Deep-Learning-Assisted Resonant Ultrasound Spectroscopy for Cubic Solids

概要

The elastic constants of a material are related to a broad
range of physical properties because they are used in the
material’s equation of state [1]. They are also crucial for
assessing the reliability of theoretical calculations [2–4].
The accurate measurement of these constants remains an
issue in condensed-matter physics. For solids that can be
cut into regular shapes, such as a sphere, cylinder, or rectangular parallelepiped, resonant ultrasound spectroscopy
(RUS) is a powerful method for determining all independent elastic constants Cij [5–9]. This method involves
measuring a number of free-vibration resonance frequencies of the specimen and then, by an inverse method, such
as the nonlinear least-squares optimization method based
on the Levenberg-Marquardt algorithm, extract a set of Cij
that closely matches the measured resonance frequencies.
Although the RUS method is superior to other methods for accurately measuring the Cij of small specimens on a millimeter or submillimeter scale [10], it has
several drawbacks that have not been fully resolved.
Firstly, one must do a complex mathematical eigenvalue extraction. Secondly, the inverse method requires
an arduous code-development process, only accessible by
specialists. ...

この論文で使われている画像