17.1.1.2. Truncation Error

Detailed Summary

Truncation error is a significant aspect of numerical methods used in solving Ordinary Differential Equations (ODEs). It results from approximating an infinite process with a finite one, an unavoidable consequence of using numerical techniques to gain practical solutions when analytical solutions are challenging or impossible.

Key Components of Truncation Error:

1. Local Truncation Error (LTE): This is the error introduced in a single step of the numerical method, dependent on how closely the method approximates the actual derivative at that step. For example, in Euler's method, the LTE is given by:

   $LTE = y(x_{n+1}) - y_{n+1}$

   where $y(x_{n+1})$ is the exact solution and $y_{n+1}$ is the numerical solution obtained using the finite step size $h$. The order of LTE gives insights into error magnitude—in Euler's method, it is $O(h^2)$, while for a fourth-order Runge-Kutta method, it is $O(h^5)$.

2. Global Truncation Error (GTE): This error accumulates from all previous LTEs over multiple steps. If $N$ is the number of steps, then:

   $GTE = N imes LTE = \frac{(b - a)}{h} imes O(h^p) = O(h^{p-1})$

   highlighting that, for instance, Euler’s method has a GTE of $O(h)$, and the Runge-Kutta method has $O(h^4)$.

3. Order of a Method: The higher the order of a method, the faster the error decreases as the step size $h$ diminishes. An important point to remember is that the order of the method (denoted as $p$) indicates how the error is proportional to $h^p$.

4. Stability and Convergence: Stability assesses how errors evolve during computation, requiring that small perturbations do not yield larger errors. A method converges if as $h$ approaches zero, the numerical solution approaches the exact one.

Understanding truncation errors and effectively controlling them ensures numerical solutions of ODEs remain reliable for practical applications.