DMD eigenvalues (unit circle)
Temporal coefficients & reconstruction
DMD (Schmid 2010): given snapshots X = [x₀,...,x_{n-1}] and X' = [x₁,...,x_n], find the best-fit linear operator A: X'≈AX.
Algorithm: SVD of X → project A into low-rank subspace → eigendecompose → lift back. DMD modes are spatial structures, each associated with a complex frequency.
The vortex shedding field here is synthetic: mean flow + two von Kármán frequencies + noise. DMD separates these into distinct spatial modes at distinct frequencies.