Figure 2.7
A typical calibration curve.that, to avoid confusion, the analytical signal is plotted on the y-axis and the amount of analyte on the x-
axis.
Various examples of calibration curves will be found in later pages of this text. A good calibration
curve will be linear over a good range of analyte quantities. Ultimately curvature must be anticipated at
the higher ranges and uncertainty and deteriorating precision at the low ones. Ideally the calibration line
should pass through the origin but does not always do so. It is not invalidated if it does not but
nevertheless, it is prudent to consider carefully the possible reasons for this.
A particular issue that must be considered for all calibration procedures is the possibility of matrix
effects on the analyte signal. If such effects are present they may be allowed for in many cases by
matrix matching of the standard to the sample. This of course requires an accurate knowledge of the
sample matrix. Where this is not available, the method of standard addition is often effective. This
involves 'spiking' at least three equal aliquots of the sample with different amounts of the analyte, and
then measuring the response for both spiked and unspiked aliquots. A plot of response vs analyte,
extrapolated back, will give abscissae intercepts from which the amount of analyte in the sample may
be deduced (Figure 2.8).
Even this method, however, does not guarantee to overcome all matrix effects, and is often limited
where the matrix is extremely complex and remains largely unknown, as in naturally occurring
materials such as rocks and biological specimens. The use of recognized standards has proved
invaluable for the standardization and assessment of methods in these circumstances. Pioneered in
geochemistry, this involves the collection and thorough mixing of suitable standard materials. These are
then distributed for analysis by recognized laboratories using as many different techniques