JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 573 Influence of Solvent Physical Properties on Drop Size Distribution Transport and Sensitivity in Flame Atomic Absorption Spectrometry With Pneumatic Nebulization* Juan Mora Vicente Hernandis and Antonio Canals Division of Analytical Chemistry University of Alicante 03080 Alicante Spain The aim of this paper has been to study the influence that the main physical properties of the solvent (surface tension viscosity and volatility) exert on the magnitudes of the following parameters the drop size distribution of the primary aerosol; the transport efficiencies for analyte and solvent; and the sensitivity in flame atomic absorption spectrometry with pneumatic nebulization. Two series of experiments have been carried out one with pure solvents and another with methanol-water mixtures of variable composition always using 2 pg ml-' Mn as the analyte.The results show that surface tension and to a lesser extent viscosity determine the mean size of the distribution. The span on the other hand does not appear to follow any simple relationship with the physical properties of the solvent. The analyte transport rate is improved by a high solvent volatility andalso by a distribution with a small mean drop size and a high span. Absorbance values are essentially consistent with analyte transport rates this tendency being modified by the diluting effect of the solvent carried to the flame. Keywords Flame atomic absorption spectrometry; pneumatic nebulization; drop-size distribution; transport efficiency The most common method for sample introduction in flame atomic absorption spectrometry (FAAS) is through the pneumatic nebulization of solutions. In the nebuliza- tion step a spray (primary aerosol) is generated the characteristics of which have a great influence on the transport parameters (analyte transport rate en and solvent transport rate EJ.In turn these parameters together with the characteristics of the tertiary aerosol i.e. the aerosol that enters the flame base will determine the analytical signal. Usually the equation of Nukiyama and Tanasawa' has been used to describe the features of the primary aerosols generated using pneumatic nebulization. Several experimental studies have been accomplished recently on the applicability of this equation to the primary aerosols generated in atomic spectrometry under normal operating condition^.*-^ According to these studies the predictions of the equation differ noticeably from the experimental results.This applies to both inductively coupled plasma atomic emission spectrometry (ICP-AES)2-4 and FAAS.5 Owing to this discrepancy empirical equations3 or mathe- matical models6 have been proposed for the description of the primary aerosols generated by concentric nebulizers. Even though the number of basic experimental studies on aerosol generation in the field of atomic spectroscopy is not very great it is beyond doubt that the physical properties of the solvent have a great influence on the droplet formation ~ t e p ~ - ~ and also on the aerosol transport through the spray chamber to the atomization ell.^-^ Organic solvents or mixtures of water and organic solvents are currently used in atomic spectrometry (i) to separate interfering elements; (ii) to concentrate the ana- lyte; and (iii) to improve sensitivity.1° Several workers have tried to correlate the sensitivity improvements with solvent physical properties.l-l However these correlations appear to be fairly artificial in some instances as they do not try to understand and hence to explain the nebulization and transport processes as a whole. In order to do that it is necessary to perform a detailed individual study on each of the processes. By taking into account previous works on the nebuliza- tion step it appears that the physical properties of the ~~~ ~~~ * Presented at the Fifth Biennial National Atomic Spectroscopy Symposium (BNASS) Loughborough UK 18th-20th July 1990.solvent that show a greater influence on the characteristics of the pneumatically generated aerosols are surface tension viscosity and v ~ l a t i l i t y . ~ * ~ J ~ J ~ The mean drop size of the resulting aerosol decreases with decreasing surface tension (the main factor) and viscosity and also with increasing volatility. The surface tension of the solvent can be modified by using (a) aqueous solutions of surfactants; (b) mixtures of water and a compatible organic solvent; and (c) pure organic solvents. The effect of the forces of surface tension in situation (a) is the cause of a controversy frequently found in the literature. In a previous study by the same workers16 a possible explanation for the action of these forces in aqueous solutions of surfactants has been given.Following this explanation the forces require a certain time to exert their influence. This would be the time necessary for the surfactant molecules to migrate from the bulk of liquid to the surface being generated so that they can exhibit their surfactant action. Because this time is longer than that necessary for the nebulization step at least with long chain surfactants the resulting aerosols show the same characteristics as those obtained without the addition of surfactants i e . with the solvent alone.2 This study deals with the effect of surface tension and other physical properties of the solvent on the nebulization transport and atomization processes in situations (b) and (c) mentioned above.To this end drop size distributions of the primary aerosols transport efficiency for both analyte and solvent and absorbance have been sequentially measured under identical experimental conditions for a series of water-methanol mixtures and for a series of pure solvents with different values of their physical properties. Experimental All of the reagents employed were of analytical-reagent grade. The water used was distilled and de-ionized. The solvents employed together with their main physical properties are listed in Table 1. All of the solutions contained 2 pg ml-1 Mn as the analyte and were prepared in the same way as the standards from a stock solution of 2000 pg ml-1 Mn in water to which a small amount of concentrated HC1 was added (1% v/v) and diluted with the appropriate solvent to the working concentration.574 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL.6 The nebulizer the systems used for the control of the gas and liquid flows and the system used to measure (by an indirect method) the transport efficiencies have been described previously.16 All of the experiments were per- formed with a Perkin-Elmer Model 373 spectrometer under the experimental conditions shown in Table 2. Drop size distributions of the primary aerosols were measured at a distance of 28 mm from the nebulizer tip by means of a particle sizer based on laser Fraunhofer Table 1 Physical properties of solvents at 20 "C Pure solvents- Solvent Water Benzaldeh yde Formic acid Acetic acid Propan- 1-01 Methanol d r x lo3/ J t x lo3/ Nm-' N smm2 RVS 72.6 1 .oo 0.08 40.0 1.52 0.0 I 37.6 1.85 0.30 27.8 1.32 0.20 23.8 2.2 1 0.27 22.7 0.60 1 .oo Methanol-water ( v h ) mixtures- O X 103/ J X 1031 Methanol + water Nm-I N ~ r n - ~ RV o+ 100 10+ 90 25+ 75 50+ 50 60+ 40 80+ 20 90+ 10 loo+ 0 72.6 59.0 46.4 35.3 33.0 27.3 25.4 22.7 1 .oo 1.22 1.56 1.76 1.63 1.25 0.93 0.60 0.08 0.14 0.23 0.39 0.48 0.66 0.80 1 .oo * 0 =Surface tension.t J =Viscosity. $ RV=Relative volatility ratio of liquid volume of solvent to liquid volume of methanol necessary to saturate a given empty volume. Table 2 FAAS operating conditions Wavelength 279.5 nm Slit-width 0.2 nm Lamp intensity 35 mA Observation height 8.0 mm CZH2 flow 2.7 1 rnin-' Total air flow 19.6 1 min-' Nebulizer air flow (Q,) 5.6 1 min-' Liquid flow (Q,) 4.5 ml min-I Integration time 5 s diffraction (Malvern Instruments 2600~).All of the measurements were made with a lens of 100 mm focal length which encompasses a droplet diameter range of between 1.9 and 188 pm. The software employed was version M5.4 and the calculations were made in 'model independent' mode which does not presume any particular function for the distribution as has been recommended by Jackson and Sam~elson,~~ and is therefore the most appropriate mode for multimodal distributions or for those containing a large percentage of fine particles. The system used for drop size distribution determination has the inherent advantage of being non-intrusive to the aerosol. The accuracy is kd0/o for volume median diameter as reported by the manufacturer.'* Results and Discussion Pure Solvent Drop size distribution of the primary aerosol The results obtained for the drop size distributions of the six solvents are shown in Table 3.The description of a distribution assumed to be approximately log-normal requires two parameters at least one for the location of the centre of the distribution and another for the characteriza- tion of its width. In this study Dso (the volume median diameter) and the span have been employed to this end. The parameter of DsO was chosen for the central tendency because it corresponds to the maximum of the log-normal volume distribution curve. The first point to be recalled is that volume concentration (VC) values are not the same for all solvents even though Q (liquid flow) is the same (4.5 ml min-l).There are two reasons for this. The first is solvent evaporation which allows a small amount of liquid volume to be lost in the measurement area. The second is the presence of droplets smaller than 1.9 pm which is a smaller diameter than can be measured by the particle sizer as the photons diffracted by these droplets fall out of the detector. As the liquid fraction contributed by these droplets is always relatively small and can be estimated by the particle sizer through the obscuration (OB) value (see Table 3) it seems clear that the differences in the VC values fairly noticeable in some instances are mainly due to evaporation. Closely related to this the sequence of the VC values is exactly opposite to that of the relative volatility (RV) values viz.VC Benzaldehyde> water> acetic acid-propan- l-ol RV Benzaldehydetwater<acetic acid=propan- l-ol =formic acid>methanol -formic acid(methano1 Table 3 Distribution parameters for pure solvents v c * DSO$l V,.~II V1.9fl SSAII/ Solvent (O/O) OBt pm Spang (%) (Oh) m2 ml-3 Water 0.0034 0.1464 16.9 3.7 7.0 4.1 0.7105 Benzaldehyde 0.0038 0.2072 12.1 3.1 9.6 4.8 0.9127 Formic acid 0.0026 0.1578 12.9 3.8 12.3 7.1 1.0145 Acetic acid 0.0026 0.1693 9.8 3.4 12.5 6.5 1.0916 Propan- 1-01 0.0028 0.1761 10.0 2.2 10.6 5.4 1.0272 Methanol 0.0018 0.1607 7.7 3.0 20.1 11.1 1.4744 * VC=Volume concentration. This value indicates the ratio of total volume of liquid aerosol contained in the measurement volume to t OB=Obscuration.This value indicates the proportion of incident light that is being scattered out of the beam by the aerosol. $ DSo=Droplet distribution diameter below which 50% of the cumulative aerosol volume is found. Hence D90 and DlO=90 and lo% 8 Distribution span [(D90-Dlo)/Dso]. A measurement of the distribution spread. 1 V3.0 and V,.,=Percentage of volume found in droplets smaller than 3.0 and 1.9 pm respectively. 11 SSA = Specific surface area of the aerosol. the total measurement volume (1 0 mm diameter x 14.3 mm beam length active). ~espectivel y .JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 575 It should be noted that solvent evaporation proceeds at a particularly rapid rate during the first moments in the life of the aerosol9 for several reasons (i) the liquid surface is enormous; (ii) the air layer surrounding a given droplet is not saturated and is being continuously renewed as the droplets decelerate more slowly than the gas at the beginning of their path because of their inertia; and (iii) the vapour pressure of a given liquid increases with decreasing curvature radius of the surface.'O The differences in VC values from one solvent to another have an immediate effect on the OB values.If two solvents of equal VC value are compared the OB values should increase as D50 decreases because two droplets intercept a greater fraction of the laser beam than does a single droplet the volume of which is the sum of the volumes of the first two droplets. Likewise for two aerosols of equal distribu- tion the one with the higher VC value will show a higher OB value.In good agreement with this benzaldehyde is the solvent with the highest OB value whereas water in spite of its relatively high VC value shows the lowest OB value because the distribution is the coarsest (a higher DsO value i.e. 16.9 pm). Formic acid acetic acid and propan-1-01 have similar OB values as their distributions and their VC values are also similar. The OB value for methanol is similar to those of the last three solvents because the VC is fairly low while the distribution is finer (a lower Dso value i.e. 7.7 pm). The characteristics of the drop size distribution of the primary aerosol and hence the Ds0 and span values are mainly determined by the surface tension (a) and viscosity (J) of the l i q ~ i d ~ - ~ but solvent evaporation and droplet coalescence tend to modify them from the very beginning evaporation in terms of diminishing the Ds0 and increasing the span and coalescence in terms of producing the opposite results9 The graphs of the experimental results reveal this situation effectively.Water which is the solvent with the highest c value by far appears to be the one with the highest Ds0 and lowest specific surface area of the aerosol (SSA) even though its viscosity is fairly low. The influence of viscosity becomes clearer on comparing methanol and propan-1-01 shown in Table 3 as their a values are similar while their J values are clearly different The D50 value for methanol is 7.7 pm whereas for propan-1-01 is 10.0 pm.A comparison between benzaldehyde and acetic acid reveals the influence of surface tension as the viscosities are similar. Increasing solvent viscosity enhances the capability of the solvent to damp the oscillations that appear on the surface during nebulization leading in turn to a lengthening of the liquid vein before the collapse of the droplets. This effect results in an increase in the mean drop size of the spray.2 The influence of the surface tension on the nebulization can be described as follows:19 the energy available for the pneumatic generation of an aerosol comes solely from the kinetic energies of the gas (and liquid) streams. This energy is partly employed in the formation of the new surface in addition to being used in the acceleration of the liquid stream and the resulting droplets.The energy employed for the formation of the new surface is proportional to the aerosol surface area and also to the surface tension of the solvent. In a series of experiments where the kinetic energies of the gas and liquid streams remain roughly unchanged as in this instance the energy available for surface formation should be also unchanged. Thus solvents with the lowest values for surface tension should generate more surface. This is accomplished through the production of a finer aerosol. From Table 3 it can be seen that a discussion of some of the discrepancies that arise on comparison of the values of Djo and SSA is necessary. Two distributions of equal span but different Ds0 should have different SSA the smaller the value for D the greater the SSA (e.g.water and formic acid). The situation is not so obvious if the span values are not the same. It might then be that a distribution with a smaller D50 shows a lower SSA (e.g. benzaldehyde and formic acid). Formic acid provides a higher fraction of the total volume enclosed in small droplets than does benzal- dehyde in spite of its larger D50 value as its span is also higher [see Fig. l(a)]. Something similar happens between formic acid and propan- 1-01 [see Fig. 1 (b)]. Whereas the D50 value is larger for formic acid the SSA values are similar because of the noticeable differences between their span values. As the D50 values for each of the solvents studied depend on o and J simultaneously it is not unlikely that Dso could be linearly related to a combination of both physical magnitudes. There is good correlation between Dso and the magnitude x = s x a+ J where s is a parameter the value of which is around 0.1 (provided that a is written in N m-' x lo3 and J i n N s m-2 x lo3) as shown in Fig.2. The correlation between D and x is much better than with CT or J separately. Transport eficiency If the spray chamber. is considered as a type of filter with a given cut-off diameter (dc) droplets smaller than d will succeed in reaching the flame whereas larger droplets will drain away. Obviously this description of the chamber action is simple and the tertiary aerosol will always contain some droplets larger than d and some droplets smaller than d will have drained away before reaching the flame.6 However this simplified mechanism can be retained as it is useful for the following discussion on the relationship between the drop size distribution of the primary aerosol and transport rate.The amount of analyte and solvent transported to the flame ( W, and S,, respectively) along the spray chamber is likely to depend on the characteristics of the primary aerosol (mainly D50 and span) and on solvent volatility. Among the factors that modify the aerosol distribution 100 (a) dip m Fig. 1 and B benzaldehyde; and (b) A formic acid; and B propan-1-01 Cumulative drop size distribution for (a) A formic acid;576 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1991 VOL. 6 Table 4 Transport and signal results for pure solvents &I*/ ESt ~ 1 0 N &"§ Solvent ml min-I (%o) pg min-I (Yo) A,! A,," Water 0.4 1 9.1 0.40 4.5 1.00 2.5 Benzaldeh yde 0.43 9.6 0.44 4.9 1.01 2.3 Formic acid 1.42 31.6 1.61 17.9 2.64 1.6 Acetic acid 1.52 33.8 1.85 20.6 3.91 2.1 Propan- 1-01 1.28 28.4 1.75 19.4 3.49 2.0 Methanol 2.22 49.3 2.79 31.0 4.67 1.7 * S, = Solvent transport rate. t &,= Solvent transport efficiency.$ W, = Analyte transport rate. 9 E = Analyte transport efficiency. fi A,= Relative absorbance (A,/AHBtcr). 11 A,=Ratio between relative absorbance and analyte transport rate (Ar/ WloJ. 5 L " " " ' 2 3 4 5 6 7 8 9 X Fig. 2 Variation of D verms x (a linear combination of surface tension and viscosity) for pure solvents. x=O.l xa+J a in N m-'x lo3 and J in N s m-2x lo3. Correlation line D,,=(2.6 +_(IS)+( 1.74&00.10) X; r2=0.987 along its way (impaction settling coalescence solvent volatility etc.) volatility shows a peculiar aspect in that it modifies the analyte and the solvent distribution curve in different ways (Fig.3). Thus the analyte concentration gradients being a function of the droplet size arise as a result of solvent evaporation i.e. after a given period of time the analyte concentration in a droplet will be higher as the droplet is smaller. As the relative rate of solvent loss increases as the droplet size decreases the effect of evaporation on the DsO value is to increase it in the solvent distribution curve and to diminish it in the analyte distribution curve. On the other hand the effect on the span values is to reduce them in the solvent distribution curve and to increase them in the analyte distribution curve.Hence assuming that only V3.0 for instance will reach the flame (in other words the dc of the spray chamber is 3.0 pm) then two solvents of equal initial size distribution (equal D50 and span) will transport different amounts of analyte to the flame if their RV values are different. The fraction of analyte transported to the flame will always be greater than the fraction of liquid volume transferred into the flame (see Fig. 3). The difference between these two fractions increases on increasing solvent volatility. The value of W, will also be enhanced on increasing solvent volatility as shown in Fig. 3. This is approximately the situation between benzaldehyde and formic acid (see Table 4). The comparison is somewhat different for solvents with different distributions.If two solvents show similar values for Dso and RV then that with the greater span will have a higher transport rate for both solvent and analyte. This is Log d- Fig. 3 Ideal distribution curves (a) for solvent; and (b) for analyte. A Initial distribution curve before evaporation; B distribution curve after evaporation. For (a) the area under curve A corresponds to 100% of the solvent. The area between curves A and B corresponds to the amount of solvent evaporated. For (b) the area under curve A=area under curve B= 100% ofthe analyte. The cut- off diameter of the spray chamber 3 pm is indicated the situation between propan- 1-01 and acetic acid. Acetic acid because of its greater span contributes a larger V3.0 as shown in Fig. 4.Consequently higher W, and S, values can be expected for acetic acid than for propan-1-01. For two solvents with equal RV and span values and different Dso values the solvent with the smaller D50 value will be the one with higher W, and S, values not only because it contains a greater fraction of liquid volume below the d but also because its evaporation is more rapid. Some predictions about the transport of analyte and solvent can be made on the basis of the above discussion and the distributions obtained for pure solvents (Table 3). Methanol is likely to be the most efficient solvent in terms of transport as it shows the highest values of V and RV (this is why it shows the lowest VC value). Water and benzaldehyde should be placed at the other end as they show the smallest values in these magnitudes.Water shows the lowest V3.0 followed by benzaldehyde which is the least volatile followed by water (this explains the fact thatJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 577 100 - 5 50 3 0 > - a b " 1 10 100 d h m 1000 Fig. 4 Cumulative drop size distribution for A acetic acid and B propan-1-01 showing the effect of their different span values on V3.0; a= 12.5% b= 10.6% benzaldehyde is the solvent with the higher VC followed by water). The rest of the solvents will probably lie in the intermediate zone relative to their transport efficiency. It can be seen in Table 4 that these predictions hold reasonably well in practice for all of the solvents investi- gated. Only W, for formic acid appears to be somewhat lower than expected (but not the S,,,).However there is a certain lack of precision associated with the indirect methods of measuring transport efficiency as quoted by Smith and Browner.2o The analyte concentration in the drained away fractions increases on increasing solvent RV from approximately 2.1 pg ml-1 for benzaldehyde to 2.7 pg ml-l for methanol. A mass balance for analyte and solvent allows the estimation with the uncertainty associated with these calculations that the solvent load to the flame in vapour form (S,) varies from 0.2 ml min-l for water and benzaldehyde to 1.2 ml min-' for methanol these volumes being expressed as liquid volumes. Analytical signal The analytical signal depends mainly on the analyte transport rate and the atomization efficiency. The analyte transport rate is determined by the drop size distribution of the primary aerosol and the transport phenomena along the spray chamber.The atomization efficiency depends on the size distribution of the tertiary aerosol the drop nature and composition of the flame and also on the element to be determined. A controversy exists in the literature on the changes induced in the flame and in the atomization efficiency when the analyte is introduced using an organic solvent. Some workers14y21722 allude to variations (increases and decreases) in the flame temperature whereas others indicate variations in the composition of the flame23 or in its dimen~ions.~~ All of these factors will affect the atomi- zation efficiency and hence analytical signal on changing the solvent nature.However W, will be the dominant influence on the analytical signal as Mn is relatively insensitive to variations in the nature and composition of the flame. A clear parallelism between the W, values and the absorbance values can be seen in Table 4. Nevertheless the atomization efficiency decreases with increasing transport rate. Thus on going from water to methanol W, increases by a factor of 7.0 whereas the signal increases by only 4.7. As Mn is only slightly sensitive to variations in the temperature and reducing character of the flame and the drop size distributions of tertiary aerosols are expected to be ~imilar,~ the decrease in the atomization efficiency might come from two possible sources ( i ) a decrease in the slope of the analytical curve due to the increase of the analyte concentration in the flame; and (ii) a dilution of the analyte in the flame as a result of variations in the geometry as organic solvents can cause flame enlargements.This second explanation seems reasonable. If 7.50 9.40 and 10.70 cm2 are used for the cross-sectional areas of the flame with water propan- 1-01 and methanol24 as solvents respectively the resulting values for the analyte transport rate per unit of cross-sectional area (qi) are 0.0533 0.186 and 0.261 and the relative values (qiJ 1.00 3.49 and 4.90 respectively. These values are in better agreement with the relative absorbances (1 .OO 3.49 and 4.67) than those of relative W, (1.00 4.38 and 6.97). In order to observe this behaviour it is necessary that the hollow cathode lamp beam width at the analyte wavelength be smaller than the flame width.The cross-sectional area of the flame using methanol as the solvent is taken as 10.70 cm2 which is the value corre- sponding to the use of ethanol as reported by Sziv6s et al.,24 taking into account that the behaviour is very similar and that they did not investigate methanol. Methanol-Water Mixtures Drop size distribution of the primary aerosol Table 5 summarizes the most relevant parameters of the drop size distributions of the methanol-water mixtures. Firstly it can be noticed that although the and VC values for water in Table 3 (pure solvents) agree with those in Table 5 (mixtures) this situation does not hold for methanol the values being higher in Table 5 than in Table 3.Secondly the span values are generally higher in Table 5 than in Table 3. The discrepancies between the two distributions for methanol may be assigned to several causes; the most significant being the period of time that elapsed (about two months from September to December) between the series of measurements. During this time room tempera- ture undoubtedly dropped (this variable was not moni- tored) and the nebulizing air temperature and solutions temperature probably dropped. This could cause the mean drop size and VC of the first series (higher tempera- ture) to be lower than those of the second series (these differences increase with increasing solvent RV). The increase of the span value may also be related to this temperature difference but it is difficult for an adjustable nebulizer to be in the same geometrical placement exactly for such long periods of time.It is not surprising there- fore that different results were obtained for methanoL It can be stated that the results obtained within a given series of measurements are strictly comparable. For dif- ferent series of measurements the tendencies are compar- able but not the absolute values. In Table 5 it can be seen that the VC values diminish at a steady rate as the proportion of methanol increases owing to the enhancement of the solvent volatility. The Dso value also decreases with increasing methanol content. These two factors size decrease and RV increase have an opposite influence on the OB value. The experimental results show that the decrease in drop size is the dominant effect and so the OB increases until a maximum is reached for the water-methanol mixture (90+ lo) followed by a slight decrease for pure methanol ( 100%).Also in this series the D50 value is controlled by surface tension and viscosity. Surface tension decreases markedly with increasing methanol content whereas viscosity in- creases until it reaches a maximum for the water-ethanol mixture (50+ 50) and then decreases steeply. Once again in this series the Ds0 values correlate much better with a linear combination of surface tension and viscosity than with578 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1991 VOL. 6 Table 5 Distribution parameters for methanol-water (vh) mixtures Solvent vc Methanol (O/O) Water (O/O) (O/o) OB 0 10 25 50 60 80 90 100 100 90 75 50 40 20 10 0 0.0036 0.0038 0.0039 0.0037 0.0036 0.0034 0.0034 0.0030 0.1434 0.1687 0.1854 0.1971 0.2000 0.2056 0.2 190 0.2094 D5d v3.0 vl.9 pm Span (Yo) (To) 17.0 5.8 7.0 4.6 16.5 5.9 7.8 4.9 15.5 6.4 8.6 5.3 13.8 5.3 10.4 6.3 13.2 5.3 11.1 6.8 11.9 4.9 12.7 7.7 10.9 4.9 13.6 7.7 9.7 5.1 15.4 8.5 SSAI m2 ml-3 0.7092 0.7503 0.8020 0.91 11 0.9538 1.0519 1.1013 1.1966 Table 6 Transport and signal results for methanol-water (v/v) mixtures Solvent &Ot/ Es Methanol (O/O) Water (O/O) ml min-' (O/o) 0 10 25 50 60 80 90 100 100 90 75 50 40 20 10 0 0.4 1 0.52 0.64 1.06 1.42 1.77 2.00 2.26 9.1 11.2 14.1 23.0 28.7 36.3 40.9 50.2 W,O,/ pgmin-1 (2) A A 0.4 1 4.5 1.00 2.4 0.48 5.1 1.35 2.8 0.57 6.2 1.78 3.1 0.89 9.5 2.21 2.5 1.56 15.7 2.57 1.7 2.24 23.0 3.32 1.5 2.56 26.1 3.97 1.6 2.82 31.3 4.87 1.7 either of them individually (Fig.5). The discussion of drop size distribution of the primary aerosol under Pure Solvent is also valid for solvent mixtures. The span values in the series using methanol-water mixtures are higher than in the series using pure solvents as has already been mentioned but the variations are not significant. Transport eflciency As the D50 value for the aerosol decreases and the volatility increases the loss of analyte in the spray chamber will decrease with increasing methanol content of the mixture. Table 6 shows that both the W, and S, values increase with increasing methanol content in addition to the transport efficiencies E and e,. However the increase in the transport efficiency is slower at the beginning of the series [from water to mixture (50+50)] than at the end [from mixture (50+ 50) to methanol] probably because the RV 5 15 -.0 a" 10 2 3 4 5 6 7 8 9 X Fig. 5 Variation of Dz0 versus x (a linear combination of surface tension and viscosity) for methanol-water (v/v) mixtures. x=O. 1 x a+J a in N m-l x lo3 and J in N s m-2 x lo3. Correlation line D5,=(6.1 ?0.6)+(1.41 kO.10) x; 9=0.971 increases in a similar way. It is evident that the increase in transport efficiency from the beginning to the end of the series is far greater than the increase in V3.0 values. This behaviour can be justified by taking into account that V3.0 values were measured close to the nebulizer tip i.e. a short time after the droplets were generated whereas the effici- ency measurement includes the much larger evaporation time associated with the passage of droplets through the spray chamber.When using solvent mixtures the vapour phase should become enriched in the more volatile component i.e. methanol. This makes the composition and density of the drained fraction of the solvent dissimilar to the fraction fed to the flame except for pure solvents e.g. methanol (100%) and water (1 00%). The composition and density of both fractions will not be the same as that of the initial mixture. These circumstances that were not taken into considera- tion on making the calculations will certainly introduce a given level of inaccuracy in the S, values,20 which were determined by an indirect method.For the same reasons the solvent load to the flame in vapour form (S,) and the solvent load to the flame in liquid form (S,) will not have the same composition either. Both fractions should increase with increasing proportions of methanol; S because of the RV increase and S because of the decrease in D50 values. Analytical signa I The values given in Table 6 show that the absorbance increases with increasing W,,,. However the increase in relative absorbance (about %fold) is lower than the corresponding relative transport increase (about 7-fold). Once again this divergence is more pronounced along the second part of the series (increased methanol) when Stot becomes important making W, less efficient for the production of the signal as seen in the last column of Table 6.The widening of the flame because of the solvent load might be one of the reasons for this behaviour as discussed for analytical signal under Pure Solvent. This explanation agrees with that of Gustavsson2s concerning the effect ofJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 579 analyte transport on inductively coupled plasma emission intensity. Conclusions The organic solvents used in this work both in pure form and in mixtures with water result in a sensitivity enhance- ment compared with the use of water alone. This enhance- ment is due to an improvement in the analyte transport which in turn is the result of a finer primary aerosol and also higher solvent volatility. Surface tension together with viscosity determine the volume median diameter of the primary aerosol formed but the span is not easily correlated with the solvent physical properties.Both distribution parameters are im- portant for transport processes. The spray chamber can be considered as a sort of ‘filter’ with a given cut-off diameter for a given set of gas and liquid flows. Solvent volatility greatly alters the initial drop size distribution thus allowing an increased fraction of the analyte to be carried to the flame. However the flame widening caused by the high solvent load does not increase the absorbance as much as expected from transport enhancement. This fact counter- balances a significant part of the beneficial effect of the improved analyte transport on the sensitivity. Nevertheless the effect of the solvent load in FAAS seems not to be so critical as in ICP-AES.In order to achieve a higher sensitivity in FAAS solvents with low surface tension and viscosity and high volatility should be used. A desolvation system would also be advisable to avoid the-effect of solvent load although this point has not been tested. The Comisidn Interministerial de Ciencia y Tecnologia (CICYT Spain) is acknowledged for financial support (Grant No. PB88-0288). J. M. also expresses his apprecia- tion to the Instituto de Estudios Juan Gil Albert (Diputa- cion Provincial de Alicante Spain) for a scholarship. References 1 Nukiyama S. and Tanasawa Y. in Experiments on the Atomization of Liquids in an Air Stream tr. Hope E. Defense Research Board Department of National Defense Ottawa Ontario Canada 1950.2 Sharp B. L. J. Anal. At. Spectrom. 1988 3 613. 3 Canals A. Wagner J. Browner R. F. and Hemandis V. Spectrochim. Acta Part B 1988 43 132 1. 4 Canals A. Hernandis V. and Browner R. F. J. Anal. At. Spectrom. 1990 5 61. 5 Robles C. Mora J. and Canals A. unpublished work. 6 Gustavsson A. Anal. Chem. 1983 55 94. 7 Gustavsson A. Anal. Chem. 1984 56 815. 8 Sharp B. L. J. Anal. At. Spectrom. 1988 3 939. 9 Canals A. Hemandis V. and Browner R. F. Spectrochim. Acta Part B 1990 45 591. 10 Cresser M. S. Prog. Anal. At. Spectrosc. 1982 5 35. 11 Feldman F. J. Bosshart R. E. and Christian G. D. Anal. Chem. 1967 39 1175. 12 Lemonds A. J. and McClellan B. E. Anal. Chem. 1973 45 1455. 13 Attiyat A. S. Microchem. J. 1987 36 228. 14 Allan J. E. Spectrochim. Acta 1961 17 467. 15 Farino J. and Brbwner R. F. Anal. Chem. 1984 56 2709. 16 Mora J. Canals A. and Hemandis V. J. Anal. At. Spectrom. 1991 6 139. 17 Jackson T. A. and Samuelsen G. S. Proc. Photo-Opt. Znstrum. Eng. 1985 573 73. 18 Malvern Instruments Particle Sizer Reference Manual. Ver- sion 5.4. October 4th 1987. 19 Faske A. J. Ph.D. Thesis Georgia Institute of Technology Atlanta GA USA 1986 p. 84. 20 Smith D. D. and Browner R. F. Anal. Chem. 1982,54,533. 21 Robinson J. W. Anal. Chim. Acta 1960 23 479. 22 Avni R. and Alkemade C. Th. J. Mikrochim. Acta 1960 3 460. 23 Harrison W. W. and Juliano P. O. Anal. Chem. 1969 41 1016. 24 Szivos K. Pungor E. and Kiss L. Talanta 1979 26 849. 25 Gustavsson A. in Inductively Coupled Plasmas in Analytical Atomic Spectrometry ed. Montaser A. and Golightly D. W. VCH Weinheim 1987 p. 4 19. Paper 1 /006 15K Received February I1 th 1991 Accepted May 23rd I991