SECTION I

Introduction

1 Overview

There is an increasing need for sensitive, accurate and convenient analytical techniques for analysis and determination of rare earth elements (REEs) with their expanding applications. REEs are now widely used in various areas of industry, agriculture, material science and other modern technologies, for instance, special alloy steels, non-ferrous alloys, magnetic materials, fluorescent powders, various kinds of growth promotion fertilizer in agriculture, catalysts in oil refining, additives in new ceramic materials, etc.

Being a powerful tool for elemental analysis, inductively coupled plasma atomic emission spectroscopy (ICP-AES) plays a more and more important role in the purity analysis of REEs owing to its high performance, such as: low detection limits (<0.0x µg ml-1), good precision, wide linear dynamic range, simple sample treatment, etc. As shown in many papers, ICP-AES has the potential to determine individual REEs directly without mutual separation, provided the spectrometer has adequate resolution and dispersion.

However, one of the basic problems of AES is spectral interference from the concomitant and matrix elements of the sample. In addition, REEs are well-known spectral line-rich elements. Improper analytical line selection may result in significant loss of detection power or accuracy for samples containing REE matrices. Therefore, it is crucial to choose appropriate analytical lines in order to avoid interference and ensure the quality of analysis. It is obvious that the line coincidence atlases are always the most emphasized and important fundamental research for ICP-AES, as stressed by many spectrochemists at a workshop held in Scarborough, Ont., Canada, in 1987: Needs for Fundamental Atomic Reference Data for Analytical Spectroscopy, Spectrochim. Acta, Part B, 1988,43, No. 1.

Although quite a few extensive tables have been published for REEs, few of them meet the demands for optimum selection of analytical line(s) for samples with REEs as major constituents. Some of them lack the mutual interferent information among the REE lines, and most of them do not provide spectral coincidence profiles. For those atlases with spectral profiles, the overlay information is still insufficient or not so accurate owing to the use of a low -resolution spectrometer for obtaining the spectral information. Many weak lines of REEs are missed or submerged in profiles of other sensitive lines. Furthermore, the general assessments adopted for rational line selection were sensitivity, detection limit, and signal-to-background ratio (SBR), etc. Another significant criterion for selecting analytical line(s), the true detection limit proposed by Boumans, has not been applied in most of these atlases.

The aim of this atlas is to provide spectral interference data for REEs, with emphasis on the spectral interferences occurring among REEs themselves. The atlas will cover the following aspects:

1. to record the detailed spectra of all REEs in the wavelength region adjacent to each of the prominent lines of a particular REE with a high-resolution ICP-AES instrument;

2. to provide reliable evaluation, based on recorded spectra, of the powers of detection of the chosen prominent lines for samples with the matrices of the other REEs;

3. to recommend the best analytical line(s) of an analyte with less interference and higher sensitivities for analysis with a particular REE matrix;

4. to provide data for calculating the criteria to choose the best line(s) with mixed REEs matrices.

2 Interpretation

Spectral Coincidence Profiles

Spectral coincidence profiles, according to our experience, are the simplest and clearest illustration for interpreting the interfering spectra. The graphic format not only provides us with the interference data, but also gives us a visual impact of the spectra in a certain wavelength range, from which we could obtain information on different kinds of interference, such as direct overlap, wing overlap and line broadening interference.

The coincidence profiles are presented in Part IV. These profiles for each prominent line are divided into four separate plots, each of them consisting of linear-scaled profiles of analyte spectral profile superimposed with interferent profiles La, Ce, Sm, Dy in plot A, Er, Ho, Tb, (Dy) in plot B, Eu, Gd, Lu, Tm in plot C, and Nd, Pr, Y, Yb in plot D, respectively. The magnitude of an interference can be estimated by a simple comparison of the analyte and interferent intensities at the analytical wavelength. The wavelengths of some unlabelled interferent lines in the plots may be found in the Tables of Interfering Lines in Part III. The full view of the spectra can be zoomed with programs on the disk enclosed with this book. This is especially useful for those profiles in which the analyte and interferent peaks are of significantly different magnitude.

Detection Limits

Detection limits (CL) in ICP-AES are usually defined as the concentrations that yield a net line signal equal to an arbitrary factor (K) times the relative standard deviation (σB) of the background signal (XB):

CL = K · σB · XB/SA (I-1)

where

SA = XA/CA (I-2)

SA: sensitivity of analyte;

XA the net analyte signal;

CA: the analyte concentration to produce XA.

Commonly, the relative standard deviation of the background σB is conservatively assumed to be 0.01. However, various values of K have been used by different workers. By convention, K is equal to 3, which corresponds approximately to 95% confidence. In the definition of the conventional detection limit given by Boumans in case of interference, K is equal to 2 [square root of (2)]. Here, in order to be consistent with other literature, a value of 3 is used for K. Therefore, the detection limit may be written as:

CL = 0.03 · XB/SA = 0.03 · CA/ XA/XB = 0.03 · CA/SBR (I-3)

Background Equivalent Concentration

Background equivalent concentration (BEC) is usually defined as the concentration that yields a net line signal equal to the net solvent blank signal, which may be estimated as:

BEC = XB/SA (I-4)

BEC is another empirical evaluation of detection power of the analysis.

True Detection Limit and Q Values

True detection limit (CL,true) assessment was proposed by Boumans and Vrakking for rational selection of wavelength lines. It has been proved capable to be a simple, applicable and standard criterion for estimating the type and magnitude of spectral interference in the analysis of real samples.

According to Ref. 11, for a multi-component system, in case of interference, CL,true is defined as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (I-5)

where λa,: the peak wavelength of an analysis line;

SI,j (λa): the sensitivity of component j at the interfering line at λa;

SA: sensitivity of analyte;

CI,j: the interferent concentration of component j;

CL,conv: the conventional detection limit.

The conventional detection limit can be written as:

CL,conv = 2 [square root of (2)] x 0.01 x RSD Bl x XBL/ SA (I-6)

in which, RSD Bl (%) is the relative standard deviation of the total background signal XBL, which is written as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (I-7)

where XB: the net background signal at λa;

XI,j: the net interferent signal of component j at λa;

XW,j: the net wing signal of component j at λa;

So the ratio XBL/SA may be written as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (I-8)

Boumans thinks that the quotients of SI,j and SW,j with SA are more meaningful than the values themselves:

QI,j(λa) = SI,j(λa)/SA (I-9)

QW,j(Δλa) = SW,j(Δλa)/SA (I-10)

The Q values are more convenient and have universal significance in that they are independent of (a) the transmission characteristics of the spectrometer and the response characteristic of the detector; (b) the transport efficiency of the nebulizer, at least to a first approximation; and (c) the units in which the sensitivities are expressed.

Then the conventional detection limit is obtained:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (I-11)

Finally, we have

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (I-12)

The basic parameters, XB, Xt, and XW, are illustrated in Figure I-1.

3 Apparatus and Procedures

Apparatus

A commercial sequential high-resolution ICP-AES spectrometer was employed for collection of the coincidence profiles. The specification of the instrument and the optimized operating conditions are summarized in Table I-1.

Sample Solutions

The stock solutions (1 mg ml-1) were provided by Shanghai Institute of Material Science (spectral purity > 99.995%). The standard solution and the artificial 'pure matrix solution of each rare earth element (listed in Table 1-2) were prepared by diluting the stock solutions with distilled and deionized water prepared by the MilliQ water purification system. For example, 1 ml Ce standard solution (1 mg ml-1) and 10 ml nitric acid (AR) were mixed and diluted to 200 ml to obtain the analytical standard solution. The final concentration of HNO3 in all test solutions was 0.8 mol l-1. 0.8 mol l-1 nitric acid prepared with distilled and deionized water was used as the blank solution.

Selection of Prominent Lines

Consulting other references, lines with large signal to background ratio (S/B) were selected for the present study. In this work, four to six prominent analytical lines were selected to investigate the spectral interference for each rare earth element. The detection limits of the selected prominent lines of each REE obtained by the authors are listed in Table III-1.

Experimental Procedures

As indicated in the section on Sample Solutions, the blank, standard and 'pure' matrix solutions were used to obtain the spectral data of background, analytes, and matrices respectively. The window width of the sequential spectrometer used in this study was set to 0.2 nm. Within the 0.2 nm window, every spectrum of background, analytes, and matrices was recorded in steps of approximate 0.9 pm. The integration time is 0.2 s per step with a constant voltage setting for the photomultiplier (PMT). Under these conditions, there are more than 200 measurement points per scan and 15 scans per profile for each of the 65 spectral lines. Thus, approximately 198 000 intensity measurements for the coincidence profiles have been made.

The coincidence profiles were obtained by superimposing the profiles of signals of background, analyte, and matrices that were located in the same wavelength range. In addition to the visualization of interference, the relevant criteria, CL, CL,conv, CL,true and Q values may be finally determined. A computer program compiled by the authors was used to locate positions of peaks and to calculate these critera automatically. Readers may contact the authors if they are interested in this program.

To maintain the excitation conditions consistent, the intensities of Dy II 364.540 nm and Sm II 359.260 nm were checked every working day with 5 µg ml-1 of standard solution under the optimum conditions. If the intensities of those lines monitored were out of the tolerance range, the carrier gas flow rate was adjusted to maintain the initial intensities. When investigation for one matrix spectrum was completed, the plasma torch and nebulizer were cleaned thoroughly with concentrated nitric acid (50%) to avoid memory effects.

Wavelength Accuracy and Reproducibility

Wavelength accuracy and reproducibility are decisive to the superimposibility of the coincidence profiles. The grating drive mechanism of the ICP 2070 is based on a harmonic drive. It works with a tolerance, reported by the instrument manufacturer, of less than one step. Consequently, the accuracy of the peak position located should be less than two steps, i.e. 1.8 pm. From the wavelength range of 200-500 nm, a prominent line of a known element was selected to check the shift of the grating position for about every 40 nm. For instance, Cu (1 µg ml-1) I1 213.598 nm and Lu (1 µg ml-1) II 307.760nm were used in the corresponding windows for such calibration purposes in this work.

The experimental results show that the shifts are actually less than [+ or -] 2 pm for five individual scans of the same line. Therefore the accuracy of the estimated wavelength for a peak in replicated scan should be less than [+ or -] 5 pm.