苯酚和麝香草酚等酚类化合物对人体和动植物有着严重危害, 且这些酚类化合物往往同时存在于水体。 由于苯酚和麝香草酚的激发和发射光谱重叠严重, 常规荧光方法不能实现直接快速测定。 基于三维荧光光谱结合四维平行因子（4-PARAFAC）算法, 对存在未知干扰物的湖水中苯酚和麝香草酚进行定性和定量分析。 利用三维平行因子和四维平行因子算法分解光谱数据, 探索三阶校正算法的“三阶优势”。 通过引入温度维来构建四维数据阵, 将不同温度下扫描得到的激发发射矩阵沿样本维叠加得到四维数据阵, 结合基于四维平行因子的三阶校正算法对目标分析物进行定性定量分析。 为避免溶剂散射和仪器的影响, 需要对扫描得到的激发发射矩阵信号进行预处理。 通过空白扣除法和Delaunay三角内插值法去除激发发射矩阵中散射信号, 再进一步进行激发发射校正, 得到真实光谱。 然后分别使用基于平行因子的二阶校正算法和基于四维平行因子的三阶校正算法对光谱数据进行分析, 对比两种算法的分析结果。 结果表明, 四维数据阵并不是三维激发发射矩阵简单的叠加, 得到的四维数据可能含有丰富的高维信息, 有助于改善对分析物的测量结果。 四维平行因子算法解析得到的湖水中苯酚和麝香草酚的平均回收率分别为97.7%±9.2%和96.5%±8.8%, 预测均方根误差为0.047和0.057 μg·mL-1, 预测相对误差低于10%, 分析结果优于三维平行因子（平均回收率分别为105.7%±15.3%和111.0%±3.6%, 预测均方根误差为0.090和0.056 μg·mL-1, 预测相对误差高于10%）。 实验表明, 样本中存在复杂干扰背景和数据共线性严重时, 三阶校正算法能够得到比二阶校正算法更满意的结果, 为复杂体系中苯酚和麝香草酚的检测提供了可靠方法。

Phenol, thymol and other phenolic compounds seriously harm the human body, animals and plants. They often exist in water at the same time. Because the excitation emission spectra of phenol and thymol are overlapped severely, and the conventional fluorescence method cannot achieve direct and rapid determination. Identification and quantification of two phenolic compounds (phenol and thymol) in lake water with unknown interferences using the three-dimensional fluorescence spectroscopy combined with four- dimensional parallel factor (4-PARAFAC) algorithm. The spectral data were analyzed and processed by the three-way parallel factor and the four-way parallel factor algorithm to explore the “high-order advantage” of the third-order correction algorithm. In this paper, the method of introducing an extra solvent mode to construct four-way data set was firstly used. The four-dimensional data array was obtained by superimposing the excitation emission matrixs obtained by scanning at different temperatures along the temperature dimension. Identification and quantification of two analytes using third-order calibration method based on four-way PARAFAC coupled with four-dimensional data. In order to avoid the influence of the instrument and scattering of solvent, firstly, the data obtained by scanning was preprocessed. The scattering signal in the excitation emission matrixs was removed by subtracting the blank sample and Delaunay trigonometric interpolation, and further excitation emission correction was performed to obtain true spectra. Then the data were analyzed using the second-order calibration algorithm based on the parallel factor and the third-order calibration algorithm based on the four-dimensional parallel factor, and the analysis results of two algorithms were compared. The results showed that the four-dimensional data array does not stimulate the simple superposition of emission matrices, and the four-dimensional data have rich high-dimensional information, which helps to improve the measurement results of the analytes. The average recoveries were obtained by the four-way parallel factor analysis algorithm, being the results: 97.7%±9.2% for phenol and 96.5%±8.8% for thymol. And the root-mean-square error of prediction values, for phenol and thymol, were 0.047 and 0.057 μg·mL-1, and the values of prediction relative error were less than 10%. Which were better than the results of three-way parallel factor analysis (105.7%±15.3% for phenol and 111.0%±3.6% for thymol, and the root-mean-square error of prediction values, for phenol and thymol, were 0.090 and 0.056 μg·mL-1, and the values of prediction relative error were larger than 10%). In short, the third-order calibration algorithm was superior to the second-order calibration algorithm in the sample with high background interference and severe collinearity. It provides a reliable method for the determination of phenol and thymol in the complex system.