Crosscorrelation kernel in the Green function retrieval and time reversal acoustics

TL;DR

This paper provides a rigorous mathematical analysis of crosscorrelation structures in Green function retrieval, revealing the importance of source-specific wavefield correlations and deriving analytical kernels for specific boundary geometries.

Contribution

It introduces a new framework for Green function retrieval using crosscorrelation kernels, extending beyond the traditional method by accounting for different sources and boundary geometries.

Findings

Derived analytical crosscorrelation kernels for plane and circular boundaries.

Showed that kernels reduce to delta functions for high-frequency waves.

Proposed numerical schemes for general boundary shapes.

Abstract

Crosscorrelation structures in the Green function retrieval by crosscorrelating wavefields are revealed using rigorous mathematical theory on integral equations. The previous practice on extracting the Green function by crosscorrelating the wavefields recorded at two locations produced by the same source and then summing such crosscorrelations over all sources on a boundary is inadequate and has limitations in recovering the low frequency content in the Green function. To recover the exact Green function, we need crosscorrelate the wavefields observed at the two locations generated by different sources, respectively. The crosscorrelation structure can be viewed as a matrix in the discrete case or an integral operator in the continuous case. The previous Green function retrieval method corresponds to the identity matrix multiplying a constant. If the matrix is diagonal, the wavefield crosscorrelation is still within the wavefield due to the same source but different weighting should be applied for different sources. We have derived analytically the crosscorrelation kernels for two important cases, the plane boundary and the circular boundary, and both kernels are symmetrical difference kernels, which correspond to convolutional operations. For high frequency waves or far-field sources, the kernels reduce to the know result which is the Dirac delta distribution (or an identity matrix). For other boundaries of general geometric shapes, numerical schemes like the boundary element method can be used to solve for the kernel matrix.