GB2291184A - Inductively-coupled plasma atomic emission spectrometer - Google Patents

Abstract

A method for correcting drift in an Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-AES) during a sampling run includes the steps of: at intervals, conducting calibrations using a series of standard solutions containing known concentrations of elements being analysed; monitoring background emission lines from argon and from water components which are present in the plasma background; and producing a model which allows the drift on sample emission lines to be predicted from the drift on the background emission lines only. <IMAGE>

Description

INDUCTIVELY COUPLED PLASMA ATOMIC EMISSION SPECTROSCOPY The present invention relates to Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES), which is widely used for multi-element analysis of a wide range of sample matrices in, for example, environmental, biological and industrial applications.

In ICP-AES a high temperature (6-10,000K) plasma consisting of argon atoms, electrons and argon ions is sustained in a flowing stream of argon in a quartz tube. This is referred to as a plasma torch.

Energy is passed to the plasma by inductive coupling from a load coil surrounding the torch and carrying an electric current oscillating at radio frequency (usually either 27 or 40 MHz). The plasma appears as a very intense, brilliant white teardrop shaped discharge. A sample solution in water, usually in the form of a fine aerosol, is introduced into the plasma through a channel which is punched through the centre of the discharge by a separate argon flow. As the sample enters the plasma it is desolvated, vaporised and finally sample atoms and ions are formed. These atoms and ions, initially in a low energy, or "ground" state, interact with the plasma to form higher energy or "excited" state atoms which decay back to the ground state emitting energy in the form of light.The wavelength of the light produced is specific for a particular element and the intensity of the light is proportional to the concentration of the element in the sample solution. The emission is dispersed into its component wavelengths by an optical system, typically a diffraction grating, and the intensity of light at each wavelength is measured with an optical sensor.

Calibration is carried out using a series of standard solutions containing known concentrations of elements to be analysed. These are nebulised into the ICP and the light intensities at the specific wavelength characteristics of the analyte elements are measured and plotted, usually using a dedicated computer system, on calibration curves relating emission intensity to element concentration. The calibration curves are linear over 4 to 5 orders of magnitude.

In use, after calibration, samples having unknown concentrations are nebulised into the ICP and the resulting emission intensities compared with the calibration curves to give the concentration.

Whilst all commercial ICP instruments are designed to minimise any changes in operating conditions during use, it has been found that even small changes in these conditions can lead to changes in the relationship between element concentration and emission intensity.

Examples of changes in operating conditions are small variations in the sample introduction system, differences in sample matrix, temperature effects, changes in energy transfer into the ICP caused by variations in gas flow rates or thermal distortion of the load coil, fogging of optical components and thermal distortion of the optical components. This change in instrument readings is known as drift, and occurs over periods of minutes to hours (change occurring in the order of seconds is referred to as noise).

Demands for Quality Assurance data are becoming increasingly more stringent so drift is a problem in ICP operation. There are several methods for minimising drift. The simplest and most commonly used is to run calibrations at regular intervals. Unfortunately the nature of drift is erratic, and it cannot be assumed that interpolation between two calibrations will give correct results.

Calibrations must therefore be carried out at short intervals, which is clearly very time consuming.

Another possible calibration method is to use an internal standard where a known concentration of an element is added to the standards and samples. The drift of this element can be monitored throughout a sampling run and used to correct the drift in the other elements. Unfortunately it has been found that the drift in each wavelength differs according to the element and on whether the spectral line is the result of an atomic or an ionic transition.

Whilst the drifts on some wavelengths are highly correlated it is impossible to choose one internal standard line that corrects accurately for the wide range of elemental wavelengths that are commonly used in this multi-element technique.

An extension of the internal standard technique, known as the Generalised Internal Reference Method (GIRM), described in Analytical Chemistry, 1984, volume 56, page 43 onwards1 uses a number (usually 5) of internal standards, one of which is usually an argon background line. Using composite correction equations, solved by numerical analysis, this method has been shown to give an impressive improvement on the precision of analysis. Unfortunately there are some drawbacks with this method. Firstly there is the difficulty in practice of choosing five or so internal standards that are not present in the samples or will not interfere with analyte elements. Secondly there is the time and trouble, common to any method involving the addition of internal standards, of preparing and adding accurate amounts of the standard in such a way that contamination of the samples is avoided.

A method using standard additions of analyte to the sample over specified time periods, known as the General Standardised Addition Method (GSAM), is described in Analytical Chemistry, 1984, volume 54, page 560 onwards. In this, the emission lines used for analysis are also the internal standard lines. Here again the need for frequent very accurate additions to each sample is extremely time consuming, and also is not cost effective when large numbers of samples are to be analysed.

Methods involving combinations of GIRM and GSAM have also been tried.

On a particular ICP system it has been found that ICP drift appears to be dominated by changes in forward power and in sample uptake rate. A correction method using this fact, known as the Parameter Related Internal Standard Method (PRISM) is described in Analyst, 1985, volume 110, page 519 onwards. Two internal standard elements, each responsive to one only of these parameters, are chosen.

The response for each analyte element is related, by a simple mathematical expression, to the variation in each of the internal standards when its respective parameter is varied. During analysis the variation of the two internal standards is combined using the mathematical expressions previously calculated for each analytical line to give an estimate of the actual variation in the analytical line. Whilst this method requires initial experimentation to set up the correction equations relating the drift in the internal standards to the analyte line drift, these parameters are constant over long time periods and only need remeasuring every few months.The major drawbacks with this method are again the difficulty in finding two internal standard elements with the required drift characteristics which are not present in the samples and which do not interfere with the measurement of the analyte lines, and the time and expense of adding these elements to each sample.

A method using an artificial reference signal has recently been suggested in Journal of Analytical Atomic Spectrometry, 1993, volume 8, page 795 onwards. This has the practical drawback that in order to calculate the artificial signal the drift in all the other lines must be known.

There is therefore a need for a drift correction method which is reasonably easy to implement.

According to the present invention a method for correcting drift in an Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-AES) during a sampling run includes the steps of: at intervals, conducting calibrations using a series of standard solutions containing known concentrations of elements being analysed; continuously monitoring background emission lines from argon and from water components which are present in the plasma background; and producing a model which allows the drift on sample emission lines to be predicted from the drift on the background emission lines only.

The model may be based on the inter-relationship between drift in the various background emission lines. and may be produced by Principal Component Analysis (PCA) or by a related technique such as, for example. Partial Square Analysis (PSA). The model can consist of a number of abstract drift profiles which are related to the true underlying drift profiles. Linear combinations of these profiles can be used to model the drift of any sample emission lines.

Correction of drift in the sample emission lines can be effected by a statistical method such as, for example, Iterative Target Factor Analysis (ITFA) or PLS. Such methods allow prediction and correction of drift in sample emission lines using the measured drift in background emission lines in that same sample.

According to another aspect of the present invention an ICP-AES instrument includes a plasma generator and means for introducing a sample solution into the plasma, an optical system adapted to disperse light emission from the plasma into its component wavelengths, an optical sensor for measuring the intensity of light at each wavelength, and a computer having an intake from the optical sensor and being programmed: to compare results from calibrations, using a series of standard solutions containing known concentrations of elements being analysed and run at intervals during the sampling run, with the drift in continuously monitored background emission lines from argon and from water components which are present in the plasma background; and to use the resultant comparison in order to predict the drift in sample emission lines.

The comparison may be set up according to a model based on the inter-relationship between drift in the various background emission lines, and used to correct the sample emission lines for drift in periods between the calibrations.

An ICP-AES arrangement, and a series of experiments carried out to verify the invention, will now be described with reference to the accompanying diagrammatic drawings and graphs, in which: Figure 1 shows-an ICP-AES arrangement, Figures 2a, 2b and 2c show the comparison of drift corrections for a 39 point model with respectively 1, 2 and 3 factor corrections, Figures 3a, 3b and 3c show similar comparisons for a 10 point model, Figures 4a, 4b and 4c show similar comparisons for a 5 point model, and Figures 5a, 5b and 5c show a comparison of drift profiles before and after correction for lines representative of, respectively, Barium, Calcium and Lithium.

An ICP-AES arrangement 10 (Figure 1) has outer 11 and inner 12 concentric quartz tubes connected to delivery pipes 13, 14 adjacent closed ends 15, 16 respectively. Along the axis of the tubes 11 and 12 and projecting through the closed end 16 of the inner tube 12 is a sample pipe 17. The tubes 11 and 12 and the sample pipe 17 have open ends, respectively 18, 19 and 20. Surrounding the outer tube 11 in the vicinity of the open ends 18, 19, 20 is a load coil 21 connectable to a power supply 22.

In use argon is passed through the delivery pipes 13, 14 to the tubes 11, 12 and power, in the form of an electric current oscillating at radio frequency (for example 27 or 40MHz), is supplied to the load coil 21. Argon in the vicinity of the open ends 13, 14 of the tubes is heated by inductive coupling to a high temperature, typically (6-10,000K) to form a plasma consisting of argon atoms, electrons and argon ions. This is referred to as a plasma torch. The plasma appears as a very intense, brilliant white teardrop shaped discharge 23. A sample solution in water, usually in the form of a fine aerosol, is introduced into the plasma through a channel 24 which is punched through the centre of the discharge 23 by a separate argon flow through the sample pipe 17. As the sample enters the plasma it is desolvated, vaporised and finally sample atoms and ions are formed.

These atoms and ions, initially in a low energy, or "ground" state, interact with the plasma to form higher energy or "excited" state atoms which decay back to the ground state emitting energy in the form of light. The wavelength of the light produced is specific for a particular element and the intensity of the light is proportional to the concentration of the element in the sample solution. The emission is dispersed into its component wavelengths by an optical system 25, typically a diffraction grating, and the intensity of light at each wavelength is measured with an optical sensor 26, results being passed to a computer 27, where they are compared with calibrations and models to give the concentrations of elements emitting the wavelengths.

When used with the method of the present invention sample testing is carried out as normal. At intervals during a sample test programme calibrations are conducted using a series of standard solutions containing known concentrations of elements being analysed.

Also some of the background emission lines from argon and from water components which are present in the plasma background are monitored in samples and standards. A typical selection of five such emission lines are wavelengths 320.037, 404.442 and 515.139nm from argon, 306.36nm from the OH molecule, and 486.133nm from hydrogen.

The drift in the background emission lines is then compared with a model established during the analytical run in the computer according to the inter-relationship between drift in the various background emission lines. The comparison is then used to correct the sample emission lines for drift in periods between calibrations.

Principal Component Analysis (PCA) or a closely related technique such as Partial Least Squares Analysis (PLSA) is used to produce the model relating the drift in the background emission lines to the drift in the sample emission lines. The model consists of a number of abstract drift profiles which are related to the true underlying drift profiles. Linear combinations of these profiles are used to model the drift of any sample emission lines.

Correction of drift in the sample emission lines is effected by a statistical method such as, for example, Iterative Target Factor Analysis (ITFA) or PLS, which allows prediction and correction of drift in sample emission lines using the measured drift in background emission lines in the same sample.

It will be realised that this method, whilst similar to the correction method using calibration samples at regular intervals, has a number of distinct advantages. Drift is measured on each sample without interpolating, no internal standard elements have to be added to sample solutions, and the method will accurately correct for drift on a wide range of sample emission lines.

A drift correction simulation was carried out using a Perkin Elmer Plasma II sequential scanning ICP-AES with twin vacuum monochronometers, one with a 3600 line/mm grating and a wavelength range of 160-400nm, the other with a 1800 line/mm grating and a wavelength range of 160-800nm. The monochronometer gratings, plasma gas flows, plasma viewing height, a 50 position autosampler and a nebuliser peristaltic pump were all under computer control.The ICP operating conditions were: Plasma gas flow 15 1/mien Argon Auxiliary gas flow 1 1/mien Argon Nebuliser gas flow 1 1/mien Argon Forward power 1000 w Nebuliser pump feed lml/min The simulation used a tap water sample spiked with trace elements, in the following concentrations::

Element Wavelength (nm) Spiked Natural Concentration Concentration (-g/l) (mg/l) Aluminium 396.152 10 Barium 455.403 10 Boron 249.773 10 Cadmium 214.438 10 Calcium 317.93 12 -Chromium 205.552 10 Cobalt 238.892 10 Copper 324.754 lo Iron 259.94 10 - Lanthanum 379.478 lo Lead 220.353 10 Lithium 670.781 lo Magnesium 279.079 12 Manganese 257.61 10 Molybdenum 201.511 10 Nickel 231.604 10

Potassium 766.49 ~ 13 Scandium 424.683 lo Silicon 251.611 10 Sodium 330.237 - 60 Strontium 407.771 lo Sulphur 180.731 - 25 Vanadium 292.402 10 Zinc 206.200 10 Zirconium 339.198 lo The synthetic sample was analysed forty times with three replicates of every measurement on the ICP.The emission values for each line were scaled by subtracting the emission for the first sample from each emission value, thus centring the emissions on the value obtained from the first sample, and dividing this value by the standard deviation for all samples for a given emission line (that is, scaling the standard deviation for each line to 1). The data were then subjected to PCA using the "NIPALS" algorithm described in Chemometric and Intelligent Laboratory Systems, 1982, Volume 2, page 37 onwards.

The process of target factor iteration, described by Malinowski in Factor Analysis in Chemistry, John Wiley and Son, 1990, was used to test the ability of the background lines to predict the drift on the analyte lines. This involved setting up a test vector, for each of the forty sample analyses, that contained the actual drift for the background lines and estimated values for the analyte lines. In this case zero drift was used as the starting value. The vector was subjected to ITFA procedure which used the score values from the principal components calculated in the first step. During the iteration the drift values for the background lines were held constant and the analyte lines were allowed to converge to values that should model the original data.The iteration was performed using the scores from the first principal component, the first and second principal components and finally the first, second and third principal components. The three sets of reconstructed data were then used to correct for drift in the original data by subtracting the reconstructed results form the original results.

If the reconstructed data are a good model of the original data the drift for each data point should tend to zero. The effectiveness of the three combinations of principal components is illustrated in Figure 2. The average value for each emission line is plotted to give a figure of merit for drift (for no drift the average should be zero) along with the standard deviation to give a measure of precision (for the original data this has been scaled to 1). Clearly the one component reconstruction gives the best performance with greatly improved average drift values for emission lines with widely varying excitation potentials. In all cases the precision over the whole analytical run was virtually the same in reconstructed data as in the original data.It is thought that the reason why the long term drift is improved and the shorter term sample to sample variations was not improved is that the lines are measured sequentially so that there is a time delay between the measurement of the background line and the analyte line and therefore the measurement conditions for the two lines are slightly different. The background lines therefore monitor the long term trends but not the shorter term fluctuations occurring over time periods of the same order as that between line measurements.

It is likely that the use of an ICP-AES which measures all lines simultaneously would show an improvement in the precision of the reconstructed data over the actual data and could be used to correct for short term drift, i.e. noise.

The results in Figure 2 show the ability of the background lines to monitor the drift in the analyte lines. In practice the drift data for all the samples will not be known and therefore it is necessary to show that the method works when the PCA drift model is based on a few samples throughout the analysis. For real analysis these samples would be standards analysed at intervals during the run. To simulate this situation with this test data set PCA models were set up using five and ten points of the original 39. Using the same procedure for all the data, the five point and ten point PCA models were used to produce drift corrected data using the iterative target testing technique. The results are shown in Figures 3 and 4.The most striking feature of these results is that both the five and ten point models, although not as good as the 39 point model, show an improvement in drift for the majority of elements in the samples not included in the model.

The fact that the first principal component seems to be the most successful for each of the PCA models set up suggests that drift is being modelled only in this first principal component. This may not always be the case. However, if only one principal component is required to model drift then the target factor iteration is not required to calculate the drift. The drift for any element can be calculated very simply using just one of the background lines and the scores for each element in the first principal component.The drift for any of the analyte elements is given by: A a = Ba/B5 x As Where: Aa= The drift of element A in sample a As= The PC score for background line A from the PCA model B5= The drift in the background B line in sample a Bs= The PC score for background line B from the PCA model.

It has been reported (Journal Analytical Atomic Spectrometry 1992, volume 7 page 791 onwards and 1987 Issue 2 page 561 onwards) that, in the case of power variation, the change in line intensity is related to the energy sum (sum of ionisation and excitation energy Ecum) for ion lines, although this may not be the case for atom lines.

Although the cause of the drift in the simulated analytical run used here is unknown, Figure 5 shows three examples of scaled drift profiles from the simulated analytical run showing how the five point model was able to correct for both positive and negative drift from atom and ion lines with varied energy sums.

Claims (9)

1. Claims

   What is claimed is: 1. A method for correcting drift in an Inductively Coupled Plasma Atomic Emission Spectrometer (ICP-AES) during a sampling run including the steps of: at intervals, conducting calibrations using a series of standard solutions containing known concentrations of elements being analysed; monitoring background emission lines from argon and from water components which are present in the plasma background; and producing a model which allows the drift on sample emission lines to be predicted from the drift on the background emission lines only.
2. 2. A method as claimed in Claim 1 wherein the model is based on the inter-relationship between drift in the various background emission lines
3. 3. A method as claimed in Claim 1 or in Claim 2 wherein the model is produced by a technique such as Principal Component Analysis or Partial Least Squares Analysis.
4. 4. A method as claimed in any one of Claims 1 to 3 wherein the model conforms to a number of abstract drift profiles which are related to true underlying drift profiles.
5. 5. A method as claimed in any one of Claims 1 to 4 wherein correction of the sample emission lines is effected using a statistical method.
6. 6. A method as claimed in Claim 5 wherein the statistical method is Iterative Target Factor Analysis.
7. 7. An ICP-AES instrument including a plasma generator and means for introducing a sample solution into the plasma, an optical system adapted to disperse light emission from the plasma into its component wavelengths, an optical sensor for measuring the intensity of light at each wavelength, and a computer having an intake from the optical sensor and being programmed: to compare results from calibrations, using a series of standard solutions containing known concentrations of elements being analysed and run at intervals during the sampling run, with the drift in monitored background emission lines from argon and from water components which are present in the plasma background; and to use the resultant comparison to predict the drift on sample emission lines.
8. 8. A method as described herein with reference to Figures 1 to 5 of the accompanying drawings.
9. 9. An ICP-AES instrument substantially as herein described with reference to Figure 1 of the accompanying drawings.