TheoremBase

The reflection ψ(u) = a+b-u of [a,b] is its own inverse and has derivative -1, so the chain rule moves solutions between the terminal and the reversed initial value problem; the reversed right-hand side inherits continuity and the Lipschitz constant, and Picard–Lindelöf applies to it.

Proof

Computations with real numbers are those of The Real Numbers as a Type (vi). By Basic Facts about Intervals of the Real Line (v), [a,b][a,b] is an interval containing the two distinct points aa and bb; it serves as both II and JJ of Rules for Derivatives of Vector- and Matrix-Valued Maps (xi) below.

The reflection. Let ψ:[a,b]→[a,b]\psi:[a,b]\to[a,b] be the function with ψ(u)=a+b−u\psi(u)=a+b-u for every u∈[a,b]u\in[a,b] of The Reflection of a Closed Interval (i). For all u,v∈[a,b]u,v\in[a,b], ψ(ψ(u))=u\psi(\psi(u))=u by The Reflection of a Closed Interval (ii) and ∣ψ(u)−ψ(v)∣=∣u−v∣|\psi(u)-\psi(v)|=|u-v| by The Reflection of a Closed Interval (iii); and ψ\psi, read as a function into R\mathbb{R}, is differentiable, derivatives being taken relative to [a,b][a,b], with ψ′(u)=−1\psi'(u)=-1 for every u∈[a,b]u\in[a,b], by The Reflection of a Closed Interval (v). Moreover ψ(a)=a+b−a=b\psi(a)=a+b-a=b and ψ(b)=a+b−b=a\psi(b)=a+b-b=a.

Step 1 (reflected functions). Let w,w~:[a,b]→Rnw,\tilde{w}:[a,b]\to\mathbb{R}^{n} be functions with w~(u)=w(ψ(u))\tilde{w}(u)=w(\psi(u)) for every u∈[a,b]u\in[a,b], and let ww be differentiable. Then w~=w∘ψ\tilde{w}=w\circ\psi, the two functions having the same values (Maps Formed Pointwise: Constant Maps, Pairs of Values, Maps Applied to Values, and Families of Maps (iii)). Let u∈[a,b]u\in[a,b]. As ww is differentiable at ψ(u)∈[a,b]\psi(u)\in[a,b], Rules for Derivatives of Vector- and Matrix-Valued Maps (xi), with [a,b][a,b] for II and JJ, ψ\psi for ψ\psi and uu for s0s_{0}, shows that w~\tilde{w} is differentiable at uu with

w~′(u)=ψ′(u) w′(ψ(u))=(−1) w′(ψ(u))=−w′(ψ(u)),\tilde{w}'(u)=\psi'(u)\,w'\bigl(\psi(u)\bigr)=(-1)\,w'\bigl(\psi(u)\bigr)=-w'\bigl(\psi(u)\bigr),

the last step by Rules for Pointwise Operations on Real-Valued Maps: a Commutative Ring, Scalar Multiples and Evaluation (iv), points of Rn\mathbb{R}^{n} being maps [n]→R[n]\to\mathbb{R} with the operations of Euclidean Space Rn\mathbb{R}^n: n-Tuples of Real Numbers with Their Componentwise Operations (iv). So w~\tilde{w} is differentiable. In the same way, by the matrix half of Rules for Derivatives of Vector- and Matrix-Valued Maps (xi) and by Rules for Pointwise Operations on Real-Valued Maps: a Commutative Ring, Scalar Multiples and Evaluation (iv) with the operations of Real Matrices: the Set of Real m x n Matrices with Their Entrywise Operations (iv): if W,W~:[a,b]→Mp×q(R)W,\tilde{W}:[a,b]\to\mathcal{M}_{p\times q}(\mathbb{R}) are functions with W~(u)=W(ψ(u))\tilde{W}(u)=W(\psi(u)) for every u∈[a,b]u\in[a,b] and WW is differentiable, then W~\tilde{W} is differentiable with W~′(u)=−W′(ψ(u))\tilde{W}'(u)=-W'(\psi(u)) for every u∈[a,b]u\in[a,b].

Proof of (i). First, we describe GG and yy. By Projection, Evaluation and Currying Maps (i), there are functions π1:[a,b]×Rn→[a,b]\pi_{1}:[a,b]\times\mathbb{R}^{n}\to[a,b] and π2:[a,b]×Rn→Rn\pi_{2}:[a,b]\times\mathbb{R}^{n}\to\mathbb{R}^{n} with π1(u,z)=u\pi_{1}(u,z)=u and π2(u,z)=z\pi_{2}(u,z)=z for all u∈[a,b]u\in[a,b] and z∈Rnz\in\mathbb{R}^{n}. By Maps Formed Pointwise: Constant Maps, Pairs of Values, Maps Applied to Values, and Families of Maps (iii), applied first to π1\pi_{1} and ψ\psi and then to ψ∘π1\psi\circ\pi_{1}, π2\pi_{2} and FF, there is a function K:[a,b]×Rn→RnK:[a,b]\times\mathbb{R}^{n}\to\mathbb{R}^{n} with K(u,z)=F(ψ(π1(u,z)),π2(u,z))=F(a+b−u,z)K(u,z)=F(\psi(\pi_{1}(u,z)),\pi_{2}(u,z))=F(a+b-u,z) for all u∈[a,b]u\in[a,b] and z∈Rnz\in\mathbb{R}^{n}. By Maps Formed Pointwise from Vector and Matrix Operations (i), with [a,b]×Rn[a,b]\times\mathbb{R}^{n} for SS, nn for qq and KK for xx, there is exactly one function [a,b]×Rn→Rn[a,b]\times\mathbb{R}^{n}\to\mathbb{R}^{n} whose value at every point (u,z)(u,z) is −K(u,z)=−F(a+b−u,z)-K(u,z)=-F(a+b-u,z); this function is GG. For a function x:[a,b]→Rnx:[a,b]\to\mathbb{R}^{n}, the composite x∘ψ:[a,b]→Rnx\circ\psi:[a,b]\to\mathbb{R}^{n} is, by Maps Formed Pointwise: Constant Maps, Pairs of Values, Maps Applied to Values, and Families of Maps (iii), the only function whose value at every u∈[a,b]u\in[a,b] is x(ψ(u))=x(a+b−u)x(\psi(u))=x(a+b-u); this function is yy.

Now let xx and yy be as in (i). For t∈[a,b]t\in[a,b], y(ψ(t))=x(ψ(ψ(t)))=x(t)y(\psi(t))=x(\psi(\psi(t)))=x(t).

Suppose that xx is differentiable with x′(t)=F(t,x(t))x'(t)=F(t,x(t)) for every t∈[a,b]t\in[a,b] and x(b)=ξx(b)=\xi. By Step 1, with xx for ww and yy for w~\tilde{w}, yy is differentiable, and for u∈[a,b]u\in[a,b]

y′(u)=−x′(ψ(u))=−F(ψ(u),x(ψ(u)))=−F(a+b−u,y(u))=G(u,y(u));y'(u)=-x'\bigl(\psi(u)\bigr)=-F\bigl(\psi(u),x(\psi(u))\bigr)=-F\bigl(a+b-u,y(u)\bigr)=G\bigl(u,y(u)\bigr);

and y(a)=x(ψ(a))=x(b)=ξy(a)=x(\psi(a))=x(b)=\xi.

Conversely, suppose that yy is differentiable with y′(u)=G(u,y(u))y'(u)=G(u,y(u)) for every u∈[a,b]u\in[a,b] and y(a)=ξy(a)=\xi. By Step 1, with yy for ww and xx for w~\tilde{w} (as x(t)=y(ψ(t))x(t)=y(\psi(t))), xx is differentiable, and for t∈[a,b]t\in[a,b]

x′(t)=−y′(ψ(t))=−G(ψ(t),y(ψ(t)))=−(−F(ψ(ψ(t)),x(t)))=F(t,x(t)),x'(t)=-y'\bigl(\psi(t)\bigr)=-G\bigl(\psi(t),y(\psi(t))\bigr)=-\Bigl(-F\bigl(\psi(\psi(t)),x(t)\bigr)\Bigr)=F\bigl(t,x(t)\bigr),

using −(−v)=v-(-v)=v (Additive Laws for Vectors in Rq\mathbb{R}^q and for Real p×qp\times q Matrices: Group Laws and Rules for Differences (ii)); and x(b)=y(ψ(b))=y(a)=ξx(b)=y(\psi(b))=y(a)=\xi.

Finally, by Solution of an Initial Value Problem on [a,c][a,c] (i) and Solution of an Initial Value Problem on [a,c][a,c] (iii), with bb for cc, a solution of the initial value problem of The Initial Value Problem for a First-Order System of Ordinary Differential Equations (i) with right-hand side GG and initial value ξ\xi is exactly a function [a,b]→Rn[a,b]\to\mathbb{R}^{n} that is differentiable relative to [a,b][a,b] with y′(u)=G(u,y(u))y'(u)=G(u,y(u)) for every u∈[a,b]u\in[a,b] and y(a)=ξy(a)=\xi.

Proof of (ii). The proof of (i) applies word for word with Mp×q(R)\mathcal{M}_{p\times q}(\mathbb{R}), FF, GG, Ξ\Xi, XX, YY and ZZ in place of Rn\mathbb{R}^{n}, FF, GG, ξ\xi, xx, yy and zz, with Maps Formed Pointwise from Vector and Matrix Operations (i) applied to KK in the role of ff (with pp and qq as here) for the existence of GG, with the matrix half of Step 1, with −(−V)=V-(-V)=V for matrices (Additive Laws for Vectors in Rq\mathbb{R}^q and for Real p×qp\times q Matrices: Group Laws and Rules for Differences (ii)), and with Solution of an Initial Value Problem on [a,c][a,c] (ii) and The Initial Value Problem for a First-Order System of Ordinary Differential Equations (ii) in place of Solution of an Initial Value Problem on [a,c][a,c] (i) and The Initial Value Problem for a First-Order System of Ordinary Differential Equations (i).

Proof of (iii). We treat the vector case; the matrix case is the same with Mp×q(R)\mathcal{M}_{p\times q}(\mathbb{R}), dp×qd_{p\times q}, (ii) and The Picard–Lindelöf Theorem on a Closed Interval: Existence, Uniqueness and Continuous Dependence for Globally Lipschitz Systems (iii) in place of Rn\mathbb{R}^{n}, dEd_{E}, (i) and The Picard–Lindelöf Theorem on a Closed Interval: Existence, Uniqueness and Continuous Dependence for Globally Lipschitz Systems (i), dp×qd_{p\times q} being a metric by The Real Matrices with the Operator-Norm Distance Form a Complete Metric Space with Entrywise Convergence (i) with dp×q(V,W)=∥V−W∥d_{p\times q}(V,W)=\lVert V-W\rVert by The Distance between Real Matrices (i), and with ∥−V∥=∥(−1)V∥=∣−1∣ ∥V∥=∥V∥\lVert-V\rVert=\lVert(-1)V\rVert=|-1|\,\lVert V\rVert=\lVert V\rVert for V∈Mp×q(R)V\in\mathcal{M}_{p\times q}(\mathbb{R}), by Rules for Pointwise Operations on Real-Valued Maps: a Commutative Ring, Scalar Multiples and Evaluation (iv) with the operations of Real Matrices: the Set of Real m x n Matrices with Their Entrywise Operations (iv) and by Rules for the Operator Norm: Bound, Least Constant, Definiteness, Homogeneity, Triangle Inequality and Products (v), in place of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (v).

Let GG be the function of (i). For all v,w∈Rnv,w\in\mathbb{R}^{n},

(−v)−(−w)=(−v)+w=w+(−v)=w−v=−(v−w)(-v)-(-w)=(-v)+w=w+(-v)=w-v=-(v-w)

by Additive Laws for Vectors in Rq\mathbb{R}^q and for Real p×qp\times q Matrices: Group Laws and Rules for Differences (ii) and Additive Laws for Vectors in Rq\mathbb{R}^q and for Real p×qp\times q Matrices: Group Laws and Rules for Differences (i), and ∥−v∥=∥v∥\lVert-v\rVert=\lVert v\rVert by Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (v). Hence, for u,u′∈[a,b]u,u'\in[a,b] and v,w∈Rnv,w\in\mathbb{R}^{n}, applying these to the points F(ψ(u′),w)F(\psi(u'),w) and F(ψ(u),v)F(\psi(u),v) and using G(u′,w)=−F(ψ(u′),w)G(u',w)=-F(\psi(u'),w) and G(u,v)=−F(ψ(u),v)G(u,v)=-F(\psi(u),v),

G(u′,w)−G(u,v)=−(F(ψ(u′),w)−F(ψ(u),v)),∥G(u′,w)−G(u,v)∥=∥F(ψ(u′),w)−F(ψ(u),v)∥.(∗)G(u',w)-G(u,v)=-\Bigl(F\bigl(\psi(u'),w\bigr)-F\bigl(\psi(u),v\bigr)\Bigr),\qquad \lVert G(u',w)-G(u,v)\rVert=\bigl\lVert F\bigl(\psi(u'),w\bigr)-F\bigl(\psi(u),v\bigr)\bigr\rVert.\tag{$\ast$}

GG is Lipschitz. For u∈[a,b]u\in[a,b] and v,w∈Rnv,w\in\mathbb{R}^{n}, by (∗)(\ast) with u′=uu'=u and by the hypothesis on FF at ψ(u)∈[a,b]\psi(u)\in[a,b], ∥G(u,w)−G(u,v)∥=∥F(ψ(u),w)−F(ψ(u),v)∥≤L ∥w−v∥\lVert G(u,w)-G(u,v)\rVert=\lVert F(\psi(u),w)-F(\psi(u),v)\rVert\le L\,\lVert w-v\rVert; exchanging vv and ww, ∥G(u,v)−G(u,w)∥≤L ∥v−w∥\lVert G(u,v)-G(u,w)\rVert\le L\,\lVert v-w\rVert.

GG is continuous. Let (u,v)∈[a,b]×Rn(u,v)\in[a,b]\times\mathbb{R}^{n} and let ε∈R\varepsilon\in\mathbb{R} with ε>0\varepsilon>0. As FF is continuous on [a,b]×Rn[a,b]\times\mathbb{R}^{n}, it is continuous at (ψ(u),v)(\psi(u),v) relative to [a,b]×Rn[a,b]\times\mathbb{R}^{n} (Continuous Map Between Metric Spaces (ii)), which gives δ∈R\delta\in\mathbb{R} with δ>0\delta>0 such that dE(F(s,w),F(ψ(u),v))<εd_{E}\bigl(F(s,w),F(\psi(u),v)\bigr)<\varepsilon for every (s,w)∈[a,b]×Rn(s,w)\in[a,b]\times\mathbb{R}^{n} whose product-metric distance from (ψ(u),v)(\psi(u),v) is less than δ\delta. Let (u′,w)∈[a,b]×Rn(u',w)\in[a,b]\times\mathbb{R}^{n} have product-metric distance less than δ\delta from (u,v)(u,v). By The Product Metric is a Metric (iii), ∣u−u′∣<δ|u-u'|<\delta and dE(v,w)<δd_{E}(v,w)<\delta, with dR(s,s′)=∣s−s′∣d_{\mathbb{R}}(s,s')=|s-s'| as in Metric Spaces: Balls, the Metric Topology, and the Real and Complex Numbers as Metric Spaces (iv). As ∣ψ(u)−ψ(u′)∣=∣u−u′∣<δ|\psi(u)-\psi(u')|=|u-u'|<\delta, the same clause shows that (ψ(u′),w)(\psi(u'),w) has distance less than δ\delta from (ψ(u),v)(\psi(u),v), and by (∗)(\ast) and Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square (ii),

dE(G(u′,w),G(u,v))=∥F(ψ(u′),w)−F(ψ(u),v)∥=dE(F(ψ(u′),w),F(ψ(u),v))<ε.d_{E}\bigl(G(u',w),G(u,v)\bigr)=\bigl\lVert F\bigl(\psi(u'),w\bigr)-F\bigl(\psi(u),v\bigr)\bigr\rVert=d_{E}\bigl(F(\psi(u'),w),F(\psi(u),v)\bigr)<\varepsilon .

So GG is continuous at (u,v)(u,v) relative to [a,b]×Rn[a,b]\times\mathbb{R}^{n}, and hence continuous on [a,b]×Rn[a,b]\times\mathbb{R}^{n} by Continuous Map Between Metric Spaces (ii).

Existence and uniqueness. Let ξ∈Rn\xi\in\mathbb{R}^{n}. By The Picard–Lindelöf Theorem on a Closed Interval: Existence, Uniqueness and Continuous Dependence for Globally Lipschitz Systems (i), with GG for FF and LL for LL, the initial value problem of The Initial Value Problem for a First-Order System of Ordinary Differential Equations (i) with right-hand side GG and initial value ξ\xi has exactly one solution y0y_{0}. Let x0:[a,b]→Rnx_{0}:[a,b]\to\mathbb{R}^{n} be the function with x0(t)=y0(ψ(t))x_{0}(t)=y_{0}(\psi(t)) for every t∈[a,b]t\in[a,b]. The function u↦x0(ψ(u))u\mapsto x_{0}(\psi(u)) has the values y0(ψ(ψ(u)))=y0(u)y_{0}(\psi(\psi(u)))=y_{0}(u), so it is y0y_{0} (Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (iv)); by (i), applied to x0x_{0} and y0y_{0}, x0x_{0} is differentiable with x0′(t)=F(t,x0(t))x_{0}'(t)=F(t,x_{0}(t)) for every t∈[a,b]t\in[a,b] and x0(b)=ξx_{0}(b)=\xi. Let x1,x2:[a,b]→Rnx_{1},x_{2}:[a,b]\to\mathbb{R}^{n} both be differentiable with xi′(t)=F(t,xi(t))x_{i}'(t)=F(t,x_{i}(t)) for every t∈[a,b]t\in[a,b] and xi(b)=ξx_{i}(b)=\xi, and let y1,y2y_{1},y_{2} be the functions of (i) for x1,x2x_{1},x_{2}, so yi(u)=xi(ψ(u))y_{i}(u)=x_{i}(\psi(u)). By (i), y1y_{1} and y2y_{2} are solutions of the initial value problem with right-hand side GG and initial value ξ\xi, so y1=y2=y0y_{1}=y_{2}=y_{0}. As xi(t)=yi(ψ(t))x_{i}(t)=y_{i}(\psi(t)) for every t∈[a,b]t\in[a,b], as shown in the proof of (i), x1(t)=y1(ψ(t))=y2(ψ(t))=x2(t)x_{1}(t)=y_{1}(\psi(t))=y_{2}(\psi(t))=x_{2}(t) for every t∈[a,b]t\in[a,b], and x1=x2x_{1}=x_{2} (Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (iv)).

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…