Escolar Documentos
Profissional Documentos
Cultura Documentos
DOI 10.1007/s00769-005-0908-4
Yoshihiro Hirano
Kyoko Imai
Kazuo Yasuda
PRACTITIONERS REPORT
Introduction
Atomic absorption spectrometry (AAS) has been widely
used for determination of trace metals in various areas,
such as environmental, biological, clinical and material
analyses. The system of the atomic absorption spectrometer is relatively simple. However, there has been no
paper on the estimation of the uncertainty for the instrument.
Several works on the estimation of uncertainty in AAS
have been reported. Kurfrst et al. [1] studied the uncertainty in the analysis of solid materials by graphite
furnace AAS comparing sample digestion AAS and solid
sampling AAS. Stringari et al. [2] evaluated sample
grinding methods for analysis of plant samples by
graphite furnace AAS. Bettencourt da Silva et al. [3]
applied their uncertainty validation method to the analysis
of leaves by acid digestiongraphite furnace AAS. Synek
et al. [4] studied the uncertainties of mercury determination by cold vapor AAS and reported that the contribution of the sample absorbance component was highest.
191
Uncertainty sources
We assessed three main uncertainty sources for the
measurand of absorbance: drift of signal, noise, and
nonlinearity of the calibration curve. Each source is
considered to be influenced by several factors. The relationship is summarized in Fig. 2.
192
Fig. 5 Drift of Cd sensitivity just after turning on the lamp (0.5 mg/
l Cd)
Fig. 4 Drift of baseline of Cd absorbance after the lamp has
warmed up: A lamp housing closed, B lamp housing open, C: airblown. a Single-beam measurement; b double-beam measurement
Table 1 Uncertainty (absorbance units) of the drift of the lamp
intensity just after turning it on
Optics
020 min
2040 min
4060 min
Single beam
Double beam
0.07028
0.00042
0.00189
0.00018
0.00564
0.00011
2040 min
4060 min
0.00382
0.00036
0.00016
193
0.00005
0.00007
0.00013
0.00028
0.00003
0.00009
0.00011
0.00010
Log transformation
The signal output from the PMT is transformed to absorbance in an electric circuit. The error in the transformation process may affect the uncertainty of absorbance.
In the log transformation process in the instruments we
used, light energy, which correspond to 100% transmittance, is adjusted to 3,000,000 counts of an analog-todigital converter output. Therefore, the electrical 1
count error is calculated as log(2,999,999/3,000,000)=
0.00000014 aborbance units. This means that the uncertainty for the log transformation process was negligible.
Nonlinearity of calibration curve
Line profile
In AAS, the line profile of the HC lamp and the absorption line profile are not exactly the same for some reasons
[6]. Doppler broadening is caused by thermal motion of
atoms. Because the temperatures of the lamp and the
flame are very different, the effects are different for
194
emission and absorption line profiles. Pressure broadening is caused by collisions of atoms of different species or
of the same species. The pressure inside the HC lamp is
different from the pressure in the flame. Therefore, the
effects are different for emission and absorption line
profiles. Also, this effect acts to shift the profile to longer
wavelength. As a result, the relationship between the
emission profile and the absorption profile becomes like
that in Fig. 8 [7]. This relationship causes the calibration
curve bend. Moreover, self-absorption of the emission
line is caused by atomic absorption inside the HC lamp.
This phenomenon also acts to bend the calibration curve.
These are well-known phenomena in AAS.
However, the calibration curve bending caused by the
difference of the line profiles occurred at over about 0.3
absorbance units. Therefore, when the measurement is
performed within 0.3 absorbance units, the uncertainty of
the calibration curve nonlinearity caused by these phenomena will be negligible.
Stray light
Theoretical calibration curve curvatures caused by stray
light are shown in Fig. 9. The stray light has been discussed in for conventional spectrophotometry. When
there is stray light, the curvature becomes significant with
increasing stray light intensity. However, this effect is
prominent only in the region of absorbance larger than
1.0. Atomic absorption analysis is usually done in a region below 1.0 absorbance units, and it may be done
below 0.1 absorbance units. In addition, stray light for
recent spectrophotometer models is less than 0.05%.
195
Fig. 11 Calibration curves for Cd and Mo with different fuel flow rates. a Cd: a fuel 2.0 l/min; b fuel 3.6 l/min. b Mo: a fuel 3.4 l/min; b
fuel 3.7 l/min; c 4.0 l/min
phenomenon exists in the horizontal and vertical directions. This fact may cause nonlinearity of the calibration
curve. The contour line also changes when the flame
condition is changed; therefore, the uncertainty of the gas
flow rate will affect the uncertainty of absorbance.
Figure 11 shows calibration curves for Cd and Mo with
different fuel flow rates. In both cases, the calibration
curves bend in the high absorbance region. In the fuel-rich
condition, the sensitivity of Cd decreased and the calibration curve bending became prominent. The bending of
the Mo calibration curve was more prominent with a
small change of the fuel flow rate.
For the data shown in Fig. 11, the calculation of the
uncertainty at 0.5 mg/l Cd was performed as follows:
1. The absorbance of 0.5 mg/l Cd is 0.0782 at a fuel flow
rate of 2.0 l/min, and 0.0432 at 3.6 l/min. Then, the
absorbance change is 0.0219 for 1 l/min fuel difference.
2. The specification of the gas flow controller is 5%.
Then, an optimum flow rate of 2.0 l/min will be
controlled within 0.l l/min. Assuming a rectangular distribution,
the uncertainty is estimated to be
p
0:1 3 0:0577l=min.
3. The corresponding uncertainty of absorbance will be
0.00126.
The calculation of uncertainty for Mo was done in the
same manner, resulting in 0.0222 absorbance units at
100 mg/l Mo.
Wavelength selection
Figure 12 shows the emission line spectrum of HC lamps
for six elements including typical complicated profiles of
Co, Fe and Ni. These line spectra include the instrumental
function (slit width); therefore, they were not substantially different from the line profile shown in Fig. 8. Most
of the line profiles for element not shown in Fig. 12 have
simple profiles. In early models in AAS history, the
wavelength had to be adjusted manually, and that could
seriously affect the uncertainty of the measurement.
However, in the latest instruments, the computer automatically searches the wavelength and sets the wavelength at the highest light intensity. Moreover, with the
development of high-performance HC lamps, the intensity
of the main resonance lines became higher than that of
neighboring lines. Therefore, the uncertainty caused by
wavelength selection is considered to be negligible.
Combined uncertainty
All the deviations caused by the previously mentioned
factors are considered to be incorporated in the measurand
(absorbance) by the linear combination. Therefore, the
combined uncertainty should be calculated from the
square root for the sum of square of each uncertainty. The
calculation is summarized in Table 5. The combined
uncertainty was estimated as 0.00133 at a Cd concentration of 0.5 mg/l (0.078 absorbance units). It was found
196
Fig. 12 Emission line profiles of hollow cathode lamps: a Ca; b Cd; c Co; d Fe; e Mg; f Ni
Table 5 Estimation of combined uncertainty for Cd measurement
Uncertainty source
Uncertainty
Negligible
Negligible
0.00126 absorbance units
0 00133 absorbance units
(0.5 mg/l Cd , 0.078 absorbance units)
References
1. Kurfrst U, Rehnert A, Muntau H (1996)
Spectrochim Acta B 51:229
2. Stringari G, Pancheri I, Mller F, Failla
O (1998) Accred Qual Assur 3:122
3. Bettencourt da Silva R, Cames M,
Seabra e Barros J (1998) Accred Qual
Assur 3:155