高压粉末制样-X射线荧光光谱法测定海洋沉积物中28种组分

常规粉末压片制样是一种简单、高效的绿色环保制样技术,但是应用于某些沉积物样品制备存在样片表面粗糙和粉末容易脱落的问题。实验采用高于常规的压力进行样品制备,建立了X射线荧光光谱法(XRF)分析海洋沉积物样品中包括硫、氧化钠、氧化镁、三氧化二铝、二氧化硅、五氧化二磷、氧化钾、氧化钙、二氧化钛、氧化锰、三氧化二铁、钴、镍、铜、锌、钒、铬、镓、铌、锆、钇、锶、铷、铅、钡、镧、钕和铪在内的28种主、微量组分的方法。探讨了压力为300 kN和1 600 kN时的制样效果,并尝试引入BP神经网络模型利用其非线性拟合能力校正主量组分的基体效应。结果表明,采用1 600 kN压力制备的样片,表面致密、光滑、不龟裂和不掉粉,制样重复性和测试精密度也有较大提高。以55个有证标准物质中17种组分的数据集为训练样本,建立了海洋沉积物样品中主、微量组分的遗传算法-BP神经网络预测模型。按照实验方法对各组分含量相对较低的实际样品连续测试12次,计算得方法的检出限在 0.63～634 μg/g之间;精密度试验结果表明,各组分测定值的相对标准偏差(RSD, n=7)为0.16%～25.1%。方法用于海洋沉积物实际样品分析,其分析结果与国标法的测定结果吻合,能够满足海洋沉积物样品中多种组分准确分析的要求。

Determination of twenty-eight components in marine sediment by X-ray fluorescence spectrometry with high-pressure powder pelleting preparation

The conventional sample preparation method by pressed powder pellet is a simple, efficient, and environmentally friendly technique. However, there are some problems when it is used for the preparation of some sediment samples such as rough sample surface and easy peeling of powder. The pressure which was higher than the traditional value was adopted for sample preparation. A method for determination of 28 major and trace components (including sulfur, sodium oxide, magnesium oxide, aluminum oxide, silicon dioxide, phosphorus pentaoxide, potassium oxide, calcium oxide, titanium dioxide, manganese oxide, ferric oxide, cobalt, nickel, copper, zinc, vanadium, chromium, gallium, niobium, zirconium, yttrium, strontium, rubidium, lead, barium, lanthanum, neodymium, and hafnium) in marine sediment samples was established by X-ray fluorescence spectrometry (XRF). The effect of pressure at 300 kN and 1 600 kN on the sample preparation was discussed. The back propagation (BP) neural network model was introduced to correct the matrix effect of major components based on its nonlinear fitting ability. The results showed that when the sample was prepared at pressure of 1 600 kN, the surface was dense and smooth without cracks and powder dropping. Moreover, the sample preparation repeatability and the measurement precision were both significantly improved. The data set of 17 components in 55 certified reference materials was used as training sample to establish the genetic algorithm-BP neural network prediction model for the major and minor components in marine sediment. The actual sample with relatively low components was determined continuously for 12 times according to the experimental method. The detection limit of method was calculated between 0.63 μg/g and 634 μg/g. The results of precision tests indicated that the relative standard deviations (RSD, n=7) were between 0.16% and 25.1%. The proposed method was applied to the analysis of marine sediment actual samples, and the results were consistent with those obtained by the national standard methods, which could meet the requirements of simultaneous and accurate determination of multiple components in marine sediment samples.