Alternative Title

Companion data for “On-shell compression and reconstruction (OSCAR) of monochromatic wave fields in weakly scattering media”

Abstract

Overview

This deposit contains one realization of each of the two simulation platforms of Ref. [1], a two-dimensional (2D) scalar field and a three-dimensional (3D) vector-field component, together with two self-contained Python scripts that perform On-Shell Compression and Reconstruction (OSCAR) on them. Each script Fourier-transforms the stored field, measures the shell radius from the spectrum, retains only the Fourier components inside the shell of full width α/ℓs around it for α = 1, 3, 5, 8, 12, reconstructs the field by the inverse transform, and plots the reconstructed intensities, panels (a)–(e), next to the original field, panel (f), on a shared color scale. Every reconstruction panel is annotated with the shell width α, the size of the compressed representation, the compression ratio C, and the measured overlap error ϵφ = 1 − |⟨E0,Er⟩|/(∥E0∥ ∥Er∥). The retained coefficient set is the complete compressed representation. No other information about the field is used in the reconstruction.

Viewing Instructions

Running the scripts

Requirements: Python 3 with numpy, h5py, matplotlib. The 3D script additionally uses scipy (its FFT preserves single precision) and needs about 7 GB of memory, so 8 GB free is recommended. Run each script in the directory containing the data files: python oscar_demo_2d.py python oscar_demo_3d.py Each prints, per α, the number of retained Fourier coefficients, the compressed size (at 8 bytes per complex coefficient), the compression ratio, and the overlap error ϵφ, and writes the panel-grid figure. The 2D metrics are evaluated over the scattering region. The 3D metrics are evaluated on the stored interior box with a boundary trim of 3ℓs/α at each end of z, which excludes the boundary layer of the periodic transform, cf. Sec. IIIB of Ref. [1]. The reconstructions use the bare shell projection. The optional power-restoring constant c2 = (1 − Pout)−1/2 of Ref. [1] rescales the overall amplitude only, which leaves the printed overlap error unchanged.


Department(s)

Physics

Comments

Acknowledgments and disclosure

Numerical simulations were carried out on the Mill high-performance computing cluster at Missouri University of Science and Technology [4]. A part of the computation was performed on the high-performance computing infrastructure operated by Research Support Solutions in the Division of IT at the University of Missouri, Columbia MO [5]. The authors declare the following competing interest: A patent application related to the OSCAR algorithm demonstrated by these scripts has been filed by Missouri University of Science and Technology.


Document Type

Data

Document Version

Citation

File Format

text

Language(s)

English

Publication Date

10 September 2026

Download

Additional files available below

Share

 
COinS