Syllabus / Part I — Foundations
Module 1
Linear Algebra, Reloaded
The geometric linear algebra your first course skipped: matrices as transformations, kernels, determinants, eigenvectors — the language of every system of ODEs and every PDE solver.
Vectors as arrows and as lists of numbers; linear combinations; span, linear independence and basis — seen geometrically.
🌊 A velocity field assigns a vector to every grid cell; everything in a fluid solver is built from vector operations.
A matrix is a function that moves space: columns are where the basis vectors land. Matrix multiplication is composition.
🌊 The velocity gradient tensor is a matrix that describes how a tiny blob of fluid is stretched, sheared and rotated.
What Gaussian elimination was really doing: when Ax = b has solutions, the null space (kernel), column space, rank, and the rank–nullity theorem.
🌊 The pressure solve in an incompressible fluid solver is a giant Ax = b; its kernel (constant pressure) is why pressure is only defined up to a constant.
The determinant measures how a transformation scales area/volume and whether it flips orientation; det = 0 means space is squashed.
🌊 Incompressible flow preserves volume: the flow map has Jacobian determinant 1.
Directions a transformation only stretches. Finding them via det(A − λI) = 0, and why they are the key to solving linear systems of ODEs.
🌊 Eigenvalues of a discretized operator decide whether a simulation is stable — the CFL condition is an eigenvalue bound.
Looking at a transformation in its eigenvector basis turns it into simple scaling: A = PDP⁻¹, powers of matrices, and symmetric matrices.
🌊 Normal modes of vibrating structures and Fourier modes of a grid are both eigenbases that decouple a hard problem into many easy ones.
When eigenvalues are complex, the transformation rotates. Complex numbers, Euler's formula e^{iθ}, and spirals.
🌊 Vortices and oscillating wakes behind a cylinder show up as complex eigenvalues of the linearized flow.
From iterating x_{n+1} = A x_n to continuous growth: defining e^{At} by power series and via diagonalization, and why it solves x' = Ax.
🌊 Every explicit time-stepper approximates a matrix exponential; how well it does so is its accuracy and stability.