ResearchPod Summary
Full-waveform inversion (FWI) is a cornerstone of seismic imaging and medical ultrasound, but its computational cost is dominated by repeated wave-equation solves. This paper investigates how to implement the necessary Born, adjoint, and Gauss-Newton actions—the building blocks of FWI—using quantum-assisted Schrodingerised propagation. Specifically, the authors address how to correctly define the physical pressure derivative when the wave equation is embedded into a higher-dimensional auxiliary space.
The researchers utilize the energy variables $\pi = c^{-1}\partial_t u$ and $q = \nabla u$ to construct an auxiliary-space Hamiltonian. A critical challenge in this framework is that physical pressure ($p = c\pi$) depends explicitly on the wavespeed $c$. The authors derive the directional derivative of this pressure observable, which consists of two parts: the propagated wavefield sensitivity and a direct receiver-calibration term. They incorporate both terms into the Born map, its adjoint, and the Gauss-Newton normal action. The study validates this construction through a nine-qubit compiled instance, using finite-difference checks, autodiff evaluations, and hybrid quantum-classical inversion runs.
The study proves that the receiver-calibration term is mathematically essential for the consistency of the Born map. Numerical experiments confirm that omitting this term leads to an $O(1)$ error in the Born approximation and fundamentally alters the regularized Gauss-Newton update direction. By including the calibration, the authors achieve second-order convergence in periodic refinement tests. Furthermore, they provide a resource model for state preparation and measurement, demonstrating that their calibrated pressure-observable formulation successfully reduces initial model error in finite-shot hybrid inversion scenarios.
This work provides a rigorous bridge between quantum Hamiltonian simulation and the practical requirements of FWI. By identifying the exact pressure-derivative structure needed for local updates, the authors enable the use of quantum algorithms for inverse scattering problems without sacrificing the physical accuracy of the gradient or Hessian actions. This is a necessary step toward deploying quantum-accelerated seismic or medical imaging workflows.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.