Lipschitz stability for a piecewise linear Schrödinger potential from local Cauchy data

Giovanni Alessandrini, Maarten V. de Hoop, Romina Gaburro, Eva SincichAsymptot. Anal 2018 · OpenReview

TL;DR

This paper tackles the challenging problem of determining a spatially varying potential in a Schrödinger-like equation from limited boundary measurements. The authors establish a significant result: they can recover piecewise linear potentials with "Lipschitz stability," meaning small errors in the measurements lead to proportionally small errors in the recovered potential. This is a crucial step towards more robust inverse problem solving, particularly relevant for applications like seismic imaging.

What Problem Does This Paper Address?

The core problem is an "inverse boundary value problem." Imagine you have a physical system described by a partial differential equation, like the Schrödinger equation (or the reduced wave equation used in geophysics), where a certain property (the "potential" or "wave speed") is unknown. You can only measure certain quantities (like the pressure and particle velocity of a wave) on a portion of the system's boundary. The goal is to use these boundary measurements to figure out the unknown property inside the system.

This problem is notoriously difficult because it's "ill-posed." This means that even tiny errors in the measurements can lead to wildly inaccurate reconstructions of the unknown property. This is especially true for the Helmholtz equation (which describes time-harmonic waves), where stability is often only "conditional" or "Hölder-type," meaning the error in the reconstruction can grow much faster than the error in the data. The authors aim to improve this by proving a stronger "Lipschitz stability" for a specific, but important, class of potentials: those that are piecewise linear.

Methods

The paper employs a sophisticated mathematical approach that combines several key techniques:

  • Green's Function Analysis: A central tool is the construction and detailed analysis of a Green's function for the given differential operator. This function essentially describes the system's response to a point source and is crucial for relating internal properties to boundary data.
  • Asymptotic Estimates: The authors derive precise estimates for the behavior of the Green's function and its derivatives, especially near interfaces where the potential changes. This is vital for understanding how small changes in the potential propagate.
  • Unique Continuation: This principle states that if a solution to a differential equation is zero on a sufficiently large set, it must be zero everywhere. The paper uses this to control the behavior of solutions and relate errors in different parts of the domain.
  • Bootstrap Argument: A key innovation is a "bootstrap" argument that leverages the piecewise linear structure of the potential. This allows them to improve an initial, weaker stability estimate (like Hölder) to the desired Lipschitz stability. This involves iteratively refining estimates by using the structure of the problem and the derived inequalities.
  • Local Cauchy Data: Instead of using the more common "Dirichlet-to-Neumann map," the authors work with "local Cauchy data," which consists of both the value of the solution and its normal derivative on a portion of the boundary. This is more directly related to practical measurements.

Datasets

Not discussed in the paper. The paper is a theoretical mathematical work and does not involve experimental data or numerical simulations on specific datasets.

Key Insights and Ideas

  • Piecewise Linear Structure is Key: The paper demonstrates that assuming the potential is piecewise linear is sufficient to achieve Lipschitz stability, a significant improvement over previous results for more general potentials. This suggests that geometric assumptions about the unknown can greatly enhance the stability of inverse problems.
  • Bridging the Gap to Practicality: By using "local Cauchy data," the results are more directly applicable to real-world scenarios where measurements are often limited to specific parts of the boundary and involve both the field value and its gradient.
  • Iterative Refinement of Stability: The "bootstrap" argument is a powerful technique for upgrading stability estimates. It shows how to use the structure of the problem to iteratively improve the bounds on the error.
  • Generalization Beyond Positive Potentials: The analysis encompasses potentials that can be positive or negative, making it relevant for both wave propagation and quantum mechanical applications (Schrödinger equation).

Key Results

The main result is Theorem 2.2, which establishes a global Lipschitz stability estimate for determining a piecewise linear potential q from local Cauchy data. This means that if two potentials, q⁽¹⁾ and q⁽²⁾, are piecewise linear on a given partition of the domain Ω, and their corresponding Cauchy data spaces, C₁ and C₂, are close in a specific metric, then the L<sup>∞</sup> norm of the difference between the potentials is bounded by a constant times the distance between their Cauchy data spaces. This provides a quantitative guarantee on the accuracy of the reconstruction.

Limitations and Future Work

  • Piecewise Linear Assumption: The primary limitation is the strong a-priori assumption that the potential is piecewise linear. Extending these results to more general classes of potentials (e.g., piecewise constant, or even general measurable functions) remains an open challenge.
  • Fixed Frequency: The analysis is performed for a fixed frequency k. Extending the results to handle multiple frequencies or the full time-domain problem would be a significant advancement.
  • Dimension: The results are established for dimensions n ≥ 3. Extending them to lower dimensions (n=1, 2) might require different techniques.
  • Future Work: The authors suggest that their approach could be extended to other types of inverse problems and potentially to more general potentials by refining the unique continuation and asymptotic estimates. The iterative nature of the proof also hints at potential algorithmic approaches for reconstruction.

Summary updated

Abstract

We consider the inverse boundary value problem of determining the potential q in the equation Δ u + q u = 0 in Ω ⊂ R n , from local Cauchy data. A result of global Lipschitz stability is obtained in dimension n ⩾ 3 for potentials that are piecewise linear on a given partition of Ω. No sign, nor spectrum condition on q is assumed, hence our treatment encompasses the reduced wave equation Δ u + k 2 c − 2 u = 0 at fixed frequency k .
BibTeX
@inproceedings{alessandrini2018lipschitz,
  author = {Giovanni Alessandrini and Maarten V. de Hoop and Romina Gaburro and Eva Sincich},
  title = {{Lipschitz stability for a piecewise linear Schrödinger potential from local Cauchy data}},
  booktitle = {Asymptot. Anal},
  year = {2018},
  url = {https://openreview.net/forum?id=XuVwulneLQW},
}

Loading related content...

Figure