摘要: As a new energy source, hydrogen (H2) detection is a hot topic in recent years. Because of the weak absorption characteristic, laser spectroscopy-based H2 detection is challenging. In this paper, a highly sensitive H2 sensor based on light-induced thermoelastic spectroscopy (LITES) technique is demonstrated for the first time. A continuous-wave, distributed feedback diode laser with emission in the 2.1 μm region was adopted as the excitation source to target the strongest H2 absorption line of 4,712.90 cm−1. A Herriott multipass cell with an optical length of 10.1 m was chosen to further improve the H2 absorption. With the feature of processing the raw input data without data preprocessing and extracting the desired features automatically, the robust shallow neural network (SNN) fitting algorithm was brought in to denoise the sensor. For the LITES-based H2 sensor, the concentration response was tested, and an excellent linear response to H2 concentration levels was achieved. A minimum detection limit (MDL) of ~80 ppm was obtained. On the basis of implementation of the H2-LITES sensor, a heterodyne H2-LITES sensor was further constructed to realize a fast measurement of resonance frequency of quartz tuning fork and H2 concentration simultaneously. The resonance frequency can be retrieved in several hundred milliseconds with the measurement accuracy of ±0.2 Hz, and the result of 30,713.76Hz is exactly same as the experimentally determined value of 30,713.69 Hz. After the SNN algorithm was applied, an MDL of ~45 ppm was achieved for this heterodyne H2-LITES sensor.