ODE Integration

Integrates ODEs while assessing accuracy, stability, and computational cost.

Conceptual context

Many physical laws are written as differential equations because they describe how a quantity changes from its current state. When no useful analytical solution exists, continuous evolution is replaced by a sequence of numerical steps. That replacement introduces non-neutral choices. $h$ sets the resolution, the integration method determines how slopes are combined, and $N$ sets the basic cost. The central task is to distinguish an accurate approximation from a curve that looks smooth but is unstable. This requires comparing $y_num$, $y_ref$, and $E_g$ over the interval from $t_0$ to $t_f$.

🟢 Essential level

Integrating an ODE means reconstructing how $y$ evolves from its instantaneous rate $\frac{dy}{dt}$.

A computer does not move continuously through time: it advances by increments $h$ and calculates a $y_{n+1}$ at each update. Euler uses one slope and is simple, but it accumulates more error. Heun corrects the prediction by averaging $k_1$ and $k_2$. RK4 combines $k_1$, $k_2$, $k_3$, and $k_4$ to represent curvature within the step more accurately. Formal algebraic details of these corrections are introduced at the formal level below.

A smaller step often improves accuracy, although it increases $N$ and computational cost. A smooth-looking curve does not by itself guarantee a correct result: the algorithm can silently invent oscillation, spurious growth, or the wrong decay rate. The basic check is repeating the calculation with a smaller step and observing whether the result settles down. Stability requires more than a plausible shape: the method must avoid artificial oscillation or growth. Comparing $y_num$ with $y_ref$ gives $E_g$ and tests convergence under refinement. The correct choice depends on physical scales, required tolerance, stiffness, and the computational budget available for repeated derivative evaluations.

🔵 Formal level

Basic formulas

The dynamics is expressed as a first-order ODE relating $\frac{dy}{dt}$ to $t$ and $y$.

First-order ODE

Explicit Euler advances from the current state using the slope at the start of the interval.

Explicit Euler step

Heun's method forms a prediction using $k_1$ and corrects it with $k_2$, so it uses information from the beginning and estimated end.

Heun step

RK4 calculates four slopes, $k_1$, $k_2$, $k_3$, and $k_4$, at different locations within the interval. Their weighted combination gives a fourth-order approximation.

Fourth-order Runge-Kutta step

Real student questions

Is a smaller step always better? Only up to a point. It reduces truncation error but increases cost and may expose roundoff.

Is RK4 always superior to Euler? No. It is often more accurate for smooth problems, but it costs more and does not solve stiffness by itself.

Does a smooth graph guarantee correctness? No. A wrong solution can be smooth. A convergence test is required.

What does stability mean? The method does not artificially amplify numerical perturbations in the regime being studied.

Why compare with half the step? It shows whether the solution approaches a limit and reveals the effective method order.

Final synthesis

$h$ controls resolution and cost; the method controls how slope information is used. A reliable solution must be dimensionally consistent, stable, and convergent. $E_g$ and refinement tests turn a computed curve into a physically defensible approximation.