Quantum Meets Classical: A New Era for Full Waveform Inversion

Full waveform inversion (FWI) is a technique renowned for its precision in reconstructing heterogeneous material properties from data captured by receivers. However, its computational demands have long been a thorn in the side of researchers. The advent of physics-informed neural networks (PINNs) offered a promising mesh-free alternative, yet convergence struggles have plagued their performance, particularly when dealing with complex velocity fields.

Marrying Quantum with Classical

Enter the hybrid quantum-classical approach, a groundbreaking innovation poised to shift acoustic FWI. By integrating quantum computing with classical machine learning, this method harnesses the strengths of both worlds. The hybrid model employs a decomposed wavefield network and a global velocity network as classical-to-quantum pipelines, culminating in parameterized quantum circuits (PQCs). These circuits, realized through differentiable JAX statevector simulators, enable effortless automatic differentiation across both the classical PINN and the quantum circuit.

The result? A significant performance boost. On a geophysical anomaly benchmark, this quantum hybrid achieved a lower L1 velocity error than its classical counterpart while requiring about eight times fewer training iterations. Moreover, it accomplished this feat using roughly 33% fewer trainable parameters, outperforming all 15 classical hyperparameter variants tested. This efficiency and accuracy stand as testament to the potential of quantum computing in tackling computationally demanding tasks.

Beyond Geophysics

While the primary focus has been on geophysical applications, the implications of this technology extend far beyond. A second benchmark, involving a checkerboard pattern, demonstrated the hybrid's ability to recover structured spatial variations, emphasizing its versatility. This architecture is set to influence a range of wave-based inverse problems, from medical ultrasound tomography to non-destructive evaluation in engineering.

Why does this matter? Traditional FWI techniques have been resource-intensive, limiting their accessibility and application scope. By slashing training iterations and reducing the number of parameters, this hybrid approach democratizes the technology, making it feasible for broader use. Is this the dawn of a new era in computational problem-solving?

The Future is Quantum

The integration of quantum computing into traditional methods is more than just a technical feat. it's a statement about where the future of computational science is headed. Quantum computing's potential to revolutionize fields like FWI can't be overstated, offering a roadmap for solving previously intractable problems. As we stand on the cusp of this new frontier, one must ask: How soon will it be before quantum-classical hybrids become the norm rather than the exception?