Study Of Two Chemometric Methods For Direct Determination Of Impurity Elements In High Purity Rich Line Matrix By ICP-AES

Inductively coupled plasma atomic emission spectrometry (ICP-AES) is one of the important means of determining trace metal elements in high purity materials at present. How to correct interference from the high purity matrix, particularly from the high purity rich line matrix elements, it has been a key technical problem in the applications field of ICP-AES. In this paper, the research objects were that ICP-AES directly determinate trace elements in high purity iron and lanthanum compounds. We researched and established two kinds of chemometric methods for correcting rich line matrix interference, namely partial least squares method (PLS) and support vector machine method (SVM). Detail works are as follows:1, In order to study the interference of trace elements in high purity iron matrix, it requires higher purity iron as matrix to preparation simulation sample, and establish interference correction models. In this paper, we researched the method that ultrapure iron oxide was prepared by using the analytical grade ferrous sulfate as raw materials, in which the content of Al, Cd, Co, Cr, Cu, Mg, Mn, Ni and Zn was below the method detection limit. By matching matrix of standard solution with prepared ultrapure iron oxide, the content of impurity elements in high purity iron GBW01402d standard substance were determined, the measured value is consistent with the certified values. Above on, the calibration method for determination of Al, Cd, Co, Cr, Cu, Mg, Mn, Ni and Zn in high purity iron and compounds were further studied by standard solution without matching iron matrix. The results showed that:Al, Co, Cr, Cu, Mg and Zn were interfered by iron matrix, and the interference was corrected to some extent by off peak background correction. The interference of Cd, Mn and Ni need use multivariate spectral fitting (MSF) to get some degree of correction. With these two instruments measured spectral interference correction method, Al, Cd, Cr and Cu have higher accuracy, and Co, Mg, Mn, Ni and Zn have less accuracy. Off peak background method and MSF method only correct the spectral interference. However, determination of impurity elements in the high purity rich line matrix, in which spectral interference and non-spectral interferences were serious, the correction effect of this two methods were not very satisfactory.2, In order to accurately determine the trace impurity elements such as Al, Cd, Co, Cr, Cu, Mg, Mn, Ni and Zn in high purity iron oxide by ICP-AES, while some methods could effectively correct spectral interference and non-spectral interferences. To get a more accurate analysis results, partial least squares (PLS) calibration method was studied, and established for this test system. The results showed that PLS method could effectively correct measurement error in high concentration of iron substrate. Compared with the MSF method and off peak background method of instrument, PLS method could withstand higher concentration of iron substrate. When the concentrations ratio of iron substrate and the impurity were50000:1, and the recoveries of this method were95%to105%, the correction capability to the interference and accuracy of analytical results were better than the off peak background method and MSF method of instrument.3, Sample analysis signal always was non linear by the effect of high purity rich line matrix, in view of SVM has its unique capability in dealing with the problems of nonlinear correlation. In this paper, we studied and established the method which SVM-ICP-AES method for determining the trace impurity elements of Al, Cd, Co, Cr, Cu, Mg, Mn, Ni and Zn. The results showed that: when the concentrations ratio of iron matrix and the impurity were50000:1, the recoveries of this method were90%to110%. The calibration results significantly better than the off peak background method and MSF method of instrument, but it slightly inferior to the PLS method. The reason is that:the Fe matrix affect for the impurity elements signal were mainly linear related. The results of this study also indicated that, once we determined the appropriate penalty parameter c and RBF kernel parameters g for SVM model, its evaluation MSE were minimum, R2closer to1, SVM was also an excellent interference correction methods.4, The establishment of PLS-ICP-AES and SVM-ICP-AES methods, in which were used for correcting ICP-AES direct determine trace impurities of Al, Ce, Cu, Er, Eu, Fe, Gd, Ho, Mg, Mn, Nd, Pr, Sm, Y, Yb and Zn in the purity lanthanum oxide. Compared with interference correct effect of PLS and SVM for La matrix, the results showed that:when the concentrations ratio of lanthanum matrix and the impurity were2500:1, it was interference seriously elements Fe and Zn by high purity La matrix, the spiked recovery of SVM-ICP-AES quantitative analysis were between90%and110%. However, the spiked recovery of PLS-ICP-AES quantitative analysis was between79.9%and123%. The spiked recovery of rest impurity elements were between90%and110%by this two method correction. For La matrix samples, RMSEP of SVM-ICP-AES quantitative analysis reached minimum, calibration model have the better prediction performance, this is because the non linear factors to matrix interference for impurity elements, La matrix more than Fe matrix, therefore, SVM significantly better than PLS in dealing with interference correct of such samples.5, Calibration set solution of PLS and SVM were designed by Standard addition subtraction method (SAS) in this paper, that is method which the spiked samples of spectral data subtract the not spiked samples of spectral data, it effectively avoided the problem of calibration set solution require high purity matrix. Fitting surface equation was established by fitting the off peak background method measured values under the absence of matrix matching, the iron matrix concentration values of tested liquid and the PLS (or SVM) calibration method measured values. By this fitting surface equation, it could be simple, fast and accurate determination of trace impurities in high purity materials with the absence of matrix matching; The high purity samples were analyzed by fitting surface equation, it just took the values of off peak background method under the absence of matrix matching and matrix concentration of tested liquid (Based on the volume of solution and the matrix of concentration values calculated) into this surface fitting equation we established, By this method, we could predict the unknown samples. The results showed that: the recoveries of surface fitting were less than the PLS and SVM correction method, but it was significantly better than the off peak background method and MSF method of instrument.