Amplification polynomial of a multilevel finite difference scheme

For a schemethe Fourier ansatz gives the amplification polynomialIts roots are the amplification factors of that Fourier mode.

Centered space and leapfrog time discretization of $u_t+c{u}_x=0$ givesIts amplification polynomial is $G^2+2i\nu \sin \theta G-1$. The two roots have unit modulus for $|\nu \sin \theta |\le 1$, but a repeated unit root at equality violates the uniform root condition.

For $u_t=u_x+u_y$ with equal spatial mesh sizes, the centered leapfrog recurrence has amplification polynomial of a multilevel finite difference scheme $G^2-2i\mu (\sin \xi +\sin \eta )G-1$. Uniform stability for arbitrary two-level starting perturbations holds exactly for $0<\mu <1/2$. At $\mu =1/2$ the phase pair $(\pi /2,\pi /2)$ gives $(G-i{)}^2$, and its companion matrix has a nontrivial Jordan block, producing growth proportional to the number of steps. For larger $\mu$ one polynomial root has modulus greater than one. Bounds on eigenvalue moduli alone therefore give a misleading non-strict endpoint.

For the centered two-level advection recurrence, the Fourier roots at Courant number one coalesce at $i$ for phase $\pi /2$. The resulting nontrivial Jordan block admits linearly growing solutions despite both eigenvalues having modulus one. Uniform two-level stability for arbitrary starting perturbations therefore requires the strict inequality $0<\mu <1$. A specially consistent startup can select the exact translating branch at the boundary, which is a restriction on the data rather than a uniform bound on the full update.

Applying a centered leapfrog step in time and the centered second difference in space to $u_t=u_{xx}$ givesFor every $\mu >0$, each nonconstant Fourier mode has an amplification root of modulus greater than one, so the scheme is unconditionally unstable.