Optimizing data analysis for broadband mid-infrared absorption spectroscopy: A hybrid dataset approach

Area of Science:

• Spectroscopy
• Analytical Chemistry
• Laser Technology

Background:

• Broadband mid-infrared (mid-IR) spectroscopy is advancing for sensitive trace gas detection using novel laser sources.
• High-resolution spectra from these sources present challenges due to instrument-specific noise and baseline drifts.
• Efficient data processing is crucial to overcome these challenges and maximize the utility of complex spectral data.

Purpose of the Study:

• To develop a simulation-based approach for enhancing data processing in broadband mid-IR gas sensing.
• To improve the accuracy and reliability of gas compound detection by optimizing spectral analysis techniques.
• To enable a priori estimation of instrument- and application-specific detection limits.

Main Methods:

• Constructing a realistic simulation environment by combining simulated absorbance spectra with measured background intensity spectra.
• Incorporating instrument-specific effects like baseline drifts into the simulated data.
• Applying the workflow to real-life measurements in the 8-11 µm region for acetone detection in exhaled breath.

Main Results:

• The simulation approach significantly improved both classical least squares (CLS) and partial least squares (PLS) models.
• The method allows for fine-tuning of CLS fitting and provides realistic training sets for PLS models.
• Successful detection of trace acetone levels in complex breath samples was achieved, demonstrating the approach's effectiveness.

Conclusions:

• The presented workflow optimizes, trains, and assesses data-processing techniques for broadband mid-IR gas spectra.
• This approach is broadly applicable to various gas absorption spectroscopic instruments with broadband capabilities.
• It offers a robust method for improving and evaluating diverse data-processing techniques for enhanced gas sensing.