Part 3: Numerical Methods

Scientific Computing with Python | COMP 536

Author

Anna Rosen

Numerical Methods for Scientific Modeling

This part translates core mathematical ideas into practical algorithms you can trust in computational science. You will move from finite-difference thinking to robust root-finding, numerical integration, and time evolution methods that preserve important physical structure.

These readings are designed as working references. Return to them when your simulation drifts, your solver fails to converge, or your numerical choices start controlling your scientific conclusions.

NoteHow to Use This Part

Work module by module, but revisit sections by need. If you are debugging derivatives, start in Module 1. If you are solving equilibrium conditions or computing totals, start in Module 2. If you are integrating dynamics over time, start in Module 3.


Why This Part Exists

Computational models are only as good as their numerical foundations. In practice, scientific code fails less often from syntax errors and more often from approximation errors, unstable timesteps, poor conditioning, and methods that violate the structure of the underlying physics.

Part 3 focuses on this reliability layer:

  • what errors we introduce when we discretize continuous math,
  • how those errors scale,
  • how to choose methods by problem characteristics,
  • and how to keep long-run simulations scientifically meaningful.

Module-by-Module Guide

Module 1: Foundations of Discrete Computing

Build intuition for finite differences, truncation versus round-off trade-offs, and Taylor-series-based error analysis.

Module 2: Static Problems and Quadrature

Learn how to find equilibrium points (roots) and compute accumulated quantities (integrals) with methods chosen for robustness and efficiency.

Module 3: ODE Methods and Conservation

Study time integration methods, stability, and structure-preserving ideas for long-run dynamical simulations.


Quick Navigation Guide

If you need to… Go to… Focus on…
Choose a safe finite-difference step size Module 1 truncation vs round-off trade-off
Diagnose unstable root-finding Module 2 convergence behavior and hybrid methods
Select an integration rule for accumulated quantities Module 2 quadrature error scaling
Keep long-time dynamics physically meaningful Module 3 symplectic methods and stability
Troubleshoot timestep explosions Module 3 stability regions and stiff systems

What to Revisit During Projects

  • Project work with derivatives or gradients: revisit Module 1 sections on finite differences and error control.
  • Project work with equilibrium or parameter solves: revisit Module 2 root-finding method selection.
  • Project work with cumulative physical quantities: revisit Module 2 quadrature accuracy and method trade-offs.
  • Project work with trajectories, dynamics, or iterative updates over time: revisit Module 3 stability and conservation topics.

These readings are meant to support decisions in real code, not just chapter-by-chapter completion.


Next step: Start with Module 1 overview.