Computational Physics And Numerical Methods

Computational physics is the practice of solving physics problems with computers when pen-and-paper mathematics runs out. Many real problems, from turbulent airflow to the motion of galaxies, have no exact formula, so physicists approximate the answer with numerical methods: recipes that turn calculus into millions of small arithmetic steps.

It matters because the modern world is designed by simulation. Weather forecasts, car crash tests, aircraft wings, microchips and movie explosions are all computed before they are built or filmed, saving years of trial and error and making experiments possible that could never be done in a laboratory.

Key Concepts

• Discretisation: chopping continuous space or time into small steps of size h.
• Truncation error: the error from approximating curves with straight steps; it shrinks as h shrinks.
• Euler's method: the simplest ODE solver: new value = old value + h × rate of change.
• Runge-Kutta (RK4): a far more accurate stepper that samples the rate four times per step.
• Numerical integration: estimating areas under curves with the trapezoidal rule or Simpson's rule.
• Root finding: locating where a function equals zero, by bisection or Newton-Raphson iteration.
• Stability: a method is stable if small errors do not explode as the computation proceeds.
• Convergence: the computed answer approaches the true answer as the step size shrinks.

How It Works

1. Write the equations: express the physics as differential or integral equations.
2. Choose a method: pick Euler for simplicity or RK4/Simpson for accuracy, balancing speed against error.
3. Discretise: replace smooth curves with steps of size h and derivatives with differences.
4. Iterate: let the computer repeat the simple update rule millions of times.
5. Verify: halve h and re-run; if the answer barely changes, it has converged and can be trusted.

Worked Example

Estimate the integral of x² from 0 to 2 using the trapezoidal rule with step h = 0.5.

1. Sample points: x = 0, 0.5, 1, 1.5, 2, giving y = 0, 0.25, 1, 2.25, 4.
2. Trapezoidal sum = h × (y₀/2 + y₁ + y₂ + y₃ + y₄/2).
3. = 0.5 × (0 + 0.25 + 1 + 2.25 + 2) = 0.5 × 5.5 = 2.75.
4. Exact value = 8/3 ≈ 2.667, so the estimate is within about 3%.

Halving the step to h = 0.25 would cut the error roughly by a factor of four, demonstrating convergence in action.

Common Misconceptions

• "Computer answers are exact." They are approximations with controlled error. A simulation is only as good as its method, step size and underlying model.
• "Smaller steps are always better." Extremely tiny steps accumulate rounding error and take forever; good computing balances truncation error against round-off and cost.
• "Simulation replaces experiments." It complements them. Every model must be validated against real measurements, because a beautiful simulation of wrong physics is still wrong.

Real-World Applications

• Weather and climate: supercomputers integrate fluid equations over the whole atmosphere.
• Engineering: crash simulations and airflow (CFD) testing replace many physical prototypes.
• Medicine: protein folding and drug-binding simulations guide new medicines.
• Electronics: chip designers simulate billions of transistors before manufacturing.
• Astrophysics: galaxy formation and black hole mergers are studied purely through simulation.

Computational physics does not replace physical insight; it extends it, letting a simple rule repeated a billion times reveal behaviour no human could calculate by hand.