Yuntao Peng, Zhirong Zhang, Pengshuai Sun, Hua Xia, Yongjun Cai, Haiming Wang, Bian Wu, Tao Pang
Accurate baseline determination remains a fundamental challenge in direct absorption spectroscopy, critically limiting its application for gases exhibiting broad-spectrum absorption, such as alkanes, where nonabsorbing regions are unavailable for traditional baseline correction. Herein, we report a physics-based dual-wavelength strategy that enables real-time baseline reconstruction under such demanding conditions, without relying on algorithmic assumptions or extensive training data. By exploiting the intrinsic, wavelength-correlated intensity variations within a multipass cell, we establish a robust linear model between a target wavelength (1686 nm, for propane and butane) and a reference wavelength (1653 nm, for methane). This model accurately reconstructs the unknown baseline of the broad-absorption band from the measurable baseline of the reference channel. Validated across dynamic temperature cycles (-10 to 30 °C), our method achieves a remarkably low relative root-mean-square error below 1.63% for baseline reconstruction. When applied to alkane mixtures, this approach introduces only 1.7% error in absorbance calculation. This work transforms a core limitation in direct absorption spectroscopy into a measurable physical correlation, paving the way for high-precision, real-time monitoring of complex gases in volatile industrial environments like petrochemical safety.