Anal. Chem. 1982, 5 4 , 1411-1419
1411
Aerosol Trainsport Model for Atomic Spectrometry Richard F. Browner," Andrew W. Boorn,' and David D. Smith School of Chemistty, Georgia Institute of Tcxhnoiogy, Atlanta, Georgia 30332
The transport of aerorol In an atomic spectrometric system Is explained In terms of the lnteractlon of a primary generation process, with varlous swondary and tertlary aerosol modlfylng steps. Secondary lmpiactlon steps may cause the generation of addltlonal small drolplets but are primarily loss processes. Aerosol loss also usuallly predomlnates In tertlary steps, as a result of Impaction, turbulence, gravitatkonal, and centrlfugal processes. Aerosol evaporation may also cause an increase In aerosol transport for volatile solvents. The relative lmporlance of all these plrocesses Is dlscussed in terms of typlcal atomic absorptlon (AA) and Inductively coupled plasma (ICP) nebullzer/sprayr chamber systems. The concept of useful and excess anallyte mass contalned In specified drop size ranges Is dlscussed. Finally, certaln aerosol transport propertles, such as tralnsport efflclency, E", and total analyte mass transported per second, WLot,are calculated on a semltheoretlcal bask and shown to correlate with experimental data.
Optimization of the sample introduction process in atomic spectrometry is generislly carried out on an empirical basis. Consequently, a suitable compromise between signal magnitude and interference severity is often only achieved after lengthy tests involving systematic changes in variables such as nebulizer and spray chamber geometry, gas and liquid flow, etc. Using such an aplproach, it is often extremely difficult to differentiate between changes in analytical signal that result from modification in aerosol properties and those that result from changes in atomizer properties. An example is the influence of varying nebulization gas flow rate on emission signals in inductively coupled plasma (ICP) spectrometry. In this instance, nebulization gas flow influences both plasma excitation characteristics ( I ) and atorn residence time, in addition to its influence on aerosol generation and transport properties. It is possible, in principle, to differentiate between the effects of system variables on aerosol properties and the effects of the same variables om atomization and excitation conditions by measuring appropriate critical aerosol parameters. Two such parameters would be droplet size distribution and analyte mass transport rate. Once known, these values could be used to subtract aerosol related variables from the net change in analytical signal, and hence determine the true extent of atomization and excitation variables. This procedure would allow system optimization to be achieved on a more rational basis than the current trial and error procedure. Realistically, however, neither the experimental capabilities nor the time are available under normal circumstances to carry out extensive studies of' aerosol properties. Nevertheless, there is a real need for effective sample introduction optimization under differing experimental conditions. For this purpose, a working model allowing at least semiquantitative estimates of the influence of system variables on aerosol properties would Present address: SpectraMetrics, Inc., 204 Andover St. Andover, MA 01810. 0003-2700/82/0354-1411$01.25/0
be of considerable value. In this paper, an attempt is made to derive a simple, comprehensive model of aerosol transport processes applicable to several typical atomic absorption (AA) and ICP nebulizer/spray chamber systems. The predictive accuracy of the model is tested against experimental data for an AA nebulizer/spray chamber system and found to give quite good agreement. T o our knowledge, this is the first attempt that has been made to derive such a model for atomic spectrometry.
EXPERIMENTAL SECTION Procedures used for droplet size distribution and transport efficiency measurements have been described in earlier publications. The size distributionswere determined by using a cascade impactor, as described by Cresser and Browner (2),and the transport efficiency measurements were made with either a cascade impactor or a filter system, as described by Smith and Browner (3). Atomic absorption and inductively coupled plasma measurements were made, under specified conditions, using a Perkin-Elmer Model 5000 AA/ICP spectrometer system. Data Presentation. The drop size distribution plots used in this paper differ from those we have published previously (2),in that the ordinate axis is normalized on a linear, rather than a logarithmic, basis. Droplet diameter is also presented as a linear scale. Normalization is necessary to allow for unequal size interval steps in the collection stages of the cascade impactor. Logarithmic normalization was used when measurements covering a wide droplet size range were measured. However, linear normalization is a simpler procedure and does not place the undue emphasis on certain points which the logarithmic procedure may. Additionally, areas under the graphs plotted linearly are proportional to the total masses per second of analyte contained in the distribution curves considered, which simplifies calculations. RESULTS AND DISCUSSION Aerosol Transport Model. The processes which occur in aerosol generation and transport can be conveniently divided into three categories (Figure 1). The primary aerosol generation step, where the bulk liquid is initially shattered by pneumatic, ultrasonic, or other means into an aerosol, will generally result in a polydisperse droplet distribution ( 4 ) , unless specialized means are used to generate a monodisperse droplet stream (5). At this stage, the droplet size distribution can conveniently be considered as the primary distribution. In many systems a secondary process of impaction then occurs, where the relatively high velocity droplet stream impinges on an impact surface, usually of approximately hemispherical geometry. Large droplets, on striking the impact surface, will generally adhere and run to waste, small dropIets will tend to follow the air stream around the surface without change, and intermediate droplets with sufficiently high velocity may shatter and generate further small droplets. The impact surface, therefore, modifies the primary drop size distribution by shifting it to a smaller median diameter and the aerosol now assumes its secondary drop size distribution. The tertiary stage involves those processes, namely, impaction, turbulent and centrifugal loss, gravitational settling, and evaporation, which modify the drop size distribution in the region beyond the nebulizer/impact surface but prior to the atomization cell. These processes will generally act to shift the aerosol size distribution to smaller droplets. Impaction and settling processes will result in preferential loss of larger 0 1982 American Chemical Soclety
1412
ANALYTICAL CHEMISTRY, VOL. 54, PRiMARY SECONDARY
NO. 8, JULY 1982
TERTiARY
i20 1
EvaDoration
in) 200
z u
-5
100
0
100
05
10
20
15
Q , ASPIRATION RATE (rnL/rnin)
Flgure 1. Stages in aerosol generation and modification from nebulizer to atomization cell.
droplets, consequently decreasing the rate of analyte transport to the atomization cell. In contrast, evaporation causes a downward shift in the size distribution, resulting in a net gain of smaller droplets and an increase in analyte transport rate. Following these steps, the aerosol will have its ultimate size distribution prior to entry into the atomization cell. This distribution may be called the tertiary or preatomization drop size distribution. The tertiary drop size distribution is the aerosol property of greatest value in predicting the properties of particle dynamics (6), droplet evaporation (7, a), and ultimately atomization in the flame or plasma. However, it is the interaction of all three stages in the generation and transport process which determine the efficiency of analyte transport (e.g., efficiency of nebulization) and the net rate of analyte mass transport to the atomization cell. Figures of Merit for Nebulizer/Spray Chamber Systems. The experimental parameter most commonly used to assess nebulizer and spray chamber performance is the transport efficiency, en, also known as the nebulization efficiency (3). The direct use of transport efficiency values to compare different nebulizer/spray chamber systems or to optimize operating parameters for a particular system is, however, of limited value in the absence of aspiration rate data. This parameter expresses only the ratio of analyte mass reaching the atomization cell to total analyte mass aspirated, and so has clear significance only when the sample volume is limited and when it is desired to make maximum use of the small volume available. A more valuable figure of merit to compare systems, because of its more direct relationship to the magnitude of the analytical signal, is the mass of analyte reaching the atomization cell per second, W. This parameter is already available experimentally from e, data, provided that the duration of the collection period is recorded. Although a superior indication of analytical performance to the transport efficiency, W still lacks any direct correlation with signal magnitude, because it contains no information regarding the size of aerosol droplets containing the analyte mass. Clearly, if the solvent is to be evaporated rapidly, and the analyte itself vaporized completely in the time available to it in the atomization cell, the preatomization drop size distribution must be narrow and centered around a small median diameter. The parameter of most direct use for correlation of aerosol properties and analytical signal will combine the W parameter with preatomization drop size distribution data. For aerosols typically produced by atomic absorption and ICP nebulizer/spray chamber combinations, such information is readily available from cascade impactor data, as the aerosol collected during a timed run is segregated according to drop size range. For a plate collection range of d to d ' ( p m ) the relationship is mass transport rate over total drop size range d'
d, d' =
CW d
(1)
Figure 2. Plots of emission intensity, useful analyte mass ( W J , and transport efficiency (en) vs. sample aspiration rate: Meinhard nebulizer, operated at 1.0 L/mh Ar flow, with dual concentric spray chamber (27);solution of 0.5 pg/mL Mg aspirated, and emission signal measured at 279.55 nm with Perkin-Elmer 5000 ICP spectrometer.
If the aerosol collected on the impactor corresponds to the entire tertiary drop size distribution, then xdd'W would represent the total aerosol mass reaching the atomizer per second, represented by Wtot. In aerosol optimization studies, it would be convenient to specify a maximum droplet diameter, d,,,, such that only droplets equal to or smaller than this diameter would contribute significantly to the analytical signal. In reality, there is no sharp cutoff diameter which differentiates between droplets which lead to totally vaporized analyte and droplets which produce no analyte vapor at all at the point of measurement. Nevertheless, the concept of a maximum useful droplet diameter can be helpful in relating the extent of vaporization in the atomizer to the aerosol size distribution. In this discussion, therefore, d,, is taken to refer to a limiting value of droplet size in the tertiary aerosol which contributes 5000 m,g/mL) where evaporation should be minimal. Comparison of Aerlosol Loss Processes for Various Spray Chamber Configurations. From the foregoing discussion, it is not immediately clear where the dominant aerosol loss process occurs in all Ai4 systems. However, a simple and unequivocal test to distinguish between impaction losses on the one hand and turbullence (or gravitational) losses on the other, is provided by transport efficiency (or Wt0J measurements. Table I shows t, values obtained for a Perkin-Elmer nebulizer and spray chamber operated in a variety of configurations, with varying: auxiliary gas flows. Nebulizer conditions, e.g., aspiration rate and gas flow rate, were held constant. In the absence of either an impact bead or mixer paddle, the transport efficiency increases by 53% for an auxiliary air flow change from 0 to 8.8 L/niin. With a paddle in place, the increase is 33% and for an impact bead 62%, for the same gas flow change. For an impaction or centrifugal dominated tertiary loss process, E , should decrease with increasing auxiliary gas flow. The degree to which e, values increase with increasing gas flow is therefore a measure of the turbulence loss occurring in the spray chamber (gravitational loss can be neglected under these conditions). Turbulence loss therefore appears to be greatest for the system operated with an impact bead, intermediate for the system operated without either impact bead or paddle and least for the system operated with i3 paddle. It is interesting to note that under all circumstances there appears to be significant turbulence loss occurring. It is also clear from these results that the aerosol generated using an impact bead is modified further by passing through the spray chamber and that this modification is primarily as a result of turbulence loss processes. Combined Model for Prediction of Aerosol Transport Properties, t, and W . The models previously discussed are not sufficiently refined to allow calculations from first principles. However, with the aid of certain experimental data, necessary to define constants, it is possible to make quantitative predictions of t, and W for certain systems. The system considered, because of its practical importance, was an atomic absorption spray chamber operated with a
15 26 36
0.6 5.4 8.5
Assumed Values effective nebulizer gas 5.0 X cm2 orifice area nebulizer gas velocity 330 m/s nebulizer liquid velocity 410 m/s 320 m/s V O 73 dyn/c3m P 1.0 g/cm 1.0 x P r) 0.6, 5.4, and 8.5 mL/min Q1 1 0 L/min Qe
ii
0
dc
5.0
10.0
d , DROPLET DIAMETER (pm)
Experimental tertiary drop size distribution curves for AA nebulizer and spray chamber, with paddle. Operating conditions are given in Table 11. Aspiration rates are as follows: (A)0.6 mL/min; Figure 7.
(X) 5.4 mL/min; ( 0 )8.5 L/min.
pneumatic nebulizer and a mixer paddle. The parameters used for the calculations are given in Table 11. The one variable considered was sample aspiration rate, and the parameters to be determined were E , and W. In order to simplify the calculations, the following assumptions were made: (1) Sample aspiration rate is controlled by a peristaltic pump, so that nebulizer gas flow and velocity remain constant. (2) The mixer paddle used in the spray chamber causes aerosol loss predominantly by impaction and centrifugal processes. Tertiary distribution curves will therefore follow a l / d 2 cutoff. The cutoff diameter and its associated mass flow term, W , are the only experimental values necessary to describe the aerosol emerging from the spray chamber. All other values are calculated from these points, using assumptions about the nature of the aerosol and the cutoff process to be described in the following sections. Experimental Drop Size Distributions. The tertiary drop size distributions, obtained with a cascade impactor, for aspiration rates of 0.6,5.4, and 8.5 mL/min of a 20000 pg/mL Mg solution are shown in Figure 7. Measurement of the 50% cutoff diameter, d,, from these curves gave a mean value for d, of 2.3 A 0.1 pm. From this cutoff diameter, the experi-
1418
ANALYTICAL CHEMISTRY, VOL. 54, NO. 8, JULY 1982
Table 111. Comparison of Calculated and Experimental Values for Wand e n u
&,, mL/min
W(calcd),b p g / s
W(exptl),c pg/s
0.60
15
13 34
5.4 8.5
33
e,(calcd),b %
41
en(exptl),c %
13
13
4.2 2.9
4.1 2. I
Values calculated for nebulizer conditions given in Table 11. Distribution curves are shown in Figures 8 and 9. e, values calculated from areas under curves in Figures 8 and 9, as indicated in text. Data obtained directly, by summation of analyte mass collected on plates of cascade impactor ( 3 ) . a
Wand
60
5 40
5 -
l n
1
s U
20
0 0 0
5.0
10.0
d , DROPLET DIAMETER (IJm)
Figure 8. Theoretical tertiary drop size distribution curves, derived from Figure 7. Details explained in text.
mental ordinate values of analyte mass per second, corresponding to this droplet size, were determined for each curve. The curves for 5.4 and 8.5 mL/min aspiration rate were sufficiently close to be treated as superimposable for this stage of the calculation. Theoretical Tertiary Distribution Curves. From each pair of experimental (x,y) coordinates a theoretical cutoff curve was generated (Figure 8), assuming a simple inverse square law relationship (as anticipated for an impaction dominated loss process). The equation for the 0.6 mL aspiration rate curve was W = 2 2 / d 2 and that for the combined 5.4 and 8.5 mL/min aspiration rate curves was W = 43/d2, where W is in pg/s and d in pm. In order to define the upper ends of the curves, a point d* was calculated, corresponding to the value of W equal to 2 times its value at the cutoff diameter. Values of d* calculated from these equations were identical a t 1.7 pm and were close to the experimental value (Figure 7) of 1.6 pm. This indicates a reasonable degree of fit between experimental and calculated cutoff curves at low d values. Above 5 pm, the degree of fit is evidently poorer, as can be seen from the figures. The portion of the curves between 0 and 1.7 pm was drawn simply by joining the points, as no a priori relationship is readily available to describe the curve shape accurately over this region. The area under each curve should now be proportional to the W value for that aspiration rate, e.g., the total mass per second in the tertiary distribution reaching the burner head. Theoretical Primary Distribution Curves. Sauter median droplet diameters were calculated for each nebulizer operating condition, using the Nukiyama and Tanasawa equation (eq 4). As discussed earlier, these can be taken practically as equivalent to mass median diameters for the aerosols con-
20
40
60
80
d . DROPLET DIAMETER (pm)
Figure 9. Theoretical primary drop size distribution curves for AA nebulizer. Parameters used for calculation are given in Table I1 and text.
sidered. The parameters assumed and calculated droplet diameters are given in Table 11. Drop size distribution curves for each aspiration rate were then plotted on the following assumptions: (1)The shape of each curve is approximately log normal. (2) The mass median diameter is that calculated from the Nukiyama and Tanasawa equation. (3) The W value for 1.7 pm diameter droplets corresponds to the peak value calculated, as shown in Figure 8. (4)The area under each curve is proportional to the total mass of analyte aspirated per second. This in turn is proportional to the aspiration rate. The curves are shown in Figure 9. W values for the different aspiration rates were calculated by integrating the areas under the tertiary distribution curves shown in Figure 8. These, together with experimental W values obtained by summing the analyte mass collected on all plates of a cascade impactor, are shown in Table 111. Transport efficiency en, values were calculated by ratioing the areas under the tertiary distribution curves to those under the primary distribution curves (Figure 9). Experimental t, values were obtained using the direct procedure of Smith and Browner (3)and ratioing the total analyte mass collected on a cascade impactor to the analyte mass aspirated during the same period, These values are also reported in Table 111. The degree of fit between calculated and experimental values for both Wand en is remarkably good, considering the many assumptions and approximations inherent in the procedure. This would indicate that the basic approach is sound and capable of extension to other systems.
CONCLUSIONS The processes which act to modify aerosols under analytical conditions are complex and only partially explicable with the present state of knowledge. Nevertheless, the application of
Anal. Chem. 1982, 5 4 , 1419-1424
even simple models to various stages of the transport process gives useful information regarding the relative importance of possible aerosol modifying mechanisms. As yet, the data available for several experimental parameters of crucial importance, such as d, for a wide range of elements in various matrices, are not available. Access to these data should allow “ideal” aerosol characteiristics to be specified more precisely for AA and ICP determinations. The production of an aerosol approximating ideal characteristics should be achievable using the models described taken together with certain experimental data, such as drop size distributions andl E, values. Further refinement of the overall aerosol transport model is clearly desirable, as this should lead to a better understanding of the complete sample introduction process in atomic spectrometry and (alsoallow more complete prediction of the transport process. A direct consequence of the improved understanding of aerosol transport mechanisms should be more accurate and precise measurements and possibly also improved dectection limits for certain elements.
LITERATURE CITED Horlick, G.; Blades, M. \N. Appl. Spectrosc. 1980, 3 4 , 229. Cresser, M. S.; Browner, R. F. Spectrochlm. Acta, Part B 1980, 358, 73. Smith, D. D.; Browner, IR. F. Anal. Chem. 1982, 5 4 , 533. Nukiyama, S . ; Tanasawe, R. “Experiments on the Atomization of Llquids in an Air Stream”; Hope, E., Transl.; Defense Research Board, Department of National Defense: Ottawa, Canada, 1950. Berglund, R. N.; Liu, B. Y. H. Environ. Scl. Techno/. 1973, 7, 147.
1419
L’vov, B. V.; Katskov, D. A.; Kruglikova, L. P.; Polzik, L. K. Spectrochlm. Acta, Part B 1976, 3 1 8 , 49. Bastiaans, G. J.; Hieftje, G. M. Anal. Chem. 1974, 4 6 , 901. Barnes, R. M.; Nlkdel, S. Appl. Spectrosc. 1976, 3 0 , 310. Boumans, P. W. J. M.; DeBoer, F. J. Spectrochlm. Acta, Part B 1978, 318, 355. Alder, J. F.; Bombelka, R. M.; Kirkbright, G. F. Spectrochim. Acta, Part B 1980, 358, 163. Boorn, A. W.; Browner, R. F. Anal. Chem. 1982, 5 4 , 1402. Novak, J. W.; Browner, R. F., unpublished work. Mercer, T. T. Health Phys. 1964, IO, 873. Fuchs, N. A. “The Mechanlcs of Aerosols”; Pergamon Press: Oxford, 1964. Friedlander, S. K. “Smoke, Dust and Haze”; Wiley: New York, 1977; pp 104-109. Novak, J. W.; Llilie, D. E.; Boorn, A. W.; Browner, R. F. Anal. Chem. 1980, 5 2 , 579. Prandtl, L. Phys. Z.1907, 8, 23. Routh, M. W. Appl. Spectrosc. 1981, 3 5 , 170. Cresser, M. S.;Browner, R. F., Appl. Spectrosc. 1980, 3 4 , 364. Skogerboe, R. K.; Olson, K. W. Appl. Spectrosc. 1978, 3 2 , 181. Scott, R. H.; Fassel, V. A,; Kniseley, R. N.; Dixon, D. E. Anal. Chem. 1974, 4 6 , 75. Walton, W. H. Brlt. J . Appl. Phys., Suppl. 1952, no. 3 , 529. Mercer, T. T.; Stafford, R. G. Ann. Occup. Hyg. 1969, 12, 41. Liu, B. Y. H.; Agarwal, J. K. Aerosol Scl. 1974, 5 , 145. Greenfield, S . ; Jones, I . L. L.; McGeachin, H. McD.; Smith, P. 8. Anal. Chim. Acta 1975, 7 4 , 225. Davies, C. N. Proc.-Inst. Mech. Eng., Part B 1952, I B , 185. Boorn, A. W.; Cresser, M. S.;Browner, R. F. Spectrochlm. Acta, Part B 1980, 358, 823.
RECEIVED for review January 7 , 1982. Accepted March 18, 1982. This work was supported by the National Science Foundation under Grant No. CHE-80 19947.
Carbon- 13 Fduclear Magnetic Resonance Spectrometry with Cross Polarization and Magic-Angle Spinning for Analysis of Lodgepole Pine Wood Waclaw Kolodziejski, James S. Frye,, and Gary E. Maclel” Department of Chemistty, (Colorado State University, Fort Collins, Colorado 80523
Carbon-13 nuclear maginetlc resonance (NMR) with cross polarization and maglc-angle spinning (CPiMAS) Is applied to whole and processed lodgepole plne wood to lnvestlgate the relative amounts of and Interactions between the major constltuents. Grlndlng the wood has no effect on the 13C spectrum, but ball milling converts “crystalllne” cellulose to the amorphous form. Extraction with dimethyl1 sulfoxlde or steam enrlches the resldue wHh “‘crystalllne” cellulose. Flve dlfferent llgnln preparations yield fractlons with distlnct 13C NMR spectra. Various preparritlons of holocellulose, hemlcellulose, and a-cellulose also give dlstlnct spectra. The differences among the llgnln and carbohydrate fractlons can arise either from chemlcal and physical modlflcatlon of the components or from dlfferent extradlon efflclencles for the components belng removed. The spectra provide lndlrect evldence for a Ilgnln-carbohydrate complex In that all llgnln fractlons exhlblt carbohydrate slgnals and vice versa.
Wood is one of the most important natural materials used by man. However, most of what we know about the chemical structure of wocd comes from studies of its components rather than from intact wood itself. These components are isolated
by destructive methods, so their structures are probably changed during preparation. This is especially true for lignin, which is easily modified. Recently there have been reports of the study of solid lignins by I3C NMR, using cross polarization and magic-angle spinning techniques (CP/MAS) (1-3). These studies assigned the peaks in lignin spectra and showed that lignin structure is dependent both on the preparation method and on the type of wood from which the lignin was isolated. Another component of wood, cellulose, has also been characterized with I3C CP/MAS (4-7). Resonances were assigned and signals from two types of glycosidic linkages were discussed. This high-resolution solid-state I3C NMR work has addressed the matter of the morphology and crystal structure of cellulose (6, 7). With preliminary 13C CP/MAS studies of wood’s major components already reported, it now seems appropriate to analyze wood itself by 13CCP/MAS, to consider interactions between wood components, and to study components and byproducts which have not been characterized previously by this method. These components include holocellulose, hemicellulose, various lignin fractions (Klason, Braun’s, and enzymatic), and a-cellulose. The present study is addressed to these goals.
0003-270O/82/0354-1419$01.25/0 0 1982 American Chemical Society
