MATH 337 – Numerical Differential Equations
Part I: Initial-value problems for Ordinary Differential equations (ODEs).
1. Review. Euler method and its modifications for 1st-order ODEs.
• Taylor series.
• Existence and uniqueness theorem for ODEs.
• Analytic solution of a linear, inhomogeneous, 1st-order ODE.
• Simple (forward) Euler method.
• Truncation error and global error of the Euler method.
• Modifications of the Euler method: Modified Euler and Midpoint methods.
• How to improve the global error: Romberg method.
2. Runge-Kutta (RK) methods for 1st-order ODEs.
• Modified Euler and Midpoint methods as particular cases of RK methods.
• Higher-order RK methods; the Classical RK method.
• Estimation of the step size to achieve a prescribed accuracy; Runge-Kutta-Fehlberg method.
3. Multi-step methods; Predictor-corrector methods; Implicit methods for 1st-order ODEs.
• The idea behind the multi-step methods.
• Adams-Bashford, Leap-frog, and a 3rd-order divergent multi-step methods.
• Predictor-corrector methods.
• Implicit 1st and 2nd-order Euler methods.
4. Consistency, stability, and convergence of numerical methods for 1st-order ODEs.
• Definitions and main theorem: “consistency+stability =⇒ convergence”.
• The concept of stability of the analytical solution of a 1st-order linear ODE.
• Investigation of stability of methods introduced in previous lectures. Conditionally stable,
absolutely unstable, and absolutely stable methods.
5. Higher-order ODEs and systems of ODEs.
• Generalizations of the studied methods to systems of 1st-order ODEs.
• Higher-order ODEs as systems of 1st-order ODEs.
• Special discretization schemes for 2nd-order ODEs: Simple central-difference and higher-order
(Numerov’s) methods.
• Symplectic methods: simple examples (symplectic Euler and Verlet-Störmer methods), and
why they (sometimes) work better than non-symplectic methods.
• Stability of numerical methods for systems of ODEs and higher-order ODEs.
• Stiff equations.
Part II: Boundary-value problems (BVP) for ODEs.
6. Introduction. (Few available theorems about) existence and uniqueness of solutions of BVPs.
7. The shooting method.
• Shooting method for linear BVPs with Dirichlet and Neumann boundary conditions.
• Multiple shooting method.
• Shooting method for nonlinear BVPs.
8. Finite-difference methods for BVPs.
• The matrix problem for the discretized solution of a linear BVP with the Dirichlet boundary
conditions. Existence and uniqueness of solution to a matrix problem where the matrix is
diagonally dominant. Gerschgorin circles and eigenvalues of a matrix.
• Higher-order discretizations and an estimate of the error.
1
• Other types of boundary conditions.
• Iterative solutions of nonlinear BVPs.
9. Concepts behind Finite-Element methods.
• The collocation method.
• The Galerkin method.
10. Eigenvalue problems.
• Definition of eigenvalues and eigenfunctions of BVPs.
• Issues in finding eigenvalues of large matrices.
• Householder transformation of a matrix to an upper-Hessenberg form.
• Power method and its modifications.
• The shooting method.
Part III: Partial-differential equations (PDEs).
11. Classification of PDEs.
• Clasification of physical problems leading to PDEs.
• Clasification of PDEs. The concept, and significance, of charcteristics.
12. Parabolic PDEs in 1 spatial dimension (Heat equation): Simple explicit method and stability
analyses.
• Formulation of the model problem and the general property of the analytical solution.
• The simple explicit method for the Heat equation.
• The matrix and von Neumann stability analyses.
• Explicit methods of higher order.
• Effect of smoothness of the initial condition on the order of consistency of a numerical method.
13. Implicit methods for the Heat equation.
• Derivation of the Crank-Nicolson (CN) scheme.
• Truncation error of the CN scheme.
• Stability analysis of the CN and related methods.
• Improving the accuracy of the CN method.
14. Numerical methods for generalizations of the one-dimensional Heat equation.
• Derivative (Neumann) boundary conditions.
• Equations with variable coefficients.
• Nonlinear parabolic equations.
15. Heat equation in 2 and 3 spatial dimensions.
• The explicit method and the CN method for the 2-D Heat equation.
• Criteria for a computationally-efficient scheme in 2 dimensions.
• Alternating-Direction Implicit (ADI) methods: Peaceman-Rachford, Douglas-Rachford, and
D’yakonov methods.
16. Wave equation and other hyperbolic-type PDEs. (Covered if time permits)
• Simple difference methods for first-order hyperbolic equations.
• The Courant-Friedrichs-Lewy stability condition.
• Other methods: method of characteristics, Lax-Wendroff method, simple implicit method.
