Initial Value Problems
u′(t)=f(t,u(t)),a≤t≤b,u(a)=u0
If the equation is linear, u′(t)=g(t)+uh(t), then with ρ(t)=exp(−∫ath(s)ds),
ρ(t)u(t)=u0+∫atρ(s)g(s)ds
(see integrating factor).
If ∂u∂f≤L for all a≤t≤b and all u, the solution exists and is unique, and the solution u(t;u0+δ) with perturbed initial value u(a)=u0+δ satisfies
∥u(t;u0+δ)−u(t;u0)∥∞≤∣δ∣eL(b−a)
Higher order An equation of order m becomes a first-order system with u1=y, u2=y′, u3=y′′,…, e.g. for y′′=g(y,y′),
u1′=u2,u2′=g(u1,u2)
One-Step Methods
ti=a+ih, h=nb−a, ui≈u^(ti) with u^ the exact solution, and
ui+1=ui+hϕ(ti,ui,h),i=0,…,n−1
The local truncation error substitutes the exact solution into the method and divides by h,
τi+1(h)=hu^(ti+1)−u^(ti)−ϕ(ti,u^(ti),h)
and the method is consistent if τ→0 as h→0. If ∣τi+1(h)∣≤Chp and ∂u∂ϕ≤L for all a≤t≤b, u and h>0, then the global error satisfies
∣u^(ti)−ui∣≤LChp[eL(ti−a)−1]=O(hp)
as h→0, but not as t→∞. So the global error has the same order of accuracy as the LTE.
Euler's Method
ϕ=f,
ui+1=ui+hf(ti,ui),i=0,…,n−1
By Taylor expansion,
u^(ti+1)−[u^(ti)+hf(ti,u^(ti))]=[u^(ti)+hu^′(ti)+21h2u^′′(ti)+⋯]−[u^(ti)+hu^′(ti)]=21h2u^′′(ti)+O(h3)
so the LTE is O(h) and Euler is consistent and first order: an O(h2) error per step over O(h−1)=O(n) steps gives a global error of O(h).