TheoremBase

Continuity is read as an epsilon-delta bound on norms; a component or entry is bounded by the norm, and the norm by the finite sum of component or entry deviations, which is continuous by the finite-sum lemma. Norms follow from the reverse triangle inequality, sums, multiples and transposes from the componentwise criterion, and products from a bilinear splitting of the difference bounded by a continuous map vanishing at the point.

Proof

Since XX is a set, so is its subclass SS (Subclasses of Sets Are Sets, the Union and Power Set of a Set Exist Uniquely, Binary Unions of Sets Are Sets, and the Universal Class Is Proper (i)); so the component and entry maps of Component Maps and Entry Maps (i) and Component Maps and Entry Maps (ii) are defined for maps on SS. For real numbers, a≤b<ca\le b<c gives a<ca<c (Uniqueness of Least and Greatest Elements, Properties of the Strict Order, and Trichotomy for Total Orders (v), Uniqueness of Least and Greatest Elements, Properties of the Strict Order, and Trichotomy for Total Orders (iv)); such chains are used below without further mention.

Distances. (R,dR)(\mathbb{R},d_{\mathbb{R}}) is a metric space by The Absolute Value Metric on the Real Numbers (i). By The Distance on the Real Numbers Is a Map with Values the Absolute Differences (i), The Euclidean Distance on Rn\mathbb{R}^n (i) and The Distance between Real Matrices (i), dR(a,b)=∣a−b∣d_{\mathbb{R}}(a,b)=|a-b|, dE(y,z)=∥y−z∥d_{E}(y,z)=\lVert y-z\rVert and dp×q(M,N)=∥M−N∥d_{p\times q}(M,N)=\lVert M-N\rVert. Hence, by Continuous Map Between Metric Spaces (i), a map FF from SS into R\mathbb{R}, Rq\mathbb{R}^{q} or Mp×q(R)\mathcal{M}_{p\times q}(\mathbb{R}) is continuous at s0s_{0} if and only if

(C) for every real ε>0\varepsilon>0 there is a real δ>0\delta>0 such that every u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta satisfies ∥F(u)−F(s0)∥<ε\lVert F(u)-F(s_{0})\rVert<\varepsilon,

where ∥⋅∥\lVert\cdot\rVert is read as ∣⋅∣|\cdot| when FF is real-valued. Absolute values and norms are nonnegative, by Rules of Arithmetic and Order in an Ordered Field (x), Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (i) and Rules for the Operator Norm: Bound, Least Constant, Definiteness, Homogeneity, Triangle Inequality and Products (i), and ∥0∥=0\lVert 0\rVert=0, by Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (iii) and Rules for the Operator Norm: Bound, Least Constant, Definiteness, Homogeneity, Triangle Inequality and Products (iv).

Step 1 (a sum of finitely many deviations). Let JJ be a finite set, and let (gj)j∈J(g_{j})_{j\in J} be a family of maps S→RS\to\mathbb{R}, each continuous at s0s_{0} relative to SS. For j∈Jj\in J let aj:S→Ra_{j}:S\to\mathbb{R} be the map u↦∣gj(u)−gj(s0)∣=dR(gj(u),gj(s0))u\mapsto|g_{j}(u)-g_{j}(s_{0})|=d_{\mathbb{R}}(g_{j}(u),g_{j}(s_{0})), and let j↦ajj\mapsto a_{j} be the family of these maps, a map from JJ to RS\mathbb{R}^{S} (Maps Formed Pointwise: Constant Maps, Pairs of Values, Maps Applied to Values, and Families of Maps (iv), applied to the map J×S→RJ\times S\to\mathbb{R}, (j,u)↦dR(gj(u),gj(s0))(j,u)\mapsto d_{\mathbb{R}}(g_{j}(u),g_{j}(s_{0})), of Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (v) and The Distance on the Real Numbers Is a Map with Values the Absolute Differences (i)). Then aj(s0)=0a_{j}(s_{0})=0, and aja_{j} is continuous at s0s_{0}: for ε>0\varepsilon>0, the δ\delta of (C) for gjg_{j} and ε\varepsilon gives ∣aj(u)−aj(s0)∣=aj(u)<ε|a_{j}(u)-a_{j}(s_{0})|=a_{j}(u)<\varepsilon for u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta. By Finite Sums of Real-Valued Maps: Continuity, Integral and Derivative (i), applied to the family (aj)j∈J(a_{j})_{j\in J}, the map h=∑j∈Jaj:S→Rh=\sum_{j\in J}a_{j}:S\to\mathbb{R}, h(u)=∑j∈Jaj(u)h(u)=\sum_{j\in J}a_{j}(u), is continuous at s0s_{0}, and h(s0)=∑j∈J0=0h(s_{0})=\sum_{j\in J}0=0 by Sums over Finite Sets in a Commutative Ring and in an Ordered Field: Distributivity, Products of Sums, Vanishing Terms, Sums over Pairs, Expanding Products of Sums, Counting, Comparison, Monotonicity and the Triangle Inequality (v). Hence, by (C) for hh: for every real ε>0\varepsilon>0 there is a real δ>0\delta>0 such that h(u)≤∣h(u)−h(s0)∣<εh(u)\le|h(u)-h(s_{0})|<\varepsilon for every u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta. The set [q][q] is finite by Finite Sets: the Pigeonhole Principle, Uniqueness of the Length, Subsets, Unions, Products, Images, Bounded Sets of Natural Numbers, Extreme Elements, Sets of Maps, Finite Unions and Finite Choice (vii), and [p]×[q][p]\times[q] is finite by Finite Sets: the Pigeonhole Principle, Uniqueness of the Length, Subsets, Unions, Products, Images, Bounded Sets of Natural Numbers, Extreme Elements, Sets of Maps, Finite Unions and Finite Choice (v); so Step 1 applies to J=[q]J=[q] and to J=[p]×[q]J=[p]\times[q].

Proof of (i). Suppose xx is continuous at s0s_{0}, and let k∈[q]k\in[q] and ε>0\varepsilon>0. Take δ\delta from (C) for xx and ε\varepsilon. For u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta, by Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations (iv) and Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (iv),

∣xk(u)−xk(s0)∣=∣(x(u)−x(s0))k∣≤∥x(u)−x(s0)∥<ε,|x_{k}(u)-x_{k}(s_{0})|=\bigl|\bigl(x(u)-x(s_{0})\bigr)_{k}\bigr|\le\lVert x(u)-x(s_{0})\rVert<\varepsilon ,

so xkx_{k} is continuous at s0s_{0} by (C). Conversely, suppose every xkx_{k} is continuous at s0s_{0}, and let ε>0\varepsilon>0. Take hh and δ\delta from Step 1 for J=[q]J=[q], the family k↦xkk\mapsto x_{k} of Component Maps of a Map into Rq\mathbb{R}^{q}, Entry Maps of a Map into Matrices, and Maps Assembled from Them (i) and ε\varepsilon. For u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta, the components of x(u)−x(s0)x(u)-x(s_{0}) are xk(u)−xk(s0)x_{k}(u)-x_{k}(s_{0}) (Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations (iv)), so by Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (iv) and Natural Numbers as Indices: ℕ ⊆ ℕ₀, 1 ∈ ℕ₀, and the Two Readings of an Operation from 1 to n (v),

∥x(u)−x(s0)∥≤∑k=1q∣xk(u)−xk(s0)∣=∑k∈[q]∣xk(u)−xk(s0)∣=h(u)<ε.\lVert x(u)-x(s_{0})\rVert\le\sum_{k=1}^{q}|x_{k}(u)-x_{k}(s_{0})|=\sum_{k\in[q]}|x_{k}(u)-x_{k}(s_{0})|=h(u)<\varepsilon .

By (C), xx is continuous at s0s_{0}.

Proof of (ii). Suppose AA is continuous at s0s_{0}, let i∈[p]i\in[p], j∈[q]j\in[q] and ε>0\varepsilon>0, and take δ\delta from (C) for AA and ε\varepsilon. For u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta, ∣Aij(u)−Aij(s0)∣≤dp×q(A(u),A(s0))<ε|A_{ij}(u)-A_{ij}(s_{0})|\le d_{p\times q}(A(u),A(s_{0}))<\varepsilon by The Real Matrices with the Operator-Norm Distance Form a Complete Metric Space with Entrywise Convergence (ii); so AijA_{ij} is continuous at s0s_{0}. Conversely, suppose every AijA_{ij} is continuous at s0s_{0}, and let ε>0\varepsilon>0. Take hh and δ\delta from Step 1 for J=[p]×[q]J=[p]\times[q], the family (i,j)↦Aij(i,j)\mapsto A_{ij} of Component Maps of a Map into Rq\mathbb{R}^{q}, Entry Maps of a Map into Matrices, and Maps Assembled from Them (ii) and ε\varepsilon. For u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta, the entries of A(u)−A(s0)A(u)-A(s_{0}) are Akl(u)−Akl(s0)A_{kl}(u)-A_{kl}(s_{0}) (Real Matrices: the Set of Real m x n Matrices with Their Entrywise Operations (iv)), so by The Operator Norm of Entries, Transposes, Multiples of the Identity, Orthogonal Factors and Diagonal Matrices (iv), Natural Numbers as Indices: ℕ ⊆ ℕ₀, 1 ∈ ℕ₀, and the Two Readings of an Operation from 1 to n (v) and Iterated Operations over Finite Sets: Singletons, Disjoint Unions, Reindexing, Products of Sets, Termwise Combination, Homomorphisms and Intervals (iv),

∥A(u)−A(s0)∥≤∑k=1p∑l=1q∣Akl(u)−Akl(s0)∣=∑(k,l)∈[p]×[q]∣Akl(u)−Akl(s0)∣=h(u)<ε.\lVert A(u)-A(s_{0})\rVert\le\sum_{k=1}^{p}\sum_{l=1}^{q}|A_{kl}(u)-A_{kl}(s_{0})|=\sum_{(k,l)\in[p]\times[q]}|A_{kl}(u)-A_{kl}(s_{0})|=h(u)<\varepsilon .

By (C), AA is continuous at s0s_{0}.

Proof of (iii). Let ε>0\varepsilon>0 and take δ\delta from (C) for xx, respectively AA, and ε\varepsilon. For u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta, ∣∥x(u)∥−∥x(s0)∥∣≤∥x(u)−x(s0)∥<ε\bigl|\lVert x(u)\rVert-\lVert x(s_{0})\rVert\bigr|\le\lVert x(u)-x(s_{0})\rVert<\varepsilon by Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (vii), and ∣∥A(u)∥−∥A(s0)∥∣≤dp×q(A(u),A(s0))<ε\bigl|\lVert A(u)\rVert-\lVert A(s_{0})\rVert\bigr|\le d_{p\times q}(A(u),A(s_{0}))<\varepsilon by The Real Matrices with the Operator-Norm Distance Form a Complete Metric Space with Entrywise Convergence (iii). By (C), both norm maps are continuous at s0s_{0}.

Proof of (iv). Constants. If FF is constant, then ∥F(u)−F(s0)∥=∥0∥=0<ε\lVert F(u)-F(s_{0})\rVert=\lVert0\rVert=0<\varepsilon for every u∈Su\in S and every ε>0\varepsilon>0, so (C) holds with any δ>0\delta>0.

For each of the other conclusions we assume only that the maps it involves are continuous at s0s_{0}.

Sums, scalar multiples and transposes. For u∈Su\in S, k∈[q]k\in[q], i∈[p]i\in[p] and j∈[q]j\in[q], by Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations (iv), Real Matrices: the Set of Real m x n Matrices with Their Entrywise Operations (iv) and Transpose of a Real Matrix (i),

(x(u)+y(u))k=xk(u)+yk(u),(φ(u)x(u))k=φ(u) xk(u),(A(u)+B(u))ij=Aij(u)+Bij(u),\bigl(x(u)+y(u)\bigr)_{k}=x_{k}(u)+y_{k}(u),\quad\bigl(\varphi(u)x(u)\bigr)_{k}=\varphi(u)\,x_{k}(u),\quad\bigl(A(u)+B(u)\bigr)_{ij}=A_{ij}(u)+B_{ij}(u), (φ(u)A(u))ij=φ(u) Aij(u),(A(u)⊤)ji=Aij(u).\bigl(\varphi(u)A(u)\bigr)_{ij}=\varphi(u)\,A_{ij}(u),\qquad\bigl(A(u)^{\top}\bigr)_{ji}=A_{ij}(u).

Whichever of xx, yy, AA and BB are continuous at s0s_{0} have component, respectively entry, maps continuous at s0s_{0}, by (i) and (ii). For each k∈[q]k\in[q] the component map (x+y)k(x+y)_{k} of u↦x(u)+y(u)u\mapsto x(u)+y(u) is the map u↦xk(u)+yk(u)u\mapsto x_{k}(u)+y_{k}(u), as both maps have the same values (by the uniqueness in Component Maps and Entry Maps (i)), and likewise for the other component and entry maps displayed above. So the component maps of u↦x(u)+y(u)u\mapsto x(u)+y(u) and u↦φ(u)x(u)u\mapsto\varphi(u)x(u) and the entry maps of u↦A(u)+B(u)u\mapsto A(u)+B(u) and u↦φ(u)A(u)u\mapsto\varphi(u)A(u) are continuous at s0s_{0} by Continuity of Sums and Products of Real-Valued Functions on a Metric Space (ii) and Continuity of Sums and Products of Real-Valued Functions on a Metric Space (iii), and the entry map of u↦A(u)⊤u\mapsto A(u)^{\top} at (j,i)(j,i) is AijA_{ij}. By (i) and (ii) these five maps are continuous at s0s_{0}.

Products. Let (F,G,P)(F,G,P) be one of the triples (x,y,u↦x(u)⋅y(u))(x,y,u\mapsto x(u)\cdot y(u)), (A,C,u↦A(u)C(u))(A,C,u\mapsto A(u)C(u)) and (A,x,u↦A(u)x(u))(A,x,u\mapsto A(u)x(u)), write F(u)G(u)F(u)G(u) for P(u)P(u), put F0=F(s0)F_{0}=F(s_{0}) and G0=G(s0)G_{0}=G(s_{0}), and let a,b:S→Ra,b:S\to\mathbb{R} be the maps u↦∥F(u)−F0∥u\mapsto\lVert F(u)-F_{0}\rVert and u↦∥G(u)−G0∥u\mapsto\lVert G(u)-G_{0}\rVert, the distances of F(u)F(u) from F0F_{0} and of G(u)G(u) from G0G_{0} (Maps Formed Pointwise: Constant Maps, Pairs of Values, Maps Applied to Values, and Families of Maps (i) and Maps Formed Pointwise: Constant Maps, Pairs of Values, Maps Applied to Values, and Families of Maps (iii), with the distance maps of The Summation Map on Real-Valued Maps over a Finite Set, and the Dot Product, Euclidean Norm and Euclidean Distance as Maps (iv) and The Operator-Norm Distance between Real Matrices Is a Map (i)). For u∈Su\in S,

P(u)−P(s0)=(F(u)−F0)G(u)+F0(G(u)−G0),P(u)-P(s_{0})=\bigl(F(u)-F_{0}\bigr)G(u)+F_{0}\bigl(G(u)-G_{0}\bigr),

by The Dot Product on Rn\mathbb{R}^n Is Symmetric, Bilinear and Positive Definite (ii) and The Dot Product on Rn\mathbb{R}^n Is Symmetric, Bilinear and Positive Definite (iii) for the dot product, by Algebraic Rules for Matrix Products, Matrix-Vector Products and Transposes (ii) and Algebraic Rules for Matrix Products, Matrix-Vector Products and Transposes (iii) for the matrix product and by Algebraic Rules for Matrix Products, Matrix-Vector Products and Transposes (v) for the matrix-vector product, a difference M−NM-N of matrices being M+(−1)NM+(-1)N by Real Matrices: the Set of Real m x n Matrices with Their Entrywise Operations (iv) and Real Matrices: the Set of Real m x n Matrices with Their Entrywise Operations (iii). The triangle inequality (Rules of Arithmetic and Order in an Ordered Field (xi) for the real absolute value, Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (vi), Rules for the Operator Norm: Bound, Least Constant, Definiteness, Homogeneity, Triangle Inequality and Products (vi)) and the product bounds ∣y⋅z∣≤∥y∥∥z∥|y\cdot z|\le\lVert y\rVert\lVert z\rVert, ∥MN∥≤∥M∥∥N∥\lVert MN\rVert\le\lVert M\rVert\lVert N\rVert and ∥Mz∥≤∥M∥∥z∥\lVert Mz\rVert\le\lVert M\rVert\lVert z\rVert (The Cauchy-Schwarz Inequality on Rn\mathbb{R}^n (i), Rules for the Operator Norm: Bound, Least Constant, Definiteness, Homogeneity, Triangle Inequality and Products (vii), Rules for the Operator Norm: Bound, Least Constant, Definiteness, Homogeneity, Triangle Inequality and Products (ii)) give ∥P(u)−P(s0)∥≤a(u)∥G(u)∥+∥F0∥ b(u)\lVert P(u)-P(s_{0})\rVert\le a(u)\lVert G(u)\rVert+\lVert F_{0}\rVert\,b(u), and ∥G(u)∥≤b(u)+∥G0∥\lVert G(u)\rVert\le b(u)+\lVert G_{0}\rVert by the triangle inequality, as G(u)=(G(u)−G0)+G0G(u)=(G(u)-G_{0})+G_{0}. As a(u)≥0a(u)\ge0,

∥P(u)−P(s0)∥≤h(u),h(u)=a(u)(b(u)+∥G0∥)+∥F0∥ b(u)(u∈S).\lVert P(u)-P(s_{0})\rVert\le h(u),\qquad h(u)=a(u)\bigl(b(u)+\lVert G_{0}\rVert\bigr)+\lVert F_{0}\rVert\,b(u)\qquad(u\in S).

The map aa is continuous at s0s_{0} with a(s0)=∥0∥=0a(s_{0})=\lVert0\rVert=0: for ε>0\varepsilon>0, the δ\delta of (C) for FF and ε\varepsilon gives ∣a(u)−a(s0)∣=a(u)<ε|a(u)-a(s_{0})|=a(u)<\varepsilon for u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta. Likewise bb is continuous at s0s_{0} with b(s0)=0b(s_{0})=0. The map h=a(b+∥G0∥)+∥F0∥ bh=a\bigl(b+\lVert G_{0}\rVert\bigr)+\lVert F_{0}\rVert\,b is formed with the pointwise operations and constant maps of Pointwise Operations on Real-Valued Maps (i) and Pointwise Operations on Real-Valued Maps (ii); by Continuity of Sums and Products of Real-Valued Functions on a Metric Space (i), Continuity of Sums and Products of Real-Valued Functions on a Metric Space (ii), Continuity of Sums and Products of Real-Valued Functions on a Metric Space (iii) and Continuity of Sums and Products of Real-Valued Functions on a Metric Space (iv), hh is continuous at s0s_{0}, and h(s0)=0h(s_{0})=0. Let ε>0\varepsilon>0 and take δ>0\delta>0 such that ∣h(u)−h(s0)∣<ε|h(u)-h(s_{0})|<\varepsilon for u∈Su\in S with d(s0,u)<δd(s_{0},u)<\delta; for these uu, ∥P(u)−P(s0)∥≤h(u)=∣h(u)−h(s0)∣<ε\lVert P(u)-P(s_{0})\rVert\le h(u)=|h(u)-h(s_{0})|<\varepsilon, as h(u)≥0h(u)\ge0. By (C), PP is continuous at s0s_{0}.

Proof of (v). By Continuous Map Between Metric Spaces (ii), a map is continuous on SS if and only if it is continuous at every point of SS relative to SS. Applying (i), (ii), (iii) and (iv) with s0=ss_{0}=s for every s∈Ss\in S gives the statements on SS.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…