A light-induced thermoelastic spectroscopy sensor can measure methane and acetylene simultaneously in four seconds, with minimum detection limits below 400 ppb for both gases.
The system, developed at Harbin Institute of Technology, China, combines parallel heterodyne modulation with deep learning to extract the concentrations of both gases from a single composite signal. Its performance was demonstrated using controlled methane-acetylene mixtures in a study.
Light-induced thermoelastic spectroscopy, or LITES, measures the thermoelastic response produced when modulated laser light strikes a quartz tuning fork after passing through a gas sample. Because the tuning fork does not contact the sample directly, the method can be used with reactive or corrosive gases.
However, adapting LITES for rapid multi-gas analysis presents a challenge. Time-division approaches measure different gases sequentially, while spatial-division systems may require multiple optical channels, tuning forks, and electronic components. Frequency-division methods can also be difficult to combine with tuning forks operating at a fixed resonant frequency.
The new parallel heterodyne LITES sensor directs two lasers, targeting methane and acetylene absorption lines at 1.65 and 1.53 μm, through the same multipass cell. Both beams then strike a four-tine quartz tuning fork, producing an overlapping transient signal containing information from the two gases.
A deep learning model combining convolutional neural networks, an attention mechanism, and bidirectional long short-term memory separates the contributions and predicts the concentration of each gas.
Under a pure nitrogen background, the researchers calculated minimum detection limits of 378 ppb for methane and 285 ppb for acetylene.
The team also designed a four-tine quartz tuning fork that generated 4.34 times the signal amplitude of a standard commercial fork.
The current model requires retraining for new combinations of gases, and performance has so far been assessed under controlled laboratory conditions. The researchers plan to extend the approach to additional gas species and use cross-validation to test its robustness.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
