JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1986, VOL. 1 75 Signal to Noise Ratios for Flow Injection Atomic Absorption Spectrometry James M. Harnly and Gary R. Beecher US Department of Agriculture, Agricultural Research Service, Beltsville Human Nutrition Research Center, Nutrient Composition Laboratory, Belts ville, M D 20705, USA Signal to noise ratios for flow injection coupled with flame atomic absorption spectrometry (FI-AAS) were compared with conventional nebulisation. Peak heights and areas were determined for a series of sample volumes (20, 100 and 500 pl) and flow-rates (3.2, 1.6,0.8 and 0.4 ml min-l). The reduced sample flow-rates (with a constant nebuliser gas flow-rate) improved nebulisation efficiencies by factors of 4-1 2. Peak-area signal to noise ratios were better than those for peak heights in every instance.The signal to noise ratios for FI-AAS peak height and area approached but never exceeded the ratios for conventional nebulisation. Keywords: Flow injection; flame atomic absorption spectrometry; signal to noise ratio Flow injection (FI), combined with flame atomic absorption spectrometry (FAAS), is a versatile tool.' The attractive features of FI-AAS are: (a) small sample volumes; (b) rapid sample throughput; ( c ) discrete nebulisation (tolerance of high salt concentrations, minimisation of viscosity effects and reduction of chemical interference effects); ( d ) alternative calibration possibilites; and (e) matrix dilution (addition of reagents for suppressing interferences). A disadvantage of FI-AAS is the inherently poorer signal to noise ratios arising from the smaller sample volumes.2 A feature of FI-AAS that has remained largely unexplored is the increased nebulisation efficiency, or signal to volume ratio, obtained using reduced sample pumping rates while the nebuliser gas flow is held constant.s6 Without a flow injector, a nebuliser will take up liquid at a rate that is determined by the gas flow-rate and the length and bore of the capillary tubing.The analyst adjusts the gas flow or (in the case of a variable nebuliser) the aspiration rate to give maximum sensitivity. The optimum sample uptake rate is usually 6-10 ml min-1. Wolf and Stewart5 showed that reducing the sample uptake rate, without changing any other parameters, reduces the signal but increases the atomisation efficiency.The net result is a 33% decrease in the peak height but a four-fold increase in the peak area (provided that all of the signals fall in the linear range). The increased nebulisation efficiency was also accompanied by an increase in the duration of the analytical signal and an increase in the area noise. As a result, peak-area signal to noise ratio increased by a factor of 1.6. Thus using reduced sample flow-rates and peak integration, improved signal to noise ratios can be obtained. It has been demonstrated that higher peak heights can be obtained for FI-AAS if the sample pumping rate exceeds the optimum aspiration rate of the nebuliser.7 This increase in the peak height, however, was achieved through the increase in the sample delivery not an increase in the nebulisation efficiency.The nebulisation efficiency decreases as can be seen by the reported less than linear response of the sample signal to the sample flow-rate.' To date, increased efficiencies for conventional nebulisers have only been achieved by reducing the sample pumping rate. This paper examines the peak-height and -area signal to noise ratios that are achieved for a series of reduced sample pumping rates while the nebuliser gas flow-rate is held constant. The signal to noise ratios are also examined as a function of the sample volume and are compared with conventional nebulisation. The trade-offs in the choice of FI-AAS parameters are considered. Experimental The FI system has been described previously.5 A standard straight-line configuration with one six-port rotary sampling valve (Model AH60, Valco Instrument Co., Houston, TX) was used with sample loops of 20 (0.05 cm i.d.), 100 (0.08 cm i.d.) and 500 p1 (0.04 cm i.d.).A 15 cm long (0.06 cm i.d.) section of tubing connected the sampling valve to a conven- tional pneumatic nebuliser. The sample pumping rate was determined by a variable rate positive displacement minipump (Milton Roy Co., Riveria Beach, FL) while the nebuliser gas flow was held constant. Flow-rates of 3.2, 1.6, 0.8, and 0.4 ml min-1 were used. Data acquisition was accomplished with a simultaneous multi-element atomic absorption continuum source spec- trometer (SIMAAC) , which has been described previously.8~9 With SIMAAC, data can be acquired for up to 30 s and the analytical signal can be integrated over any pre-set interval to the nearest 1/56 of a second.Data acquisition was triggered at the time the sample loop was switched into the main stream. Optimum integration intervals were determined for each sample volume and flow-rate by measuring the appearance time and ending time of the peaks from strip-chart recorder Table 1. Integration interval (seconds) Sample Sample size/pl ml min-1 20 100 500 3.2 4.3 7.5 15.5 1.6 6.4 10.7 26.9 0.8 10.7 20.0 0.4 17 pumping rate/ - - - Table 2. FI-AAS peak heights (absorbance)* Sample Sample size/yl ml min-1 20 100 500 pumping rate! 3.2 0.0585 0.118 0.148 1.6 0.0509 0.0947 0.123 0.8 0.0389 0.0685 0.4 - - 0.0232 - * Conventional nebulisation (9.0 ml min-1) = 0.1572 A.76 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1986, VOL.1 Table 3. FI-AAS peak areas (absorbance seconds)* ~ Table 4. FI-AAS peak-height coefficients of variation* Sample Sample size/pl pumping rate/ ml min-1 20 100 500 3.2 0.0588 0.313 1.18 1.6 0.0966 0.481 1.70 0.8 0.136 0.679 0.4 0.170 - - - * Conventional nebulisation (9.0 ml min-I) for 5 s = 0.780 A s. 0.20 0.16 cn 5 0.12 (0 2 Y a 0.08 0.04 I I I 1 .o 2.0 3.0 Sample pumping rate/ml min-’ Fig. 1. Peak area versus the sample pumping rate for a 20-1.1.1 sample volume. Nebuliser gas was held at a constant flow, which provided a conventional sample nebulisation rate of 9.0 ml min-l tracings. The intergration intervals used in this study are shown in Table 1. No data were collected at 100 pl for a sample pumping rate of 0.4 ml min-1 or at 500 pl for sample pumping rates of 0.8 or 0.4 ml min-1 as the peak lasted longer than 30 s, the integration limit of the computer program.For this study, data were only acquired for Cu, 324.7 nm, in an air - acetylene flame. Peak heights and areas were determined for each of the sample loop sizes and pumping rates listed above. Ten measurements were made for each set of conditions for a standard and a blank. Data for conventional nebulisation were obtained without the FI, using the same gas flow-rate and a 16.5 cm (0.06 cm i.d.) piece of polyethylene capillary tubing attached to the nebuliser. Results and Discussion The FI-AAS peak heights and areas for 10 pg ml-1 of Cu are listed in Tables 2 and 3, respectively, as functions of the sample volume and pumping rate.The largest peak heights were produced, as expected, by the largest sample volumes and the fastest pumping rates. The best peak-height signals for FI, however, were still less than that for conventional nebulisation, which gave a peak height of 0.157 at an uptake rate of 9.0 ml min-1. Larger peak heights can be obtained for FI-AAS with faster sample pumping rates or larger sample sizes, but the maximum peak height can only be equivalent to that for conventional nebulisation.5J0J1 The largest peak areas were obtained for the largest sample volumes but at the lowest pumping rates, and consequently, the longest integration periods (Table 3). For the smaller sample volumes (20 and 100 pl), the peak areas were proportional to the volume.At 500 pl, the areas were no longer proportional due to a much larger fraction of the sample peak exceeding the linear absorbance range (greater Sample Sample size/pl pumping ratel ml min-1 20 100 500 3.2 4.4 1 .o 0.5 1.6 1.5 0.5 0.9 0.8 2.7 1.3 - 0.4 9.6 - - * Conventional nebulisation (9.0 ml min-1); coefficient of variation = 0.003. Table 5. FI-AAS peak-area coefficients of variation* Sample Sample size/pl pumping ratel ml min-1 20 100 500 3.2 4.6 0.2 0.3 1.6 1.8 1 .o 0.3 0.8 2.1 0.8 0.4 2.6 - - - * Conventional nebulisation (9.0 ml min-I); coefficient of variation = 0.008. than 0.1; the linear range ends at a lower absorbance for SIMAAC than for conventional line source AAS). Reducing the sample pumping rate produced an accelerat- ing increase in the nebulisation efficiency.This is illustrated by the graph of the peak area versus pumping rate in Fig. 1. Injection of 100 or 500 1-11 at pumping rates of between 0.8 and 3.2 ml min-1 gave peak areas ranging from approximately a factor of two worse to approximately a factor of two better than the value of 0.78 A s obtained for conventional nebulisation for 5 s (750 pl). An integration period of 5 s was chosen for conventional nebulisation as this interval is used in our laboratory for all routine determinations. The areas of the three undetermined values in Table 3 (100 pl at 0.4 ml min-1 and 500 yl at 0.4 and 0.8 ml min-1) will all exceed the value for conventional nebulisation (0.78 A s). Unfortunately the required integration period (35-80 s) exceeded the limits of our data acquisition program.The data in Table 3 cannot be used to compare the nebulisation efficiency, or signal to volume ratios, of FI and conventional nebulisation as some of the absorbances fell in the non-linear region. Nebulisation of a standard of a lower concentration (in the linear range) permitted the signal to volume ratios to be compared. The FI efficiencies were 4.1, 6.8,9.5 and 11.9 times greater than conventional nebulisation at sample pumping rates of 3.2, 1.6, 0.8 and 0.4 ml min-1, respectively. The cost of these increased efficiencies is increased integration time (Table 1) and reduced sample throughput. Signal to noise ratios were computed for the data in Tables 2 and 3 in two ways. The first method employed ten repeat determinations of the analytical signals and used the com- puted means and standard deviations to compute the coeffi- cients of variation.The standard deviations of these measure- ments reflect the base-line noise contribution (arising from the light source)and the analytical signal and matrix background fluctuation noises. As the detected limit is approached the sample noise components diminish in significance until the base-line noise is limiting. Accordingly, the second method computed signal to noise ratios using the analytical signals and the base-line noises. Tables 4 and 5 give the coefficients of variation, based on precision of the analytical signals, shown in Tables 2 and 3. For 100- and 500-yl sample volumes, the relative precisions forJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1986, VOL.1 77 Table 6. FI-AAS peak-height signal to base line noise ratios* Sample Sample size/yl pumping rate/ ml min-1 20 100 500 3.2 39 79 99 1.6 34 63 82 0.8 26 46 0.4 15 - - - * Conventional nebulisation (9.0 ml min-l) = 105. Table 7. FI-AAS peak-area signal to base line noise ratios* Sample Sample size/yl pumping rate/ ml min-1 20 100 500 3.2 79 277 77 1 1.6 98 398 842 0.8 112 402 - 0.4 102 - - * Conventional nebulisation (9.0 ml min- l ) = 904. Table 8. FI-AAS area base-line standard deviation (absorbance seconds) * Sample Sample size/p1 ml min-1 20 100 500 pumping rate/ 3.2 0.007 0.001 1 0.0015 1.6 0.0010 0.0012 0.0020 0.8 0.0012 0.0017 - 0.4 0.0017 - - * Convention nebulisation (9.0 ml min-1) for 5 s, area base-line standard deviation = 0.0009 A s.lo-* ~ , j 10 100 10-4 1 .o In teg ration t ime/s Fig. 2. Area base-line standard deviation versus integration time both height and area were close to 1.0% or better. Sample volumes of 20 pl yielded consistently poorer coefficients of variation than 100 or 500 1-11, ranging from 2 to 10% for peak height and 2 to 5% for peak area. The peak-height and -area signal to noise ratios, based on the base-line noise levels, are shown in Tables 6 and 7. These values are inversely proportional to the detection limit. The base-line noise level for the peak-height measurements was based on the standard deviation of the base-line absorbance. The standard deviation of the base-line absorbance is indepen- dent of the number of values used for the computation; the more values that are used, the greater the confidence of the computed standard deviation.For this study, a standard deviation of 40.0015, the average of several computations, was used for determining the peak-height signal to noise ratios. The base-line noise levels for peak area were based on the standard deviation of ten or more repeat integrations for the atomisation of a blank for the time intervals listed in Table 1. The area base-line standard deviations (Table 8) are propor- tional to the square root of the integration time. Fig. 2 illustrates this relationship. A first-order least-squares fit to the logarithms of the data gave a slope of 0.4998. This means the area base-line noise is shot-noise limited: i.e. , limited by the uncertainty of the arrival of photons at the photomultiplier tube cathode.The values listed in Table 8 were used in computing the peak-area signal to noise ratios. In general, the peak-height signal to noise ratios showed the same pattern as the peak-height data (Table 2) while the area signal to noise ratios showed less of an increase than might be expected from the area data (Table 3). This reflects the constant nature of the base-line absorbance noise compared with the time dependent base-line area noise. None of the FI-AAS height or area signal to noise ratios exceeded the signal to noise ratios of conventional nebulisation for height, 105, or area, 904. The data in Tables 6 and 7 illustrate the compromises involved in the selection of operating parameters for FI-AAS. The most consistent generalisation to be found in the data is that the signal to noise ratios for peak area are better than those for peak height.Selection of other operating parameters depends on the needs of the analyst. For samples requiring the best detection limits and having no volume limitations, conventional nebulisation offers the best detection limits. FI-AAS, using 100-p1 sample volumes, offers signal to noise ratios comparable to 5 s of conventional nebulisation. Still larger sample volumes and integration intervals would offer better signal to noise ratios for both methods. However, FI-AAS attains comparable detection limits primarily by imitating conventional nebulisation (large sample volumes and high flow-rates) and thus sacrificing some of the more desirable features of FI.However, where sample volume, not detection limits, is critical, FI-AAS offers high precision for very small sample volumes. In this study, 20 p1, the smallest sample volume injected, gave relative peak-area precisions of 2% (Table 5). Smaller sample volumes, with larger relative precisions, could be used depending on the requirements of the determination. Using these volumes sample throughput would be quite high. If both the detection limits and sample volume are critical, FI-AAS again offers useful compromises. Conventional nebu- lisation requires 750 1.11 for a 5-s data acquisition (at a sample uptake rate of 9.0 ml min-1) and yields a signal to noise ratio of 900. Using 100 1-11 of sample and a pumping rate of 1.6 ml min-l, FI-AAS with peak-area measurement can achieve a signal to noise ratio of 400, i .e . , a detection limit only a factor of two worse than conventional nebulisation, but requiring just 13% of the sample volume. Finally, with no detection limit or sample size restrictions, FI-AAS and conventional nebulisation, with an autosampler, are quite similar. FI-AAS has the advantage of increased precision in sample handling, fewer physical interferences, a higher rate of determination and general versatility. While the data reported in this study are highly specific to the experimental set-up employed, the general trends can be applied to FI-AAS in general. The integration times and peak heights and areas (Tables 1, 2 and 3) are dependent on the nebuliser and the FI configuration. However, reducing the sample pumping rate (while holding the gas flow constant) can be expected to reduce the signal and increase the atomisation efficiency of any pneumatic nebuliser .The exact noise levels will depend on the AAS, but almost all spectrometers employing hollow-cathode lamps are shot-noise limited (the integrated base-line noise is proportional to the square root of78 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1986, VOL. 1 the integration time). Consequently, while it may be useful for the analyst to characterise individual spectrometers, the conclusions of this study should be helpful in predicting signals and signal to noise ratios for FI-AAS. 6. 7. 8. 9. References 10. 1. 2. 3. 4. 5. Tyson, J. F., Analyst, 1985, 110, 419. Tyson, J. F., and Sarkissian, L. L., Anal. Proc., 1985,22, 19. Jones, D. R., Tong, H. C., and Manahan, S. E., Anal. Chem., 1976, 48, 7. Szivos, K . , Polos, L., and Pungor, E., Spectrochim. Acta, Part B , 1976,31, 289. Wolf, W. R., and Stewart, K. K., Anal. Chem., 1979,51,1201. Koropchak, J. A., and Coleman, G. N., Anal. Chem., 1980, 52, 1252. Brown, M. W., and RGiiEka, J., Analyst, 1984, 109, 1091. Harnly, J. M., O’Haver, T. C., Golden, B., and Wolf, W. R., Anal. Chem., 1979, 51, 2007. Harnly, J. M., Miller-Ihli, N. J., and O’Haver, T. C., J . Autom. Chem., 1982, 4, 54. Treit, J., Nielsen, J. S., Kratochvil, B., and Cantwell, F. F., Anal. Chem., 1983, 55, 1650. Olsen, S., Pessenda, L. C. R., RGiiEka, J., and Hansen, E. H., Analyst, 1983, 108, 905. Paper J.5126 Received August 14th, 1985 Accepted September 18th, 1985