It makes intuitive sense — the higher the sensitivity of an inductively coupled plasma–mass spectrometry (ICP-MS) system,
the lower the detection limit. But there are many factors that affect the detection limit for a given isotope in a given sample.
These factors include sensitivity, background noise, and interferences.
In many applications in environmental management, semiconductor manufacture, and clinical research, some important elements
are subject to spectral interferences. For the accurate determination of these elements, some form of interference management
— collision reaction interface, collision cell, or reaction cell — often is employed.
With the recent focus on interference management techniques, less attention has been paid to one of the major benefits of
inductively coupled plasma–mass spectrometry (ICP-MS) — a benefit that has spurred its growth over the past two decades. We
are referring to sensitivity, the ratio of net signal to concentration. The importance of sensitivity should not be forgotten.
Many elements of interest in many samples are not subject to significant spectral interference, and for such elements, the
quantitation limit is set primarily by sensitivity.
Signal Noise, Sensitivity, and Detection Limits in ICP-MS
Detection limits for ICP-MS are quoted most often as "3 sigma" detection limits. These detection limits are derived using
the following equation:
detection limit = (3 × σbl)/sensitivity
where the standard deviation of the blank (σbl) is expressed in counts per second, and the sensitivity is expressed in counts per second (cps) per unit concentration (ng/L).
This equation clearly shows the direct dependence of detection limits on sensitivity. In practice, the standard deviation
of the blank also is affected by the sensitivity. This arises because part of the blank signal arises from traces of the element
of interest. We go to a lot of trouble to avoid such contamination, but its complete elimination is not possible.
The total variation in background noise (σbl) is a combination of both source flicker noise (caused by instabilities in the operation of the nebulizer, spraychamber,
and plasma, σsf) and Poisson or counting statistics noise (caused by the random nature of the arrival of ions at the detector, σcs) (2,3). Source flicker noise can be modeled as being a constant fraction of the total ion count rate, typically 0.5%. The
counting statistics noise can be modeled as the square root of the average number of counts during a measurement:
σcs = √(average number of counts)
Practically, there is always some elemental contamination present in the analytical solutions. Assuming a constant low blank
contamination, counting statistics (and hence, detection limits) on the blank count rate can be improved by increasing the
Table I shows the theoretical relationship between sensitivity, the two sources of background noise, and the resulting detection
limit. For this example, it is assumed that there is a continuous background of 10 cps, a real blank signal arising from contamination
of the blank with 1 ng/L of the element of interest, and that the integration time is 1 s. These results clearly show that
counting statistics noise becomes more dominant as a proportion of the total noise as the count rate drops, and that the best
detection limit is found when the sensitivity is greatest.
Table I: Theoretical calculations showing the effect of sensitivity on detection limits