) = ~0 for all ~x ∈ W . Example. (Simple – S 1 ) g(x, y) = x2 + y 2 − 1, (x0 , y0 ) = (0, 1) U = R2 , V = { (x, y) ∈ R2 | y > 0 }, W = (−1, 1), f (x) = Examples. (SO(3), O(3)) p 1 − x2 O(3) = { 3 × 3 real matrices R | R† R = I } SO(3) = { 3 × 3 real matrices R | R† R = I, det R = 1 } Think of the matrix a1 R = a2 a3 b1 b2 b3 c1 c2 c3 as a vector in R9 . Write R† R = I as a linear system of equation.