TheoremBase

Proof sketch: by induction on k, using the first-order derivative rules to express the partial derivatives of sums, products and reciprocals through functions of class Cᵏ.

Proof

Throughout, the arithmetic and order of R\mathbb{R} are used through The Structure of ℝ as a Field and an Ordered Ring (i), The Structure of ℝ as a Field and an Ordered Ring (ii), Arithmetic and Order Rules in the Real Numbers, Stated for ℝ (v), Arithmetic and Order Rules in the Real Numbers, Stated for ℝ (ix), Rules of Arithmetic and Order in an Ordered Field (iii) and The Quotient 1/x, Order Rules for Sums, and Nesting of Closed Intervals in ℝ (i), and the metric spaces (Rn,dE)(\mathbb{R}^{n},d_{E}) and (R,dR)(\mathbb{R},d_{\mathbb{R}}) through Rn\mathbb{R}^n with the Euclidean Distance Is a Complete Metric Space with Coordinatewise Convergence (i) and The Absolute Value Metric on the Real Numbers (i); membership in C0(U)\mathcal{C}^{0}(U) and Ck+1(U)\mathcal{C}^{k+1}(U) is read through Functions of Class Cᵏ on an Open Subset of ℝⁿ (i), and each induction on kk runs from 00 over N0\mathbb{N}_{0} (The Natural Numbers and the Natural Numbers with Zero: Arithmetic, Order, Induction and Recursion (viii)), k+1k+1 being the successor of kk. Two reductions are used for every clause. A vector-valued statement follows from the real-valued one applied to the component maps (Functions of Class Cᵏ on an Open Subset of ℝⁿ (iii)), since the components of f+gf+g and λf\lambda f are fj+gjf_{j}+g_{j} and λfj\lambda f_{j}, and those of a constant function are constant. A statement about smooth functions follows from the CkC^{k} statement for every kk (Smooth Functions on an Open Subset of ℝⁿ (i), Smooth Functions on an Open Subset of ℝⁿ (ii)).

We first compute the first partial derivatives of sums, scalar multiples, products and reciprocals. Let f,g∈C1(U)f,g\in\mathcal{C}^{1}(U), λ∈R\lambda\in\mathbb{R} and i∈[n]i\in[n]. By Basic Properties of Cᵏ Functions (v), ff and gg are differentiable at every x∈Ux\in U, with derivative vectors ∇f(x)\nabla f(x) and ∇g(x)\nabla g(x). By Rules for Derivatives of Functions on Open Subsets of ℝⁿ (v), Rules for Derivatives of Functions on Open Subsets of ℝⁿ (vi) and, when g(x)≠0g(x)\neq0 for every x∈Ux\in U, Rules for Derivatives of Functions on Open Subsets of ℝⁿ (vii), the functions f+gf+g, λf\lambda f, fgfg and 1/g1/g are differentiable at every x∈Ux\in U, with derivative vectors ∇f(x)+∇g(x)\nabla f(x)+\nabla g(x), λ∇f(x)\lambda\nabla f(x), f(x)∇g(x)+g(x)∇f(x)f(x)\nabla g(x)+g(x)\nabla f(x) and −1g(x)2∇g(x)-\frac{1}{g(x)^{2}}\nabla g(x). The ii-th component of the derivative vector of a differentiable function is its ii-th partial derivative (The Derivative Is the Jacobian Matrix, and Is Unique (i)); reading off ii-th components, the partial derivatives below exist at every point of UU, and, as functions on UU with the pointwise operations,

∂i(f+g)=∂if+∂ig,∂i(λf)=λ ∂if,∂i(fg)=f ∂ig+g ∂if,\partial_{i}(f+g)=\partial_{i}f+\partial_{i}g,\qquad\partial_{i}(\lambda f)=\lambda\,\partial_{i}f,\qquad\partial_{i}(fg)=f\,\partial_{i}g+g\,\partial_{i}f, ∂i(1g)=(−1)(1g⋅1g)∂igif g(x)≠0 for every x∈U.\partial_{i}\Bigl(\frac{1}{g}\Bigr)=(-1)\Bigl(\frac{1}{g}\cdot\frac{1}{g}\Bigr)\partial_{i}g\qquad\text{if }g(x)\neq0\text{ for every }x\in U.

We begin with (i), showing by induction on kk that every constant function U→RU\to\mathbb{R} lies in Ck(U)\mathcal{C}^{k}(U). For k=0k=0, constant functions are continuous on UU by Continuity of Sums and Products of Real-Valued Functions on a Metric Space (i). For the inductive step, let φ\varphi be constant. By Rules for Derivatives of Functions on Open Subsets of ℝⁿ (i), φ\varphi is differentiable at every x∈Ux\in U with ∇φ(x)=0\nabla\varphi(x)=0, so ∂iφ(x)=0\partial_{i}\varphi(x)=0 for every x∈Ux\in U and i∈[n]i\in[n] (The Derivative Is the Jacobian Matrix, and Is Unique (i)). Thus each ∂iφ\partial_{i}\varphi is the constant function 00, which lies in Ck(U)\mathcal{C}^{k}(U) by the induction hypothesis, and φ∈Ck+1(U)\varphi\in\mathcal{C}^{k+1}(U).

For (ii), let φ(x)=v⋅x+c\varphi(x)=v\cdot x+c. By Rules for Derivatives of Functions on Open Subsets of ℝⁿ (ii), φ\varphi is differentiable at every point of UU with ∇φ(x)=v\nabla\varphi(x)=v; it is therefore continuous on UU by Differentiable Functions Are Continuous (i), so φ∈C0(U)\varphi\in\mathcal{C}^{0}(U). Its partial derivatives ∂iφ\partial_{i}\varphi are the constant functions with value viv_{i} (The Derivative Is the Jacobian Matrix, and Is Unique (i)), which lie in Ck(U)\mathcal{C}^{k}(U) for every kk by (i); hence φ∈Ck+1(U)\varphi\in\mathcal{C}^{k+1}(U) for every kk, and φ\varphi is smooth. For f(x)=Bx+cf(x)=Bx+c, let bj=(Bj1,…,Bjn)∈Rnb_{j}=(B_{j1},\dots,B_{jn})\in\mathbb{R}^{n} be the jj-th row of BB. By Matrix-Vector Product (i),

fj(x)=(Bx)j+cj=∑l=1nBjlxl+cj=bj⋅x+cj,f_{j}(x)=(Bx)_{j}+c_{j}=\sum_{l=1}^{n}B_{jl}x_{l}+c_{j}=b_{j}\cdot x+c_{j},

the last equality by The Dot Product on Rn\mathbb{R}^n (i); so every component of ff is smooth by the real-valued case, and ff is smooth.

For (iii), πi(x)=xi=ei⋅x+0\pi_{i}(x)=x_{i}=e_{i}\cdot x+0, since ei⋅x=xie_{i}\cdot x=x_{i} by Standard Basis Vectors of ℝ^q: Coordinates, Orthonormality and Unit Length (i), so πi\pi_{i} is smooth by (ii).

For (iv), we show by induction on kk that f,g∈Ck(U)f,g\in\mathcal{C}^{k}(U) and λ∈R\lambda\in\mathbb{R} imply f+g,λf∈Ck(U)f+g,\lambda f\in\mathcal{C}^{k}(U). For k=0k=0 this is the continuity of sums and scalar multiples of continuous functions, Continuity of Sums and Products of Real-Valued Functions on a Metric Space (v). For the inductive step, let f,g∈Ck+1(U)f,g\in\mathcal{C}^{k+1}(U) and i∈[n]i\in[n]. By Basic Properties of Cᵏ Functions (ii), f,g∈C1(U)f,g\in\mathcal{C}^{1}(U), so the first paragraph gives ∂i(f+g)=∂if+∂ig\partial_{i}(f+g)=\partial_{i}f+\partial_{i}g and ∂i(λf)=λ ∂if\partial_{i}(\lambda f)=\lambda\,\partial_{i}f. By Basic Properties of Cᵏ Functions (iii), ∂if,∂ig∈Ck(U)\partial_{i}f,\partial_{i}g\in\mathcal{C}^{k}(U), and the induction hypothesis gives ∂i(f+g),∂i(λf)∈Ck(U)\partial_{i}(f+g),\partial_{i}(\lambda f)\in\mathcal{C}^{k}(U). As ii was arbitrary, f+g,λf∈Ck+1(U)f+g,\lambda f\in\mathcal{C}^{k+1}(U).

For (v), we show by induction on kk that f,g∈Ck(U)f,g\in\mathcal{C}^{k}(U) imply fg∈Ck(U)fg\in\mathcal{C}^{k}(U). For k=0k=0 this is the continuity of products of continuous functions, Continuity of Sums and Products of Real-Valued Functions on a Metric Space (v). For the inductive step, let f,g∈Ck+1(U)f,g\in\mathcal{C}^{k+1}(U) and i∈[n]i\in[n]. Since f,g∈C1(U)f,g\in\mathcal{C}^{1}(U) by Basic Properties of Cᵏ Functions (ii), the first paragraph gives

∂i(fg)=f ∂ig+g ∂if,\partial_{i}(fg)=f\,\partial_{i}g+g\,\partial_{i}f,

where f,g∈Ck(U)f,g\in\mathcal{C}^{k}(U) by Basic Properties of Cᵏ Functions (ii) and ∂if,∂ig∈Ck(U)\partial_{i}f,\partial_{i}g\in\mathcal{C}^{k}(U) by Basic Properties of Cᵏ Functions (iii). By the induction hypothesis both products on the right lie in Ck(U)\mathcal{C}^{k}(U), and so does their sum, by (iv) at level kk. Hence fg∈Ck+1(U)fg\in\mathcal{C}^{k+1}(U).

For (vi), we show by induction on kk that every g∈Ck(U)g\in\mathcal{C}^{k}(U) with g(x)≠0g(x)\neq0 for every x∈Ux\in U has 1/g∈Ck(U)1/g\in\mathcal{C}^{k}(U). For k=0k=0, let E={t∈R:t≠0}E=\{t\in\mathbb{R}:t\neq0\}. The identity map of EE is continuous on EE by Continuity of Identity Maps, Compositions and Reciprocals (i) and has no zeros, so the function r:E→Rr:E\to\mathbb{R}, t↦1/tt\mapsto1/t, is continuous on EE by Continuity of Identity Maps, Compositions and Reciprocals (iii). Since gg takes its values in EE, 1/g=r∘g1/g=r\circ g is continuous on UU by Continuity of Identity Maps, Compositions and Reciprocals (ii). For the inductive step, let g∈Ck+1(U)g\in\mathcal{C}^{k+1}(U) have no zeros, and let i∈[n]i\in[n]. Since g∈C1(U)g\in\mathcal{C}^{1}(U) by Basic Properties of Cᵏ Functions (ii), the first paragraph gives

∂i(1g)=(−1)(1g⋅1g)∂ig.\partial_{i}\Bigl(\frac{1}{g}\Bigr)=(-1)\Bigl(\frac{1}{g}\cdot\frac{1}{g}\Bigr)\partial_{i}g .

Here g∈Ck(U)g\in\mathcal{C}^{k}(U) by Basic Properties of Cᵏ Functions (ii), so 1/g∈Ck(U)1/g\in\mathcal{C}^{k}(U) by the induction hypothesis, and ∂ig∈Ck(U)\partial_{i}g\in\mathcal{C}^{k}(U) by Basic Properties of Cᵏ Functions (iii). Two applications of (v) at level kk and one of (iv) with λ=−1\lambda=-1 give ∂i(1/g)∈Ck(U)\partial_{i}(1/g)\in\mathcal{C}^{k}(U), so 1/g∈Ck+1(U)1/g\in\mathcal{C}^{k+1}(U).

Finally, for (vii), f/g=f⋅(1/g)f/g=f\cdot(1/g) (Functions of Class Cᵏ: Notation (v)), so f/gf/g is of class CkC^{k}, respectively smooth, by (vi) and (v).

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…