runge_kutta

Runge-Kutta Methods

Definition

Runge-Kutta methods are a family of numerical integration schemes for $\frac{dy}{dx}=f(x,y)$ that achieve higher accuracy than the Euler method by evaluating $f$ at intermediate points within each step.

Intuition

Rather than using only the slope at the start of the step (Euler), Runge-Kutta methods take a weighted average of slopes sampled at several interior points. This cancels out higher-order error terms from the Taylor expansion, achieving accuracy $O(\Delta x^p)$ for a $p$th-order method — with no extra cost from reducing the step size. The classical RK4 achieves fourth-order accuracy with just four slope evaluations per step.

Formal Description

First-order Runge-Kutta (Euler’s method)

$$k_1=\Delta xf(x_n,y_n),y_{n+1}=y_n+k_1.$$

Second-order Runge-Kutta

$$k_1=\Delta xf(x_n,y_n),k_2=\Delta xf(x_n+\alpha \Delta x,y_n+\beta k_1),y_{n+1}=y_n+a{k}_1+b{k}_2.$$

The parameters $\alpha ,\beta ,a,b$ must satisfy the consistency conditions

$$a+b=1,\alpha b=\beta b=\frac{1}{2}.$$

Different choices of these parameters yield distinct second-order methods (e.g.\ the midpoint method with $\alpha =\beta =\frac{1}{2}$, $a=0$, $b=1$; or Heun’s method with $\alpha =\beta =1$, $a=b=\frac{1}{2}$).

Key Results

• The classical fourth-order Runge-Kutta (RK4) method uses four $k$-evaluations per step and achieves $O(\Delta x^4)$ global error.
• Runge-Kutta methods are self-starting and do not require values at previous steps.
• They are the standard choice for numerically solving ODEs when higher accuracy than Euler’s method is needed.

Applications

• Standard workhorse for non-stiff ODE integration in scientific computing (scipy.integrate.solve_ivp default method is RK45).
• Trajectory simulation in physics, orbital mechanics, and robotics.
• Adaptive step-size control by comparing solutions of different orders (embedded Runge-Kutta pairs such as Dormand-Prince RK45).

Trade-offs

• Each step requires multiple function evaluations of $f$; if $f$ is expensive to compute, methods that reuse previous evaluations (Adams-Bashforth) may be preferable.
• Standard Runge-Kutta methods are explicit and can be unstable for stiff equations; implicit variants (e.g.\ implicit RK) are needed for stiff problems but require solving a nonlinear system per step.
• Higher-order accuracy is only realised when the solution is sufficiently smooth; discontinuities in $f$ can degrade accuracy to first order.

Links

• Euler Method