腔衰荡光谱技术(CRDS)作为一种具有高灵敏度高光谱分辨率的检测方法已被广泛用于痕量气体检测。 而目前基于CRDS痕量气体检测多针对单一气体进行测量或通过多个激光器产生的多光束进行多种组分气体浓度测量。 利用DFB激光器波长可调谐特性, 通过强弱吸收峰结合, 使用单光束实现了多种温室气体的腔衰荡光谱技术同步检测。 由于大气中水汽和二氧化碳浓度较高, 为实现同一衰荡系统对三种温室气体的同步测量, 在平衡吸收损耗的基础上, 选取1 653~1 654 nm内甲烷的强吸收峰与水汽、 二氧化碳的弱吸收峰进行测量。 通过光谱叠加反演矩阵, 分别得到甲烷、 水汽、 二氧化碳的浓度。 在计算测量灵敏度过程中发现, 通过去除衰荡过程初期的部分数据点(过滤区间), 会对噪声等效吸收系数产生影响。 多数情况下, 在测量灵敏度计算方面, 列文伯格-马夸尔特算法(L-M)会优于离散傅里叶变换法(DFT); 但当衰荡曲线的单指数性下降时, 上述结论不一定成立。 搭建了一个低精细度(F≈6×103)衰荡腔对上述结论进行了实验验证。 相较于用于测量温室气体浓度的高精细度衰荡腔(F≈1×105), 低精细度衰荡腔的衰荡速率较快, 衰荡曲线的单指数性明显低于高精细度衰荡腔。 实验表明, 在过滤区间长度较短时, 采用DFT算法计算得到的噪声等效吸收系数会小于L-M算法得到的结果。 当过滤区间长度增加时, L-M算法得到的结果优于DFT算法。 在受过滤区间长度影响方面, DFT算法的波动性要明显小于L-M算法。 根据Allan方差分析, 在512次采样平均(约8 s)下的最小噪声等效吸收系数进行计算, 该CRDS装置测量灵敏度为2.4×10-10 cm-1。 在25 ℃标准大气压下, 对应甲烷、 水汽、 二氧化碳的测量灵敏度分别为0.64 ppbv, 3.5 ppmv和4.0 ppmv。 基于该CRDS装置, 通过单光束多波长测量方法, 利用光谱叠加反演矩阵, 测得大气中甲烷、 水汽、 二氧化碳浓度分别为2.018, 3 654和526 ppmv; 而采用经典CRDS单波长测量得到的甲烷、 水汽、 二氧化碳浓度分别为2.037, 3 898和630 ppmv。 通过与温控调节波长, 逐点扫描得到的光谱吸收曲线进行对比, 采用多波长测量得到气体浓度进行复合拟合的光谱曲线残差小于单波长测量得到气体浓度进行简单拟合的光谱曲线残差。

The cavity ring down spectroscopy (CRDS) has been a proper detecting method for the trace gas with its ultrahigh sensitivity and super spectral resolution. However, the common CRDS is designed for a single gas or the measurement of multiple species with several laser sources. In this paper, a CRDS instrument has been developed for multicomponent greenhouse gas synchronous detection with a single laser. Considering the balance of the absorption losses, it utilizes the strong absorption peak of methane (CH4) and weak ones of vapor (H2O) and carbon dioxide (CO2) in the range of 1 653～1 654 nm simultaneously. The wavelength scan of that range is completed by a tunable distribution feedback laser diode. The corrected concentration of greenhouse gas has been determined by the CRDS instrument with a high finesse (F≈1×105) cavity and calculated with the spectral superposition inversion matrix. It demonstrates that the remove of data points, called the filter region, at the preliminary stage of the decay has an influence on the noise equivalent absorption coefficient, which has a interrelation with the measuring sensitivity. In most cases, the Levenberg-Marquardt (L-M) algorithm, which shows a good precision, is better than the discrete Fourier transform (DFT) algorithm on the measuring sensitivity as a fitting algorithm. But this conclusion will be dubious when the ring-down curve is deviated from the single exponential form. For studying this phenomenon, a CRDS instrument with a low finesse cavity (F≈6×103) is set up. Compared to the high finesse cavity, the low finesse cavity has a faster decay rate, and a bigger deviation from the single exponential form, which can be easily seen from the residual analyses. When the filter region is not long enough, the noise equivalent absorption coefficients calculated by L-M algorithm is larger than the ones calculated by DFT algorithm. Meanwhile, according to the definition of the fluctuation of the noise equivalent absorption coefficient, the influence of the DFT algorithm is less than that of the L-M algorithm affected by the length of the filter region in both high finesse cavity and low finesse cavity. The best length of the filter region in our CRDS instrument is 20 data points, which are basically the same at different averaging time. And according to Allan variance, the measuring sensitivity of the CRDS instrument can reach 2.4×10-10 cm-1 for an 8 s integration time. At 25 ℃ and 1 atm, the measuring sensitivities of CH4, H2O and CO2 are approximately 0.64 ppbv, 3.5 ppmv and 4.0 ppmv separately. Calculated with the spectral superposition inversion matrix, the atmospheric concentrations of CH4, H2O and CO2 in the lab are measured to 2.018 ppmv, 3 654 ppmv, and 526 ppmv separately with multiple wavelengths, in contrast to the results of 2.037 ppmv, 3 898 ppmv and 630 ppmv in the classical CRDS method. Using the temperature control of the DFB laser, an absorption spectrum of the greenhouse gas has been acquired with the wavelength scan. Compared to this measured results, the residuals of the complex fitting curve using the data from the multiple wavelength measurements are less than the ones of the simple fitting curve with data from the classical method.