This is a preview of subscription content, log in via an institution to check access.
Author information
Authors and Affiliations
Consortia
Additional information
The Analytical Methods Committee has received and approved the following report from the Instrumental Criteria Sub-Committee.
Electronic Supplementary Material
Evaluation of overall performance
Evaluation of overall performance
Although the performance of each component of the instrument is evaluated individually, it is desirable to make some evaluation of the overall system performance. It is also appreciated that light-gathering power can be as easily tested by an evaluation of sensitivity as part of a test of overall performance.
The items for consideration can be summarised as: precision, sensitivity (detection limit, related to sensitivity and precision), accuracy (comparison of subsequent readings with a reference value), drift (calibration shift), freedom from spectral interferences (resolution), linear dynamic range and analytical range. Ideally, the instrument should be evaluated over its full wavelength working range and the following experiment is designed to permit this evaluation. However, in practice this may be too time consuming or impractical. It is recognised that only a limited number of wavelengths can be tested. If such a truncated experiment is envisaged, it is essential that it be applied equally to all instruments under evaluation.
Experimental
For each wavelength/element to be tested, prepare five standard solutions; the lowest (S1) should have a concentration corresponding to about one order of magnitude above the detection limit. The other four (S2–S5) should be prepared so that a total of five orders of magnitude are covered. The preparation of such a series of solutions is facilitated by the use of a suitable automatic diluter. The above solutions should be aspirated, in accordance with a randomised sequence, with a blank, S0, in between each standard and the sequence repeated to give a replication of six readings for each of S1–S5. The sample order should be randomised.
An example of a randomised sequence is (reading left to right and top to bottom):
Set up the data system to record the intensities for the 60 sets of readings. Each set of readings should comprise six individual measurements for which the mean and standard deviation are calculated. The data system should be set to calculate the corrected signal (x−b) and x/b (total signal-to-noise ratio). The blank value (b) should be the mean of the S0 mean values bracketing the test solution to compensate for baseline shifts. Ideally this experiment should be repeated over a number of days in order to assess reproducibility and not just repeatability.
Resolution should be checked at several points in the spectrum by recording the spectrum of a combination of elements with closely spaced lines. This test can only be carried out on instruments with a profiling facility. Suitable sets of lines include the following (in nm).
Element | nm | Element | nm |
|---|---|---|---|
Al(II) | 167.001 | Pb(I) | 283.306 |
P(I) | 177.499 | Sn(I) | 283.999 |
P(I) | 178.280 | Cr(II) | 284.325 |
As(I) | 189.042 | V(II) | 292.402 |
Sn(I) | 189.989 | Fe(I) | 302.064 |
Bi(II) | 190.241 | Al(I) | 309.278 |
Ba(II) | 193.400 | Al(II) | 309.284 |
As(I) | 193.696 | V (II) | 309.311 |
Hg(II) | 194.227 | Ti(II) | 310.623 |
Se(I) | 196.090 | V(II) | 311.071 |
Mo(II) | 202.030 | Be(II) | 313.042 |
Zn(II) | 202.548 | Be(II) | 313.107 |
Mo(II) | 202.800 | Hg(I) | 313.155 |
Se(I) | 203.985 | Hg(I) | 313.188 |
Mo(II) | 204.598 | Ca(II) | 315.887 |
Be(I) | 205.601 | Ca(II) | 317.933 |
Zn(II) | 206.200 | Ti(II) | 319.080 |
Sb(I) | 206.833 | Ti(I) | 319.200 |
Al(I) | 208.215 | Y(II) | 324.221 |
B(I) | 208.893 | Pd(II) | 324.270 |
B(I) | 208.959 | Cu(I) | 324.754 |
Cu(II) | 213.598 | Cu(I) | 327.396 |
Zn(II) | 213.856 | Ag(I) | 328.068 |
Cd(II) | 214.438 | Na(I) | 330.298 |
P(II) | 214.914 | Ti(I) | 334.500 |
Pb(I) | 216.999 | Ti(II) | 334.904 |
Sb(I) | 217.581 | Ti(II) | 334.941 |
Pb(II) | 220.353 | Ti(II) | 336.121 |
Ni(II) | 221.643 | Ti(I) | 338.289 |
Cu(II) | 224.700 | Pd(I) | 340.500 |
Ag(II) | 224.874 | Ni(I) | 341.476 |
Cd(II) | 226.502 | Ni(I) | 344.476 |
Co(II) | 228.616 | Pd(II) | 348.892 |
Cd(I) | 228.802 | Y(II) | 360.073 |
Sb(I) | 231.147 | Y(II) | 371.030 |
Ni(II) | 231.604 | Fe(II) | 371.670 |
Ni(I) | 232.003 | Fe(I) | 371.849 |
Al(I) | 237.312 | Fe(I) | 371.994 |
Al(II) | 237.335 | Fe(I) | 372.256 |
Fe(II) | 238.204 | Ce(II) | 393.081 |
Ca(II) | 238.892 | Ca(I) | 393.365 |
Fe(II) | 239.562 | Al(I) | 396.152 |
Sn(I) | 242.170 | Ca(II) | 396.847 |
Au(II) | 243.779 | Hg(I) | 404.656 |
Pd(II) | 248.892 | K(I) | 404.721 |
B(I) | 249.678 | Li(I) | 413.256 |
B(II) | 249.773 | Ce(II) | 413.380 |
Sb(I) | 252.852 | Ce(II) | 418.660 |
Hg(I) | 253.652 | Ca(I) | 422.673 |
Mn(II) | 257.373 | Hg(I) | 435.835 |
Al(I) | 257.510 | Ba(II) | 455.403 |
Mn(I) | 257.610 | Cs(I) | 455.531 |
Fe(II) | 259.940 | Cs(I) | 459.300 |
Pb(II) | 261.418 | Li(I) | 460.286 |
Ge(I) | 265.118 | Ti(II) | 552.430 |
Ge(I) | 266.158 | Na(I) | 588.995 |
Cr(II) | 267.716 | Na(I) | 589.592 |
Mn(I) | 279.482 | Li(II) | 610.362 |
Mg(II) | 279.553 | Li(I) | 670.784 |
Mg(II) | 280.270 | K(I) | 766.490 |
Mo(II) | 281.615 |
Treatment of results
The first set of results should be used to establish the calibration function (x−b) versus concentration. This will permit a check on the linear dynamic range of the instrument. Statistical examination of the residuals will give additional information on the efficiency of the curve-fitting programme. Subsequent sets of results should be compared with the initial set to provide information on instrument drift, which will affect accuracy, if the common practice of calibrating daily is envisaged. Non-superimposable plots will indicate drift. A typical example is shown below:
However, data should not only be presented or analysed in the form of log–log graphs, as quite large differences in signal show only as small shifts in the graphs. For example, the top standard intensity is useful in assessing drift. In the example below day 1 shows excellent freedom from drift whereas days 2 and 3 indicate significant problems over the 3-h total run time.
Individual points can be compared by calculating the standard deviation of the residuals of the replicates, while if desired the total plot of each set of data can be compared by means of a multi-tailed “F-test” (or analysis of covariance) using the residuals.
Short-term precision should be evaluated from the standard deviations of (x) for the six replicates comprising each data point. A graph of RSD of the corrected signal (x−b) versus log x/b will provide a plot from which the analytical range and detection limit can be estimated. The detection limit is the signal (x−b) which has a S/N ratio of about 3. The limit of quantitation is usually defined as a S/N ratio of about 10. However, the actual definitions are relatively unimportant, provided that they constantly applied for evaluation purposes. The analytical range R1−R2 is the range over which the function has values of less than, for instance, three times the minimum value, m. These values can be expressed in terms of log x/b. The lower value of LOD and m and the greater the value of R′−R, the better the performance of the instrument.
Rights and permissions
About this article
Cite this article
Analytical Methods Committee. Evaluation of analytical instrumentation. Part XVII. Instrumentation for inductively coupled plasma emission spectrometry. Accred Qual Assur 10, 155–159 (2005). https://doi.org/10.1007/s00769-004-0861-7
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00769-004-0861-7