TheoremBase

Proof of Global Existence and Uniqueness for Lipschitz Ordinary Differential Equations in Integral Form

theoremthm:picard-lindelof-global-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
· 4,855 chars · 19 deps · depth 16 Reason: Kalman-Bucy phase Block B: weighted-metric contraction proof of the global Picard-Lindelof theorem; internally reviewed and validated; batch-approved by Aaron on 2026-07-31.

Proof

Let (C,d∞)(\mathcal{C},d_{\infty}) be the nonempty complete metric space of functions [a,b]→Rk[a,b]\to\mathbb{R}^{k} with continuous components from Completeness of the Space of Continuous Vector-Valued Functions under the Supremum Metric, and write ∣x∣=d(x,0)|x|=d(x,0), so that d(x,y)=∣x−y∣d(x,y)=|x-y| by claim 1 of Componentwise Estimates, Transpose Identities, and Indefinite Riemann Integrals, whose componentwise inequalities we use freely.

The solution operator. For h∈Ch\in\mathcal{C} define

(Th)(t)=ξ+∫atF(r,h(r)) dr(a≤t≤b).(Th)(t)=\xi+\int_a^tF\bigl(r,h(r)\bigr)\,dr\qquad(a\le t\le b).

By hypothesis (i) the integrand has continuous components, so the componentwise Riemann integrals exist (Continuous Functions on a Closed Interval are Riemann Integrable) and each component of ThTh is continuous on all of [a,b][a,b], including the endpoints, by claim 4 of Componentwise Estimates, Transpose Identities, and Indefinite Riemann Integrals. Thus T:C→CT:\mathcal{C}\to\mathcal{C}, and a function xx is as in the conclusion exactly when it is a fixed point of TT.

Basic integral bound. For h,h′∈Ch,h'\in\mathcal{C} and a≤t≤ba\le t\le b, claim 5 of Componentwise Estimates, Transpose Identities, and Indefinite Riemann Integrals applied to w(r)=F(r,h(r))−F(r,h′(r))w(r)=F(r,h(r))-F(r,h'(r)), together with hypothesis (ii) and monotonicity of the Riemann integral on continuous integrands (Agreement of the Riemann and Lebesgue Integrals for Continuous Functions on a Closed Interval with Linearity and Monotonicity of the Lebesgue Integral; the function r↦d(h(r),h′(r))r\mapsto d(h(r),h'(r)) is continuous by the componentwise estimates), gives, with K=kLK=kL,

d(Th(t),Th′(t))=∣∫atw(r) dr∣≤k∫at∣w(r)∣ dr≤K∫atd(h(r),h′(r)) dr.d\bigl(Th(t),Th'(t)\bigr)=\Bigl|\int_a^tw(r)\,dr\Bigr|\le k\int_a^t\bigl|w(r)\bigr|\,dr\le K\int_a^t d\bigl(h(r),h'(r)\bigr)\,dr .

Case K=0K=0. Then Th=Th′Th=Th' for all h,h′h,h', so TT is a contraction (constant 12\tfrac12) on (C,d∞)(\mathcal{C},d_{\infty}), and Contraction Mapping Theorem on a Nonempty Complete Metric Space gives a unique fixed point.

Case K>0K>0: weighted metric. With the exponential function, define, for h,h′∈Ch,h'\in\mathcal{C},

dw(h,h′)=sup⁡t∈[a,b]exp⁡(−2K(t−a)) d(h(t),h′(t)).d_{w}(h,h')=\sup_{t\in[a,b]}\exp\bigl(-2K(t-a)\bigr)\,d\bigl(h(t),h'(t)\bigr).

Since exp⁡(−2K(b−a))≤exp⁡(−2K(t−a))≤1\exp(-2K(b-a))\le\exp(-2K(t-a))\le1 on [a,b][a,b] (positivity and monotonicity from Basic Properties of the Exponential Function), every value defining dwd_w lies between exp⁡(−2K(b−a))\exp(-2K(b-a)) times and 11 times the corresponding value defining d∞d_\infty; taking suprema,

exp⁡(−2K(b−a)) d∞(h,h′)≤dw(h,h′)≤d∞(h,h′)<∞.\exp\bigl(-2K(b-a)\bigr)\,d_{\infty}(h,h')\le d_{w}(h,h')\le d_{\infty}(h,h')<\infty .

dwd_w is a metric on C\mathcal{C}: nonnegativity and symmetry are clear; dw(h,h′)=0d_w(h,h')=0 forces d∞(h,h′)=0d_\infty(h,h')=0 by the left inequality, hence h=h′h=h'; and multiplying the pointwise triangle inequality for dd by the positive weight exp⁡(−2K(t−a))\exp(-2K(t-a)) and passing to suprema gives the triangle inequality, exactly as in claim 1 of Completeness of the Space of Continuous Vector-Valued Functions under the Supremum Metric.

(C,dw)(\mathcal{C},d_w) is complete. Let (hn)(h_n) be Cauchy for dwd_w; given η>0\eta>0, applying the Cauchy property with ηexp⁡(−2K(b−a))\eta\exp(-2K(b-a)) and using d∞≤exp⁡(2K(b−a)) dwd_\infty\le\exp(2K(b-a))\,d_w shows (hn)(h_n) is Cauchy for d∞d_\infty; by Completeness of the Space of Continuous Vector-Valued Functions under the Supremum Metric it converges to some h∈Ch\in\mathcal{C} for d∞d_\infty; and dw(hn,h)≤d∞(hn,h)→0d_w(h_n,h)\le d_\infty(h_n,h)\to0 shows convergence for dwd_w.

Contraction estimate. Fix h,h′∈Ch,h'\in\mathcal{C} and t∈[a,b]t\in[a,b]. By the definition of dwd_w, d(h(r),h′(r))≤exp⁡(2K(r−a)) dw(h,h′)d(h(r),h'(r))\le\exp(2K(r-a))\,d_w(h,h') for every rr, and

∫atexp⁡(2K(r−a)) dr=exp⁡(2K(t−a))−12K:\int_a^t\exp\bigl(2K(r-a)\bigr)\,dr=\frac{\exp(2K(t-a))-1}{2K} :

indeed exp⁡(2K(r−a))=exp⁡(−2Ka)exp⁡(2Kr)\exp(2K(r-a))=\exp(-2Ka)\exp(2Kr) by Basic Properties of the Exponential Function, the function r↦exp⁡(−2Ka)exp⁡(2Kr)/(2K)r\mapsto\exp(-2Ka)\exp(2Kr)/(2K) has derivative exp⁡(2K(r−a))\exp(2K(r-a)) by Derivative of a Scaled Exponential Function and the constant-multiple rule of Sum and Product Rules for One-Dimensional Derivatives and Continuity, and Fundamental Theorem of Calculus, Part II in One Dimension evaluates the integral (degenerate t=at=a by the convention of Mean-Square Riemann Integral of a Family of Random Variables). Combining with the basic integral bound and monotonicity,

exp⁡(−2K(t−a)) d(Th(t),Th′(t))≤exp⁡(−2K(t−a)) K exp⁡(2K(t−a))−12K dw(h,h′)=12(1−exp⁡(−2K(t−a)))dw(h,h′)≤12 dw(h,h′).\exp\bigl(-2K(t-a)\bigr)\,d\bigl(Th(t),Th'(t)\bigr)\le\exp\bigl(-2K(t-a)\bigr)\,K\,\frac{\exp(2K(t-a))-1}{2K}\,d_w(h,h')=\tfrac12\Bigl(1-\exp\bigl(-2K(t-a)\bigr)\Bigr)d_w(h,h')\le\tfrac12\,d_w(h,h').

Taking the supremum over tt: dw(Th,Th′)≤12dw(h,h′)d_w(Th,Th')\le\tfrac12 d_w(h,h'), so TT is a contraction of the nonempty complete metric space (C,dw)(\mathcal{C},d_w).

By Contraction Mapping Theorem on a Nonempty Complete Metric Space, TT has exactly one fixed point x∈Cx\in\mathcal{C}. Fixed points of TT are precisely the functions in the conclusion, so such an xx exists and is unique. ■\blacksquare

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…