lacunary - Mathnotes

Euler's Method

We'll consider a first order IVP of the form

dudtf(t,u),u(0)=c,

with the goal of numerically estimating the solution on a finite time interval [0,T] for some T>0.

First we'll establish this is a reasonable thing to do.

Theorem: Well-posedness of IVPs \@{well-posedness-of-ivps}

Consider the IVP

dudt=f(t,u),u(0)=c.

Suppose that f:[0,T]×Rd→Rd is continuous in its domain and @Lipschitz continuous in u, i.e., there exists a constant L>0 such that

|f(t,u)−f(t,v)|≤L‖u−v‖,∀(t,u),(t,v)∈[0,T]×Rd.

Then it holds that:

  1. Existence and Uniqueness: There exists a unique solution u(t) for any u0∈Rd.
  2. Continuous Dependence on initial data and right hand side: Suppose r:[0,T]×Rd→Rd is a bounded @perturbation of f, i.e. there exists a constant M>0 such that ‖r(t,u)‖≤M,∀(t,u)∈[0,T]×Rd. Further let u^(t) be the solution of the perturbed IVP du^dt=f(t,u^)+r(t,u^),u^(0)=c^, which we assume exists and is unique. Then we have ‖u(t)−u^(t)‖≤eLt‖c−c^‖+ML(eLt−1),∀t∈[0,T].

Now that we've established that IVPs are well posed, we can look at how to approximate our unknown function u(t) at a discrete set of points in time.

Definition: discrete time points (also: time mesh points, time grid points) (also defines: step size) \@{discrete-time-points}

Consider a @partition 0=t0<t1<⋯<tN=T of the time interval [0,T]. We can call the points tn the discrete time points. Then our goal is to approximate the values u(tn) with a set of quantities {un}n=0N, that is, un≈u(tn) for n=0,1,…,N. We can also write δtn=tn+1−tn to denote the step size (also: time step) between between consecutive points. For simplicity we often assume that δtn=δt for all n=0,1,…,N−1, i.e. the time steps are constant. Such a mesh is called a @uniform mesh.