A Universal Approach to Mie Scatter Correction in FTIR Analysis of Microsized Samples

Area of Science:

• Optics and Spectroscopy
• Machine Learning Applications
• Materials Science

Background:

• Mie scattering significantly distorts infrared absorbance spectra, complicating the analysis of microscopic samples.
• Accurate chemical information retrieval from infrared microspectroscopy is hindered by scattering effects.
• Existing solutions for the inverse Mie scattering problem (IMSP) are often sample-specific.

Purpose of the Study:

• To develop a universal deep-learning based method for correcting Mie scattering in infrared spectra.
• To enable accurate chemical characterization of diverse microscopic samples without pre-existing knowledge of their absorption properties.
• To provide a transferable framework for solving IMSP across various scientific disciplines.

Main Methods:

• A novel deep-learning algorithm was developed to address the inverse Mie scattering problem (IMSP).
• The method was validated using diverse real-world samples including microplastic beads, lung cells, and fungi.
• Spectroscopic measurements were performed using different setups, including single-detector and focal plane array detectors.

Main Results:

• The deep-learning approach demonstrated universal applicability to infrared spectra of various sample types.
• The novel method effectively corrected Mie-like scattering biases, restoring accurate spectral information.
• A characteristic spectral distortion near the true IMSP solution was identified and utilized for effective sieving.

Conclusions:

• The developed deep-learning method offers a universal solution for Mie scattering correction in infrared microspectroscopy.
• This approach allows rapid retrieval of undistorted infrared spectra, crucial for accurate chemical analysis.
• The framework provides a transferable solution for inverse Mie scattering problems in diverse scientific fields.