Chemometrics: Multiparameter, Multisignal Approaches

Last updated
Save as PDF

Page ID: 76383

Alexander Scheeline & Thomas M. Spudich
Analytical Sciences Digital Library

\( \newcommand{\vecs}[1]{\overset { \scriptstyle \rightharpoonup} {\mathbf{#1}} } \) \( \newcommand{\vecd}[1]{\overset{-\!-\!\rightharpoonup}{\vphantom{a}\smash {#1}}} \)\(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\) \(\newcommand{\id}{\mathrm{id}}\) \( \newcommand{\Span}{\mathrm{span}}\) \( \newcommand{\kernel}{\mathrm{null}\,}\) \( \newcommand{\range}{\mathrm{range}\,}\) \( \newcommand{\RealPart}{\mathrm{Re}}\) \( \newcommand{\ImaginaryPart}{\mathrm{Im}}\) \( \newcommand{\Argument}{\mathrm{Arg}}\) \( \newcommand{\norm}[1]{\| #1 \|}\) \( \newcommand{\inner}[2]{\langle #1, #2 \rangle}\) \( \newcommand{\Span}{\mathrm{span}}\)\(\newcommand{\AA}{\unicode[.8,0]{x212B}}\)

If one emission line is good, why wouldn't two be better? More photons means less shot noise, which means better precision. It also means more chance for interferences. What is the optimal number and identity of emission lines to use for elemental analysis? There is no global answer, but there is a set of statistical tools, loosely lumped under the heading "chemometrics," that provide rigorous means to use multiple portions of a spectrum to extract multi-element information. We model emission at each wavelength as I(λ) = I_background_,_λ + Σ I_n where we sum over all n different elements in the sample. We presume each element has a particular sensitivity at each wavelength, so we can rewrite the expression as I(λ) = I_background_,_λ + Σ k(λ, n) C_n. As long as we measure over a set of wavelengths greater in number than the number of elements, and the background is either independently known or modeled well as a function of the C_n's, regression fitting is possible. The weak links are:

All signals are noisy, so errors in measuring any elemental concentration are significant. Large errors in measuring the concentration of one element will propagate to errors in other elements.
The more complex the mixture, the harder it is to detect, much less quantify, concentrations of trace elements.
Any nonlinearity means that the expression for I(λ) is even more complicated than that shown.

One approach that has proven useful in some circumstances is the method of concentration normalization. The sum of the elements in a specimen must come to 100%. Prior to emission determination, the total mass of sample is measured and the rate at which sample is delivered to the plasma is specified. One uses ordinary working curves (no internal standard) for each element. The total amount of sample is then inferred from the sum of the concentrations determined. In almost all instances, the apparent concentration of all substances together will be different from 100%. One then scales all concentrations so that the sum of concentrations gives 100% of the known weight of the sample. Thus, all elements are used as standards for all others, since a suppression or enhancement for one element is presumed to have an identical influence on others. For samples where there is uniform suppression or enhancement of excitation, this approach works. If some elements are suppressed while others are enhanced, errors are actually amplified.

It has been noted in the inductively coupled plasma that a signal may be enhanced in one part of the plasma and suppressed elsewhere. At some intermediate height, dependent on many parameters, there is a point at which there is no net change in signal as a function of matrix composition. Hieftje and co-workers are in the process of developing detection schemes based on spatially-resolved, simultaneous multi-wavelength measurements as this module is being written. Check out the possibilities in the literature!