POINTS IN SETS → interior Å (in the set), exterior (in the complement), boundary ๐ฟA (neither), closure ๐ด (Å + δ๐ด), limit (always surrounded), isolated (not surrounded by some ๐ ball). TYPES OF SETS → open (any point has some ๐ ball within), bounded (contained in some ๐ ball), compact (closed + bounded), convex (segment joining 2 points within), star-shaped (with point with convex prop), poligonally pathconnected (2 points connected by finite path within). ๐: ๐ด ⊆ โ๐ → โ๐ VECTOR-VALUED FUNCTIONS → N=1, M=2/3 gives curve, N=2, M=3 gives surface, N=3, M=3 gives vector field. CONTINUITY → ∀๐ > 0 ∃๐ฟ > 0 s.t. ๐ฅ ∈ ๐ด ∩ ๐ต๐ฟ (๐ฅ0 ) ⇒ ๐(๐ฅ) ∈ ๐ต๐ (๐(๐ฅ0 )), if arithm oper, composition then cont, limit ∀๐ > 0 ∃๐ฟ > 0 s.t. ๐ฅ ∈ ๐ด ∩ ๐ต๐ฟ (๐ฅ0 ) โ {๐ฅ0 } ⇒ ๐(๐ฅ) ∈ ๐ต๐ ((๐)) โโโโ)−๐(๐ โโ) ๐(๐โโ+โ๐ ๐ , gradient (∇๐)(๐โ) โ (๐๐ฅ1 (๐โ), ๐๐ฅ2 (๐โ), … , ๐๐ฅ๐ (๐โ)) ∈ โ๐ , directional โ โโ)−๐(๐โโ) ๐(๐โโ+โ๐ฃ โโ) = ๐(๐โ) + (๐๐โโ ๐)(โ) + deriv for unit vector (๐ท๐ฃโโ ๐)(๐โ) = lim , differentiability (all partial exist) ๐(๐โ + โ โ โ→0 ๐ฟ๐ N, M=1 → partial deriv ๐ฟ๐ฅ (๐โ) = lim โ→0 ๐ โโ|) ๐(|โ ∀๐ฃโ ∈ โ๐ → (๐๐โโ ๐)(๐ฃโ) = (๐ท๐ฃโโ ๐)(๐โ) = ∇๐|๐โโ ⋅ ๐ฃโ = |∇๐)(๐โ)| cosθ Tangent hyperplane equation by: (1) definition of diff for all (๐ฅ, ๐ฆ) close to given diff point (linearization of ๐), (2) property that given vector is orthogonal to the level set ๐(๐ฅ, ๐ฆ, ๐ง) = 0 at a given point (normal vector for tangent plane). โ โโ โ โโ โโ)−(๐๐ |๐(๐โโ+โ )−๐(๐ โโโ ๐)โ| N, M → deriv linear map (๐๐โโ ๐โ)๐ฃโ = lim = 0, Jacobian matrix (๐๐โโ ๐โ)๐ฃโ = โโ |โ | โโ →0 โโ โ δ๐1 δ๐ฅ1 โฏ δ๐1 δ๐ฅ๐ ๐ฃ1 โฎ โฎ |๐ฅโ→๐โโ [ โฎ ] δ๐๐ ๐ฃ๐ โฏ δ๐ฅ๐ ] โฎ δ๐๐ [ δ๐ฅ1 N=1, M → position of particle as a function of time, parametrized curves ๐ฅโ: ๐ผ → โ๐ : simple (one-to-one), regular (deriv ฬ ๐ฅโ (๐ก) defined and cont, never โ0โ (unit tangent vector field ๐ฬ(๐ก) โ ฬ (if diff, deriv is orthogonal to it for every interior ๐ก)). |๐ฅโ (๐ก)| N, M → chain rule functions as multiplication of matrices ๐๐โโ (๐ โ ๐) = (๐๐โโ(๐โโ) ๐โ)(๐๐โโ ๐โ), parametrized curve for ๐ and ๐: โ → โ๐ diff at 0 with derivative βฬ(0) = (๐๐ ๐)(๐ฬ (0)) that takes velocity vec of a curve through a point to the velocity vec of the image curve through the image point. ๐ ๐กโ column of the Jacobian matrix representing ๐๐ ๐ can be interpreted as the velocity at ๐(๐)of the image curve under f of the ๐ ๐กโ “coordinate line” through ๐ (parameterized at constant unit speed in the positive ๐๐ direction). PARAMETRIZED SURFACE → map ฯ: ๐ด ⊆ โ2 → โ๐ (picking ๐ด and deforming it): regular if each component function is diff with cont partial deriv at every point (columns of Jacobian matrix independent). δ๐ฅ δ๐ฆ δ๐ง ๐ฟ๐ฅ ๐ฟ๐ฆ ๐ฟ๐ง Freezing components and writing ฯ๐ข = (δ๐ข , δ๐ข , δ๐ข) = (๐ฯ)๐1 , ๐๐ฃ = (๐ฟ๐ฃ , ๐ฟ๐ฃ , ๐ฟ๐ฃ) = (๐๐)๐2 for ๐ = (๐1 , ๐2 ), vectors ฯ๐ข×ฯ span tangent plane for ฯ. Unit normal ๐(๐ข, ๐ฃ) = |ฯ๐ข×ฯ๐ฃ |. ๐ฃ PARAMETRIZATION OF GRAPHS → ฯ: (๐ข, ๐ฃ) โฆ (๐ข, ๐ฃ, ๐(๐ข, ๐ฃ)): regular, cross-product ฯ๐ข × ฯ๐ฃ = (−๐๐ข , −๐๐ฃ , 1), unit normal ๐(๐ข, ๐ฃ) = (−๐๐ข ,−๐๐ฃ ,1) √1+๐๐ข2 +๐๐ฃ2 = (−๐๐ข ,−๐๐ฃ ,1) √1+|∇๐|2 . HIGHER DERIVATIVES → Parametric curve α regular if ๐α nonzero and α ∈ ๐ถ 1 (๐ผ), parametric surface σ regular if ๐σ has δ๐ δ๐ rank 2 and σ ∈ ๐ถ 1 (๐ด). Clairaut’s/Schwarz’s ๐: ๐ด ⊆ โ๐ → โ partial deriv continuous: δ๐ฅ δ๐ฅ = δ๐ฅ δ๐ฅ (order of deriv not ๐ important). ๐ ๐ ๐ Second deriv given by Hessian: (๐ท 2 ๐)๐ = ๐๐ ∇๐ = [ ๐๐ฅ1 ๐ฅ1 (๐) โฏ ๐๐ฅ1 ๐ฅ๐ (๐) โฎ โฎ โฎ ] (๐) (๐) ๐๐ฅ๐ ๐ฅ1 โฏ ๐๐ฅ๐ ๐ฅ๐ TAYLOR EXPANSION FOR ๐ช๐ 1 ๐(๐ + โ) = ๐(๐) + (๐๐ ๐)โ + โ๐ (๐ท 2 ๐)๐ โ + ๐(|โ|2 ) 2 โ1 1 = ๐(๐) + [๐๐ฅ1 (๐) โฏ ๐๐ฅ๐ (๐)] [ โฎ ] + [โ1 2 โ๐ โฏ ๐๐ฅ1 ๐ฅ1 (๐) โฏ ๐๐ฅ1 ๐ฅ๐ (๐) โ1 โ๐ ] [ โฎ โฎ โฎ ] [ โฎ ] + ๐(|โ|2 ) ๐๐ฅ๐ ๐ฅ1 (๐) โฏ ๐๐ฅ๐ ๐ฅ๐ (๐) โ๐ GLOBAL EXTREMIZERS (N, M=1) → interior by Fermat’s: ๐ not differentiable at ๐ or ∇๐|๐ = (0, … ,0), exterior by Lagrange: ๐(๐ฅ1 , … , ๐ฅ๐ ) = ๐ถ → ๐ and ๐ diff at ๐, (∇๐|๐ ≠ 0, … ,0), ๐ is extremizer of ๐ with constraint ๐ then there exists ∇๐|๐ = λ∇๐|๐ . Finding min and max: (1) interior (2) exterior by Lagrange or parametrizing the boundary (3) compare. DIAGONALIZATION → any ๐ × ๐ matrix can be diagonalized such that ๐ต๐ฃ๐ = λ๐ ๐ฃ๐ , giving ๐ โ [๐ฃ1 λ1 0 0 ๐−1 ๐ต๐ = ๐๐ ๐ต๐ = [ 0 λ2 0 ]. 0 0 λ๐ โฏ ๐ฃ๐ ] and SIGN OF SYMMETRIC MATRIX → positive semidefinite โ๐ ๐ตโ ≥ 0 positive definite โ๐ ๐ตโ > 0, negative semidefinite โ๐ ๐ตโ ≤ 0, negative definite โ๐ ๐ตโ < 0, indefinite (1 strictly positive and 1 strictly negative eigenvalue), degenerate (0 an eigenvalue). LOCAL EXTREMIZER (N, M=1) → necessary: if ๐ is a local minimizer of ๐ข, then (๐ท^2๐ข)|_๐ is positive semidefinite, If ๐ is a local maximizer of ๐ข, then (๐ท^2๐ข)|_๐ is negative semidefinite; sufficient: ๐ Saddle point Local minimum Local maximum (๐ท 2 ๐ข)|๐ Indefinite Positive definite Negative definite det(๐ท 2 ๐ข)|๐ + + ๐ข๐ฅ๐ฅ (๐) + - IMPLICIT FUNCTION THEOREM → ๐น(๐ฅ1 , … , ๐ฅ๐ , ๐ฆ) ∈ ๐ถ 1 : โ๐+1 → โ, ๐ = (๐1 , … , ๐๐ ) ∈ โ๐ , ๐ฆ0 ∈ โ s.t ๐น(๐, ๐ฆ0 ) = 0, ๐น๐ฆ (๐, ๐ฆ0 ) ≠ 0 ⇒ connected open set ๐ ⊆ โ๐ with ๐ ∈ ๐, open interval ๐ผ ⊆ โ with ๐ฆ0 ∈ ๐ผ, unique ๐ถ 1 function ๐: ๐ → โ such that ๐น(๐ฅ, ๐(๐ฅ)) = 0 ∀๐ฅ ∈ ๐; Every (๐ฅ, ๐ฆ) ∈ โ๐+1 satisfying ๐ฅ ∈ ๐, ๐ฆ ∈ ๐ผ and ๐(๐ฅ, ๐ฆ) = 0 also satisfies ๐ฆ = ๐(๐ฅ). δ๐น δ๐ δ๐ฅ๐ | =− | δ๐น (๐ฅ,๐(๐ฅ)) δ๐ฅ๐ ๐ฅ δ๐ฆ ∀๐ฅ ∈ ๐, ๐ ∈ {1, … , ๐} ๐น(๐ฅ, ๐(๐ฅ)) = 0 ๐ฟ ๐ฟ (๐น(๐ฅ, ๐(๐ฅ)) = (0) ๐ฟ๐ฅ๐ ๐ฟ๐ฅ๐ δ๐น δ๐น δ๐ + =0 δ๐ฅ๐ δ๐ฆ δ๐ฅ๐