(24) R. M. Lowe, Spectrochim. Acta, in press. (25) P. Hannaford, unpublished work. (26)D. G. Mitchell and A. Johansson, Spectrochim. Acta, Part 6,25, 175 (1970). (27) R. M. Dagnall, G. F. Kirkbright,T. S. West, and R. Wood, Ana/yst(London), 97, 245 (1972).
(28) T. F. Fisher and C. E. Weber, J. Appl. Phys., 23, 181 (1952). (29) J. E. Greene and F. Segueda-Osorio, J. Vac. Sci. Techno/., 10, 1144 (1973).
for review
19, 1976* Accepted
19,
1976.
Noise at Detection Limit Levels in Atomic Absorption Flame Spectrometry Peter
I?.
Liddell
Varian Techtron my. Ltd., P.O. Box 222, Springvale 3 171, Victoria, Australia
The standard deviation in the absorbance at zero concentration is experimentally determined for ten elements in an airacetylene flame. The contribution of various noise sources to the total standard deviation is determined and the domlnant source at the detection limit is found to be photon noise, lamp flicker noise, or flame transmission noise.
Amplifier-Readout Noise. UAR. This is the standard deviation in the absorbance due to the electronic processing of the signal and any errors in the readout device. If these noise sources are assumed to be independent, their standard deviations add quadratically and the standard deviation in the measured absorbance of the blank solution is given by UA
There are a number of published works giving detailed theoretical analyses of noise in atomic absorption measurements, both over a broad absorbance range (1-6) and over the narrow absorbance range near the detection limit (7-11). Some experimental data have been published on the measurement of noise in the optimum working range (1-3, 6, 12-15). It is generally agreed (1,3,6,14)that most of the noise in this range is due to fluctuations in the absorption properties of the analyte, although analyte emission noise can also be important for some elements. Both of these noise sources depend on the concentration of the analyte and can be assumed to be negligible at the low concentrations corresponding to the detection limits. Factors which may affect the standard deviation in the absorbance, U A , at the detection limit are (5, 6 ) the following. Photon Noise. This consists of statistical fluctuations in the photocathode current, i, generated by light from the lamp and is sometimes called signal shot noise. The standard deviation in the absorbance due to photon noise, up, depends on i according to the relationship (5),
Lamp Flicker Noise, UL. This is the standard deviation in the absorbance due to fluctuations in the output intensity of the light source. Flame Transmission Noise, UF. This is the standard deviation in the absorbance due to fluctuations in the transmission of light through the flame. It includes the effect of aspirating a blank solution. Background Emission Noise, UBE. This is the standard deviation in the absorbance due to emission from the flame and the blank solution. It includes flicker noise caused by fluctuations in the emission intensity and photon noise caused by the increase in the intensity of light at the photomultiplier. However, if a modulated light source is used, any fluctuations which occur at a frequency less than the modulation frequency should be corrected for. Dark Current Noise, UD. This is the standard deviation in the absorbance caused by emission of electrons from the photocathode without the arrival of a photon.
= (Up2
+
UL2
+
UF2
+ UBE2 + UD2 + .AR2)1’2
(2)
In this paper, the magnitude of these noise sources is experimentally determined for a number of elements under standard operating conditions.
EXPERIMENTAL Apparatus. A Varian Techtron atomic absorption spectrophotometer, Model AA-GD, was used for all measurements. This instrument employs single beam optics with the light source modulated a t 285 Hz with a 50%duty cycle. In every case, the light source used was a Varian Techtron hollow cathode lamp. The instrumental conditions are listed in Table I. For the absorbance measurements, the digital readout on the instrument was used with fifty times scale expansion. For the %T measurements, a National chart recorder, Model VP653A, with a resolution of 1 WVor 0.001%Twas used. Procedure. For each noise measurement, a series of 101 consecutive 10-s integrations was obtained. This series was handled in the same way as an actual analysis consisting of alternating blank and sample measurements. Every second reading was treated as a sample measurement and the average of the preceding and following readings (treated as blanks) subtracted from it. This yielded 50 readings corrected for any long-term drift in exactly the same way as in an actual analysis. The standard deviation was then calculated from these readings. The following measurements were made for each element using this procedure. 1)The standard deviation in the absorbance was determined with the lamp on and the flame off, giving (up2 u~~ The photon noise here will be less than that in Equation 2 for those elements where the flame or the blank produce significant absorbance. 2) The standard deviation in the absorbance was determined with the lamp on, the flame on, and distilled water aspirating, giving UA. Prior to this step, the burner position and flame stoichiometry were adjusted for maximum absorbance using a standard aqueous solution of the element under investigation. 3) The standard deviation in the %T mode was determined with the lamp off, the flame on, and distilled water aspirating. The monochromator settings and photomultiplier supply voltage were the same as in step 2. The effect of this O%Tnoise on the noise in the absorbance mode at zero absorbance (100%T)was determined from the relationship ( 5 ) , u(%T) u (abs) = 0.4343 (3) 100 where dabs) and u(%T) are the standard deviations in absorbance
+ + +
ANALYTICAL CHEMISTRY, VOL. 48, NO. 13, NOVEMBER 1976
1931
Table I. Instrumental Operating Conditionsa Wavelength, Lamp current, Spectral band mA Element nm pass, nm 7 1.0 As 193.7 Se 196.0 10 1.0 Zn 5 0.5 213.9 Pb 5 1.0 217.0 0.2 232.0 Ni 5 0.2 5 248.3 Fe 0.2 3 285.2 Mg
cu
0.5
3 3 5
324.8 328.1 357.9
Ag Cr
0.5 0.2
a Flame: Air-acetylene. Readout: 10-s integration. Scale expansion: X 50.
+
and %T units, respectively. This gave ( U B E ~ 6 ~ ~ ) ~The ' ~ .noise due to the signal processing prior to the %T readout should actually be included but this was assumed to be negligible. 4) The standard deviation in the %T mode was determined with the lamp off, no solution aspirating,and the flame off. This was converted to absorbance using Equation 3, giving UD. In addition to steps 1-4 which were carried out for each element, the following steps were carried out once only. 5 ) The standard deviation in the absorbance was determined with the photomultiplier signal replaced by a low-noise simulated signal, giving UAR. For this step, the chart recorder was used to obtain an extra twenty times scale expansion (makinga total of 100OX) so that the noise could be resolved. 6) The standard deviation in the absorbance was determined for As with the flame off and a very narrow slit such that the photocathode current was about 1%of the current with the normal 1-nm spectral bandwidth. Under these conditions up >> UL and the measured standard deviation can be assumed to be (up2 + U D ~+ l7*R2)"2.
7) The standard deviation in the %T mode was determined under the same conditions as step 6 but with the lamp off. After conversion to absorbance, this gave UJJfor these conditions.
RESULTS AND DISCUSSION The procedure outlined in the previous section enabled the individual contributions to the noise to be calculated for each element. 1)Amplifier-Readout Noise. Step 5 gave UAR = 2 X absorbance. ab2) D a r k C u r r e n t Noise. Step 4 gave UD < 1 X sorbance for every element. The resolution of the chart recorder was not sufficient for an accurate measurement to be made.
3) Photon Noise. The standard deviation obtained in the absorblow intensity measurement of step 6 was 2.1 X ance and the dark current component measured in step 7 was 3.0 X absorbance. Substitution for UD and UAR gave u p 2.1 X absorbance (i.e., UD and UAR both proved to be negligible) for the particular conditions of step 6. The photon noise applicable to step 1(flame off) and step 2 (flame on) was then calculated by determining the change in photocathode current and using Equation 1. 4) Lamp Flicker Noise. UL was calculated from the standard deviation measured in step 1using the previously determined values of UAR, UD, and up. (UD and UAR again proved to be negligible for all elements investigated.) The value of UL obtained from this calculation is less reliable if aLis much smaller than the measured standard deviation. 5 ) Background Emission Noise. UBE was obtained from the standard deviation measured in step 3. 6) Flame Transmission Noise. UF was calculated by substituting the UA measured in step 2 into Equation 2 along with the previously determined values of up, ul,UBE, UD, and UAR. As with the determination of UL, this calculation is less accurate if UF
