没食子酸（GAa）, 学名为3,4,5-三羟基苯甲酸（分子式为C7H6O5）, 通常以水合物的形式存在, 作为一种重要的有机原料, 广泛的存在于植物中。 有研究证明GAa具有抗氧化、 抗炎、 抗肿瘤、 抗病毒、 抗突变等多种作用。 因此GAa常作为抗氧化剂添加于化妆品中。 对羟基苯甲酸（p-HA）, 分子式为C7H6O3, 其中的R基为甲基、 乙基、 丙基、 丁基或庚烷基时分别称为对羟基苯甲酸乙酯、 对羟基苯甲酸丙酯、 对羟基苯甲酸丁酯和对羟基苯甲酸庚酯。 p-HA酯类的抗菌性强、 毒性低、 抑菌作用不受pH影响, 因此常添加于化妆品及药物中用作防腐剂。 间苯二酚（RE）又称1, 3苯二酚或间二苯酚（分子式为C6H6O2）。 RE具有杀菌作用, 可作为防腐剂添加于化妆品中。 以没食子酸（GAa）、 对羟基苯甲酸（p-HA）和间苯二酚（RE）三种化妆品常用添加剂为目标分析物, 通过引入第四维—溶剂, 构建四维荧光光谱数据, 使用甲醇（光谱级）、 乙醇（光谱级）、 超纯水分别获得三组实验样本, 三组样本的配置方法与加入药品量相同。 使用FS920稳态荧光光谱仪对样本进行检测, 设置激发波长为210～330 nm, 间隔4 nm记录一个数据； 发射波长为280～480 nm, 间隔2 nm记录一个数据。 初始发射波长总是滞后激发波长10 nm, 由此可消除一级瑞利散射的干扰。 随后使用空白扣除法对初始荧光数据进行预处理, 去除了溶剂的拉曼散射。 最后, 采用核一致诊断法确定待测样本的组分数为3, 使用交替加权残差约束四线性分解（alternating weighted residual constrained quadratic decomposition, AWRCQLD）算法对预处理后的三维荧光光谱数据进行分解。 结果表明, AWRCQLD算法分解得到GAa、 p-HA和RE的激发、 发射光谱图与目标光谱几乎重叠, 能实现光谱重叠严重的GAa、 p-HA和RE的快速定性和定量分析。

Gallic acid (GAa), known as 3, 4, 5-trihydroxybenzoic acid (C7H605), usually in the form of hydrates, as an important organic raw material, is widely found in plants. Studies have shown that GAa has many effects such as anti-oxidation, anti-inflammatory, anti-tumor, anti-viral and anti-mutation. Therefore, GAa is often added to cosmetics as an antioxidant. P-Hydroxybenzoic acid (p-HA), whose molecular formula is C7H603, wherein the R group is methyl, ethyl, propyl, butyl or heptyl, respectively, referred to as ethyl p-hydroxybenzoate, Propyl p-hydroxybenzoate, butyl p-hydroxybenzoate and heptyl p-hydroxybenzoate. The p-HA ester has strong antibacterial property, low toxicity, and antibacterial action against pH, so it is often used as a preservative in cosmetics and medicines. Resorcinol (RE) is also known as 1,3 benzenediol or m-diphenol (formula C6H602). RE has a bactericidal action and can be added to cosmetics as a preservative. In this paper, three kinds of cosmetic additives such as gallic acid (GAa), p-hydroxybenzoic acid (p-HA) and resorcinol (RE) were used as target analytes, and four-dimensional fluorescence spectra were constructed by introducing a fourth dimension solvent. Three sets of experimental samples were obtained for methanol (spectral level), ethanol (spectral level), and ultrapure water, and the configuration of the three groups was the same as that of the added drug. Using the FS920 steady-state fluorescence spectrometer (spectral wavelength response range of 200~900 nm, liquid nitrogen refrigeration range of 77~320 K, excitation source power is 450 W, signal-to-noise ratio is 6 000∶1) to test the sample, and setting the excitation wavelength to 210～330 nm, a piece of data was recorded at intervals of 4 nm; the emission wavelength was 280～480 nm, and a piece of data was recorded at intervals of 2 nm. The initial emission wavelength always lagged the excitation wavelength by 10 nm, thereby eliminating the interference of the first-order Rayleigh scattering. The initial fluorescence data were then pretreated using a blank subtraction method to remove Raman scattering of the solvent. Finally, the nuclear consistent diagnosis method was used to determine the number of components of the sample to be tested to be 3, and the three-dimensional fluorescence spectrum data after pretreatment were decomposed using the alternating weighted residual constrained quadratic decomposition (AWRCQLD) algorithm. The results showed that the AWRCQLD algorithm decomposes the excitation and emission spectra of GAa, p-HA and RE almost overlapping with the target spectrum, and can achieve rapid qualitative and quantitative analysis of GAa, p-HA and RE with severe spectral overlap. Using the AWRCQLD algorithm to decompose the samples, the average recoveries of GAa, p-HA and RE were 98.3%, 98.4% and 98.1%, respectively, and the root mean square error (RMSEP) was 0.081, 0.111 and 0.001 μg·mL-1. Three-dimensional fluorescence combined with the AWRCQLD algorithm enables rapid detection of GAa, p-HA and RE in cosmetics.