Abstract:In this paper, we prove that, under precise spectral assumptions, some finite difference approximations of scalar leftgoing transport equations on the positive half-line with numerical boundary conditions are $\ell^1$-stable but $\ell^q$-unstable for any $q>1$. The proof relies on the accurate description of the Green's function for a particular family of finite rank perturbations of Toeplitz operators whose essential spectrum belongs to the closed unit disk and with a simple eigenvalue of modulus $1$ embedded into the essential spectrum.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper aims to study the semigroup stability of a specific type of discretization method when solving hyperbolic initial - boundary value problems. Specifically, the author focuses on the explicit finite - difference scheme for the scalar left - traveling transport equation (i.e., velocity \(v < 0\)) with numerical boundary conditions on the positive half - axis.
The main research question is: when the discretization operator \(T\) has a simple eigenvalue with modulus 1 in its essential spectrum, is this numerical scheme stable? Specifically, the author has proven that under certain precise spectral assumptions, some finite - difference approximations are stable in the \(\ell^1\) norm, but are unstable in the \(\ell^q\) norm for any \(q > 1\).
### Main conclusions
The main conclusions of the paper can be summarized as follows:
- **Theorem 1**: Suppose assumptions 1 and 2 are satisfied.
- If \(B(1,\dots,1)^T\in BE_s(1)\), then the numerical scheme (1.3) is \(\ell^q\)-stable for all \(q\in[1,+\infty]\).
- If \(B(1,\dots,1)^T\notin BE_s(1)\), then the numerical scheme (1.3) is \(\ell^1\)-stable, but is \(\ell^q\)-unstable for all \(q\in]1,+\infty]\). Moreover, for all \(q\in]1,+\infty]\), there exists a positive constant \(C\) such that:
\[
\forall n\in\mathbb{N}\setminus\{0\}, \|T^n\|_{L(H_q)}\geq Cn^{1 - \frac{1}{q}}.
\]
### Methods and tools
To prove the above conclusions, the author uses the following methods and tools:
1. **Green function**: By introducing the space - time Green function \(G(n,j_0,j)\), the long - time behavior of the numerical scheme is described.
2. **Spectral analysis**: Using the properties of the adjoint matrix \(M(z)\) and the Lopatinskii determinant \(\Delta(z)\), the spectral structure of the discretization operator \(T\) is analyzed in detail.
3. **Asymptotic expansion**: By analyzing the asymptotic behavior of the generalized Gaussian wave, the stability results of the numerical scheme are obtained.
### Significance and impact
The significance of this research lies in revealing the stability behavior of numerical schemes under specific spectral conditions. In particular, it shows that even when the Godunov - Ryabenkii condition holds but the uniform Kreiss - Lopatinskii condition does not hold, the numerical scheme may still exhibit different stability characteristics. This finding helps to gain a deeper understanding of the behavior of numerical discretization methods when dealing with hyperbolic partial differential equations and provides a theoretical basis for future research.