A family of single-step numerical integration schemes for initial-value ordinary differential equations that combine weighted evaluations of the derivative at intermediate stages to achieve a desired order of accuracy and stability properties.

Principle

Advance the solution from one time level to the next by computing a set of stage evaluations (slopes) of the right-hand side, combine them with specific weights (Butcher tableau), and produce a high-order consistent update without multi-step history.

Demonstration

Use a classical fourth-order Runge–Kutta (RK4) to integrate y' = f(t,y): compute four slopes at intermediate times, combine them with weights 1/6, 1/3, 1/3, 1/6 and update y_{n+1} = y_n + h * weighted_sum to obtain fourth-order local accuracy.

Misapplication

Applying an explicit Runge–Kutta method with too large a time step on a stiff ODE can cause numerical instability; in stiff regimes implicit methods or special stiff-aware RK variants are required.

Consequence

Runge–Kutta methods offer flexible control of local truncation error, easy step-size adaptation (via embedded pairs), and straightforward parallel evaluation of stages in some implementations, making them widely used for ODE integration.

Reversal

Contrasts with linear multi-step methods: Runge–Kutta uses only current-step information and intermediate evaluations, while multi-step methods rely on past steps and can be more efficient for smooth problems when history is cheap.

Boundary

Designed for ordinary differential initial-value problems; boundary-value problems, differential-algebraic systems, or highly stiff problems may necessitate specialized methods or implicit RK variants.

Semantic Tension

Tension between explicit RK simplicity and implicit scheme stability: explicit RK is easy and cheap per stage but limited by stability regions, whereas implicit RK expands stability at the cost of nonlinear solves per step.

Synthesis

Runge–Kutta methods construct high-order single-step integrators by combining staged derivative evaluations according to a Butcher tableau, balancing accuracy, stability, and computational cost according to problem stiffness and step control.