Numerical methods in computing

Many equations that arise in physics, chemistry, and engineering — most differential equations describing real systems, for instance — simply have no closed-form solution: no finite combination of elementary functions (polynomials, exponentials, trigonometric functions, and so on) can be written down that satisfies the equation exactly. Rather than giving up, numerical methods solve such problems approximately, by breaking a difficult continuous problem into a large number of small, simple, repeated computational steps that a computer can execute very quickly, even though no human could realistically carry out the same calculation by hand.

Euler's method, one of the simplest numerical techniques, illustrates the general idea clearly. Given a differential equation describing how a quantity changes (its derivative) as a function of its current value, and a known starting point, Euler's method approximates the solution curve by taking a small step forward, computing the derivative at the current point, and using it to estimate the quantity's new value a short step later — as if the true, generally curved solution were locally a straight line over that short interval. This estimated new point becomes the starting point for the next step, and the process repeats: compute the local slope, step forward, repeat, building up an approximate solution curve step by step, entirely through repeated arithmetic rather than symbolic manipulation.

Because each step approximates a genuinely curved solution with a straight line, Euler's method introduces error at every step, and these small errors tend to accumulate over many steps. The step size controls this trade-off directly: a smaller step size tracks the true curve more closely, since a straight line is a better local approximation to a curve over a shorter interval, reducing the error introduced at each step — but reaching the same total distance along the curve then requires proportionally more steps, so the total computation takes longer. Choosing an appropriate step size, small enough to keep accumulated error acceptably small but large enough to finish the computation in reasonable time, is itself a central practical skill in applying any numerical method, and it is why numerical methods, though only approximate, remain indispensable throughout physics and chemistry, wherever a problem's exact solution is unknown or impractical to compute directly.