[Paper Review] On Inverting Asteroseismic Data

M. J. Thompson, J. Christensen‐Dalsgaard|arXiv (Cornell University)|Oct 19, 2001

Geophysics and Gravity Measurements3 references3 citations

TL;DR

This paper presents a linearized inversion method for asteroseismic data using least-squares fitting of basis functions to infer stellar internal structure, demonstrating that qualitative recovery of structural differences is possible even with noisy data. It emphasizes model calibration and optimally localized averaging (OLA) kernels to improve resolution and reduce bias from assumptions.

ABSTRACT

Some issues of inverting asteroseismic frequency data are discussed, including the use of model calibration and linearized inversion. An illustrative inversion of artificial data for solar-type stars, using least-squares fitting of a small set of basis functions, is presented. A few details of kernel construction are also given.

Motivation & Objective

• To develop a robust inversion framework for asteroseismic frequency data in solar-type stars, where data resolution is limited compared to helioseismology.
• To address the ill-posed nature of inverse problems in asteroseismology by incorporating physical constraints and error modeling.
• To evaluate the effectiveness of model calibration and basis function fitting in recovering structural differences between stellar models.
• To assess the impact of data noise and mode identification uncertainty on inversion accuracy.
• To demonstrate that physically motivated basis functions and OLA kernels improve resolution and reduce reliance on arbitrary assumptions.

Proposed method

• Apply linearized inversion using a small set of basis functions to represent structural differences relative to a reference model.
• Use least-squares fitting to minimize the chi-squared difference between observed and predicted mode frequencies.
• Construct optimally localized averaging (OLA) kernels to assess the resolution and localization of the inversion results.
• Perform artificial data experiments with controlled noise (Gaussian, σ = 0.1–0.3 μHz) to test inversion robustness.
• Use homology-scaled structure variables (u) to compare differences between stellar models (e.g., ZAMS stars, evolved models).
• Incorporate uncertainty in mode identification by adding a frequency-dependent correction term to the inversion equation.

Experimental results

Research questions

• RQ1 Can a least-squares inversion with basis functions recover the true structural differences between stellar models from noisy asteroseismic data?
• RQ2 How does data noise affect the accuracy and resolution of asteroseismic inversions?
• RQ3 To what extent do OLA kernels and model calibration improve the reliability and physical plausibility of inversion results?
• RQ4 How sensitive are inversion results to uncertainties in mode identification (e.g., radial order n)?
• RQ5 Can physically motivated basis functions enhance the recovery of structural features without introducing spurious artifacts?

Key findings

• The least-squares inversion with basis functions successfully recovers the qualitative structure of the true difference in homology-scaled u between a 1.05M☉ ZAMS star and a 1.0M☉ ZAMS star, even with noise.
• Inversions of noisy data (σ = 0.3 μHz) show increased scatter but still preserve the overall shape of the true difference, indicating robustness to moderate noise levels.
• For a more evolved star (model S₁ vs. solar model), the inversion recovers the core downturn in δu/u, though discrepancies suggest basis function limitations in evolved structures.
• The use of a constant correction term in the inversion equation effectively mitigates errors from radial order (n) misidentification.
• OLA kernels confirm that resolution is limited by the mode set, and that additional assumptions are needed to infer structure in regions with low sensitivity.
• Model calibration using large and small frequency separations can improve inversion accuracy, provided underlying physics (e.g., rotation, magnetic fields) are not neglected.