Mass size distributions of traffic aerosols at Vienna - ACS Publications

Matti Happonen , Fanni Mylläri , Panu Karjalainen , Anna Frey , Sanna ... and Risto E. Hillamo , Juha K. Vilhunen , Leena Rantanen , Nicole Havers, A...
0 downloads 0 Views 673KB Size
J. Phys. Chem. 1980, 84, 2079-2083

2079

Mass Size Distributions of Traffic Aerosols at Vienna Axel Berner* and Christlan Lurrer University of Vienna, Institute for Experimental Physics, A 1090 Vienna, Austria (Received: October 9, 1979)

The mass size distributions of aerosols can be measured directly with low-pressurecascade impactors. Because of the wide span from 0.06 to 16 pm of aerodynamic equivalent diameter, the accumulation mode and part of the nucleation and coarse mode fall into the measuring range of the impactor. Urban aerosols have been sampled near a traffic road by three identical impactors at different heights above street level, and the structure of the mass size distributions and their change with increasing height has been investigated.

Introduction A set of three logarithmic normal size distributions has been introduced by Whitby and co-workers in order to represent the size distribution of atmospheric aerosols. The corresponding modes, i.e., the nucleation mode, the accumulation mode, and the coarse mode, reflect certain 'aspects of the genesis and the formation of atmospheric aerosols. The nucleation mode has modal sizes from 0.015 to 0.04 pm, with respect to volume size distributions, while those for the accumulation and coarse modes fall into the ranges from 0.15 to 0.5 pm and from 5 to 30 pm, respectively. The corresponding standard deviations are u = 1.6 for the nucleation mode, u = 1.6-2.2 for the accumulation mode, and u := 2-3 for the coarse rn0de.l The trimodal model is a fairly simple way to describe the structure of atmospheric aerosols. It should be mentioned, however, that the surface and volume size distributions which are used with the model are not measured directly, but are inferred from number size distributions measured by electrical mobility analyzers and optical particle counters. Most of these distributions do not show three distinct modes but rather the accumulation mode, and measurements have to be very accurate in order to justify a transformation of number size distributions into surface and volume size distributions. As has been stated,l "there is an overwhelming amount of evidence" for the modal structure of m,msand volume size distributions, and more information on the modal structure comes from mass size distributions of aerosol components isuch as sulfates.2 However, the direct way of measuring mass size distributions by cascade impactors has not been attempted very often. Recently, special impactors have been developed for measuring mass size distributions to a high degree of a accuracy. In the impactors described by Chuan3 the particles are collected directly on quartz crystal balances built into the impactor stages. A sensitivity of 7 x lo8 Hz/g is reported, and a sufficient amount of material for mass size distributions is collected in short periods of time. These impactors have a wide working range spanning from 0.06 pm of aerodynarnic equivalent diameter up to 50 and 77 pm, respectively, thereby registrating the modes of atmospheric mass size distributions almoe t completely. In the impactors descriibed by Berr~er,"~ the particles are collected on thin foils. A sensitivity o f 1 pg has been achieved, justifying the use of electronic microbalances for the evaluation of the samples.6 The Berner impactors facilitate chemical analysis of the deposits, because the material of the foils can be matched to the analytical techniques. The usefulness of these impactors has been demonstrated by P u ~ b a u m . ~ , ~ Measurements of urban aerosols by the low-pressure impactor have been performed e a ~ d i e r , but, ~ . ~ as these 0022-3654/80/2084-2079$01 .OO/O

aerosols have been sampled in an enclosed backyard of a five-story building, the mass size distributions exhibited a very strong accumulation mode with almost no material in the nucleation mode and coarse mode size ranges. Traffic aerosols have been reported to contain higher amounts of nucleation mode and coarse mode material, therefore measurements of traffic aerosols have been performed in order to investigate the modal structure by means of the low-pressure impactor.

Experimental Section A special low-pressure impactor has been developed for sensitive measurements of mass size distributions. The impactor has a flow rate of 30 L/min and a measuring range from 0.06 to 16 pm of aerodynamic equivalent diameter. The impactor has nine stages, and the cutoff points of subsequent stages differ by a factor of 2, dividing the size range into eight logarithmic equally sized intervals. Besides the entrance stage, which has a single orifice in order to precipitate particles larger than 16 pm, each stage has several equal size orifices lined up equidistantly on a circle centered to the axis of the stage (Figure 1). This arrangement warrants symmetric flow patterns and symmetric particle deposition, and therefore the deposition spots beneath the orifices contain equal amounts of mass, within the limits of a few percent of the average spot mass.1° The aerosol jet coming from the orifices are deflected by the stagnation plate, which has a large center hole for the flow to pass to the subsequent stage. This design which has already been described earlier"-13 is advantageous in sofar as particles which are carried away from their location of deposition are not transferred into the deposits of the subsequent stages. (This fact is demonstrated by Figure 2, which shows two mass size distributions of the same aerosol. Two impactors have been operated in parallel, and the particles have been collected on greased aluminum foils in one of the impactors and on dry aluminum foils in the other one. The deposits have been evaluated on a balance. Evidently a certain amount of coarse particles is missing when dry foils are used, but they do not occur in the finer fractions, as the (unnormalized) mass size distributions are identical in the size range below 0.70 pm of aerodynamical equivalent diameter). The distance between the orifice and stagnation plates is made by a removable circular spacer. A critical orifice, which is placed behind the final impactor stage, controls the total volume flow rate and keeps the impactor flow free from fluctuations. The deposits are collected on thin foils which line the stagnation plates completely and which are held in place by the spacer. For the purpose of gravimetric analysis, aluminum foils of 10 pm thickness have been used because of the low blank weight of 70 mg, the weight stability, and 0 1980 American Chemical Society

2080

The Journal of Physical Chemistry, Vol. 84, No. 16, 1980

Berner and Lurzer Ili9 ?/stage]

60

P

Im

50 LO

30

20 10

Flgure 1. Impactor stage (partly dissected in order to show the arrangement of the stage elements). The collection foil not shown here covers the stagnation plate completely and is held in place by the spacer: P", stagnation plate; S, spacer; OP, orifice plate. 0

1

I

greased foils

o dry foils

71

50

i

?Irn

, 11

b AED 09

17

35 70 14 28 56 113 Lpml

Flgure 2. Mass size distributions of an aerosol measured by two identical Impactors. One of the impactors has been charged with dry aluminum foils, the other one with greased foils (Apiezon grease). The diameters are aerodynamic equivalent diameters.

the ease of handling. For the measurements, those foils which collect particles of 2 pm diameter and larger have been coated with a thick layer of Apiezon grease in order to prevent the particles from bouncing and blowing off. The other foils are left blank. The deposits are weighed on an electronic microbalance with a sensitivity of 1pg (Mettler Me 30). The total errors of weighing the samples amount to 2-3 pg, justifying the use of the sensitive balance. The series of measurements reported here has been performed in order to investigate the variation of the urban aerosols at different heights above ground. Three impactors have been mounted in the windows of a building at levels of 1, 9, and 20 m above ground. The windows are facing a street with heavy automotive traffic. The impactors have been operated for 23.5-h periods, from 8.30 a.m. to 8-00a.m. of the next day. A series of 20 samples was collected in February, 1979. Results The distributions represented in Figure 3 are typical with respect to the structure of the urban mass size distributions and their variation with increasing height at the measuring site. The accumulation mode appears very distinctly at sizes around 0.6 pm (aerodynamic equivalent diameter). The distribution is definitly asymmetric due

11

12

I

l

5

1

l

l

1

'

1

l

2

1

I

5

i

, '

I

101

20 [ p m ~ diameter

Flgure 4. A set of mass size distributions measured by the impactors at high relative humidity of the air. The diameters are aerodynamic equivalent diameters.

to the minor, but noticable, contributions of the nucleation mode in the size range around 0.1 pm. The gap at 2 pm separates the accumulation mode from the coarse mode, where the latter is incomplete due to the upper limit of the measuring range of the impactor. Within the series of measurements this basic form undergoes certain variations. In a case of extremely high relative humidity the accumulation mode occurred at a modal size of 1pm (Figure 4). Numerical parameters of the modes of the mass size distributions have been obtained by means of a mathematical model. The model assumes a logarithmic normal size distribution for each mode, where a distribution is given by

P,(xI)= (l/ML)exp(-(x,

- xJ2/2u,2)

The index j represents the stage number, and xI = In (D,/D,) represents the mean particle size of the stage. Do = 1 pm is the reference diameter. The index j runs from j = 1 for the finest fraction to j = 8 for the coarsest fraction. The index i indicates the mode, with i = 1 for the nucleation mode, i = 2 for the accumulation mode, and i = 3 for the coarse mode. The probability density, P,(X.~), is characterized by the modal size zl of the mode i, Its standard deviation uI, and the normalizing factor, M,, which represents the total mass of the mode. Finally, the value of P,(x,) is the mass contribution of mode i to the mass collected on stage j . Then CPL(xl) is the total mass collected on stage j . It should be noted that the data (ML, R,, a,) of the modes follow from the analysis of the measured mass size distributions and consequently the sole

The Journal of Physical Chemistry, Vol. 84, No. 16, 1980

Mass Size Distrlbutions of Traffic Aerosols

2081

TABLE I: Mass Concentrations (pg/m3)at Different Heights

c ( T 0 T ) z(NM)

x(AM)

z(CM)

mean max min std dev

1 m above Street Level 15.8 68 3116 20 97 445 8 24 95 f3.1 + 30 +74

242 361 50 i 74

mean max min std dev

9 m above Street Level 63 210 10.8 15 112 303 84 4 21 i 5:1 + 2.9 t 27

140 227 40 f 50

mean max min std dev

20 m above Street Level 6.9 67 175 13 120 22'7 8!3 3 29 f 2.1 + 26 f 3'7

102 150 40 i 30

I

TABLE 11: Nodal Sizes at Different Heights

mean

1 m above Street Level AM CM DZ D, 0.59 him 1.68 1 3 pm

%3 2.8

mean

9 m above Street Level AM CM D* D, 0.59 pm 1.80 12pm

%3 2.9

mean

20 ni above Street Level AM CM DZ sg2 D3 0.61 p,m 1.79 12pm

%3 3.2

bias of these data is the assumption of log normal size distributions for the modes. (Parts of this model have been presented at the 1979 GAF conference at Dusseldorf.6 The model itself is part of a thesis, which will be published soon.) The average mass (concentrationsof the modes, together with the extreme concentrations and the standard deviations of the ensembles, are listed in Table I, where the numerical values represent the mass concentrations within the measuring range of the impactor. As demonstrated by the data, the total mass concentrations (C(T0T))of the aerosols are decreasing with increasing height, from an average of 316 pg/m3 at the 1-m level, to 210 pg/m3 at the 9-m level, and to 175 pg/m3 at 20-m level. In detail, the average mass concentrations of the accumulation mode (C(AM)) is almost independent of the height, whereas the average mass concentrations of the nucleation mode (1(NM)) and the coarse mode (C(CM)) decrease considerably. The correlation coefficients show more clearly the interdependence of the mass concentrations. The correlation coefficient for the mass of the nucleation mode and the accumulation mode is CNA = 0.07, and the coefficient for the accumulation mode-coarse mode correlation is CAC = -0.06. Consequently there is no interdependence between the accumulation mode and the other modes. On the other hand, the coefficient for the correlation of the nucleation mode and coarse mode is CNC = 0.47, indicating a weak, but certain correlation between the mass concentrations of these modes. Table I1 represents the average modal data, i.e., the modal diameters, D, = Do exp(Z,), and the geometric standard deviations, ug,&= exp(uJ, of the log normal distributions for the accumulation and coarse mode. The scattering range of the individual data, Le., the geometric ~ , represented also. The standard deviation of D, and u ~ ,are

Flgure 5. Graphical representation of accumulation mode and coarse mode data showing average values and ranges of modal diameters and modal geometric standard deviations for different levels. The dlameters are aerodynamic equivalent diameters.

scattering range is larger than the difference of the average values for the accumulation mode and coarse mode data throughout, therefore the variations with height are less important in comparison to the variations between different samples (Figure 5). The nucleation mode data, which lack accuracy and are incomplete due to the limited measuring range of the impactor, are not incorporated into Table 11. However, the mathematical model indicates a modal diameter of 0.09 pm. This value is evidently larger than the geometric diameters for the nucleation mode,l but differences are to be expected as the impactor measures an aerodynamic equivalent diameter. For the conditions in the final impactor stage (the mean free path of the gas molecules is larger than the particle diameters), these diameters are , = ppDgeom,where pp is the density of the correlated by D particles. For a density of 1.8 g/cmS, the aerodynamic diameter of 0.09 pm would correspond to a diameter of 0.05 pm, which is close to other data,' in order to be seriously taken into account.

Discussion As shown by the averaged data of Table I as well as by the individual measurements represented in Figures 3 and 4, the concentrations of the accumulation mode are fairly independent of the height. This result indicates that most of the accumulation mode material does not stem from local sources, but is already contained in the air masses transported to the measuring site. This conclusion is further supported by the correlation coefficients which indicate independency of the accumulation mode from the nucleation and coarse modes. The geometric standard deviations of the accumulation modes fall into fairly narrow ranges with average values of 1.7-1.8 (see Figure 5). These values are compatible with other data.l This correspondence and the fact that the modal diameters are also narrowly distributed justify the assumption that the normalized accumulation mode mass size distributions are fairly uniform during the sampling periods of 23.5 h, at least in this investigation. This assumption implies that the accumulation modes of size distributions from subsequent short time measurements

2082

The Journal of Physical Chemistry, Vol. 84, No. 16, 1980

will also show a high degree of uniformity, with respect to modal diameters and standard deviations. The mass concentrations of the accumulation modes are the only parameters then which are subjected to major changes within days, as indicated by the data (Table I), or within hours as has been shown el~ewhere.~ The mass size distributions shown in Figure 4 are exceptional as the modal diameters occur at 1pm, approximately. In this case the aerosol was exposed to relative humidities of more than 90% during the measuring period, and the particles could have grown by uptake of water. This explanation is supported by the results of sulfate analysis performed on some of these samples, but unfortunately not on the sample considered here. The sulfate mass size distributions correspond almost perfectly to the aerosol mass size distributions of the accumulation mode and the sulfate content of this mode is in the order of 20-30% of the accumulation mode mass.14 Sulfate particles are hygroscopic and grow in atmospheres with more than 70% relative humidity. This explanation of the accumulation mode shift implies that the impactor is classifying the particles correctly in their wet state. This implication is the subject of further investigations. For 19 samples of this series, the model diameters and geometric standard deviations of the coarse mode fall into narrow ranges and, as is demonstrated by Figure 5, modal sizes and geometric standard deviations do not vary with height. Consequently, the normalized coarse mode mass size distributions are uniform at different heights. It is obvious then that sedimentation, which would shift the modal diameters to smaller values, does not play an important roll in removing particles near ground levels. (There is other evidence for this conclusion, as some of the measurements show almost identical mass size distributions for the 9- and l-m levels.) This statement which might not be true in general refers to the particles sizes within the measuring range of the impactor, to the sample set, and to the unspecified atmospheric conditions at the measuring site. The mass concentrations of the coarse modes decrease markedly with height and, as sedimentation would not account for the decrease, it is to be assumed that dilution of the highly polluted ground level air masses with less polluted urban background air is responsible for the observed variations. There is evidence that this background air measured at remote sites, as i.e. an enclosed backyard of a five-story building 200 m off a traffic road, does not contain considerable amounts of coarse part ice^.^^^ The dilution of the ground level air is also responsible, at least in part, for the fairly high correlation coefficients of the nucleation mode and coarse mode mass concentrations. The coarse mode data of one of the samples have not been included in Table 11. In this case the calculations of the modal data resulted in diameters of 38 and 28 pm at the 1-and 9-m level, respectively, and geometric standard deviations of 4.6 and 4.7. Such values should be considered with precaution, in spite of the fact that modal (geometric) diameters as high as 30 pm are rep0rted.l In case of the impactor measurements, such modal diameters as well as the main part of the modal size distributions are far ouhide of the measuring range, and direct measurements up to 50 pm or more would have been to performed in order to validate the calculated data. For comparing the impactor data with other results,1the aerodynamic equivalent diameters are to be converted into geometric diameters. Impactors classify particles according to their relaxation time, 7. For spherical particles the relaxation time is T = (D2ppCp)/(18qg),where D, is the

Berner and Lurzer

geometric diameters, p, the density, and C, the slip correction factor of the particles, and qg is the viscosity of the gas. The general form of the slip correction factor is

C, = 1 + (2hg/D,)(1.25 + 0.4 exp(-1.1(DP/2X,)))

(1)

Any two particles with identical relaxation times cannot be discriminated by the impactor, and consequently the relation

holds, if T, = T~ With po = 1 g/cm3, the diameter Do is called the aerodynamic equivalent diameter. It is the diameter of a sphere which represents all other particles with relaxation time T~~~ Three cases of eq 2 are to be considered. C = 1. The slip correction factors are unity, and the geometric and aerodynamic diameters are related by D, = ( p o / p )'/2Do, where the diameter conversion factor ( p o / p P ~ ' bis independent of size. In a logarithmic representation, the geometric and aerodynamic sizes differ by the constant term (1/2) In (po/p,), and consequently the modal size distributions are shifted without changing their shape. Therefore the standard deviations, u, as well as the geometric standard deviations, ug = exp(u), are equal in both size scales. The condition C = 1holds for the coarse mode and, assuming an average density of p, = 2.6 g/cm3, which is the average density for soil and building material, the modal diameters are related by Dp,3= 0.62D0,3

(3)

C = 1 + (2Xg/D)(1.6). This representation of the slip correction factor is valid for aerodynamic and geometric sizes around 0.5 pm, and accordingly applies to accumulation mode particles. Inserting the slip correction factors into eq 1 gives the equation DZ(1 + (3.2Xg/Dp))= Dz(1 + (3.2Xg/Do))(Po/Pp) and after some conversions D, = Do(po/pp)1/2(1+ (3.2Xg/D0)

+ (l.6hg/Do)2p,)1/21.6Xg

As for accumulation mode particles the term (l.6hg/Do).2pp is essentially smaller than 3.2Xg/D0,and the the equation reduces to D, = D ~ ( p o / p ~ ) ~+/ (~~( .l ~ X , / D O )-) 1.6hg ~/~ and, after developing the root to D, = Do(po/pp)1/2(1 + l.6X,/D0) - 1.6X, With an average density of p, = 1.8 g/cm3 and a mean free path of A, = 0.07 pm, the expression D, = 0.745(1 - 0.038/Do)Do (4) is found for the diameter relation. The diameter conversion factor, 0.745(1- o.038/D0), is size dependent, and therefore the modal size distributions will not preserve their shape under the diameter conversion. Consequently the standard deviations will assume different values in the different size scales. In order to estimate the magnitude of these differences two aerodynamic equivalent diameters, i.e., DO' = Do,2ug,2 and D{ = D0,2/ug,2, are converted. The diameter is here the average modal diameter of the accumulation mode, Do,z= 0.6 pm, and the geometric standard deviation, ~ r= ~1.8,, is~representative for the

J, Phys. Chem. 1980, 84, 2083-2084

geometric standard deviations of the accumulation mode. Consequently D 1, and, as the mean free path in the final impactor stage is about 0.25 pm, it holds for the particles of the nucleation mode. Moreover as 2.5Xg/D = 6.9 is fairly large in comparison to unity the slip correction factor may be further reduced to C = 2.5Xg/D. In this approximatioii eq 1 becomes

+

D, = Do(Po/Pp)

(5)

which has already been used earlier in this paper. The diameter conversion factor is independent of particle size, and consequently the diameters are shifted, but the standard deviations are not inflicted under this conversion.

2083

In summary, the modal diameters are shifted to smaller values under the diameter conversion, however, the standard deviations can be transferred without change. Assuming densities of 2.6 g/cm3 for the coarse mode particles and 1.8 g/cm3 for the accumulation mode particles, we find almost complete correspondence between the impactor data and data reported by Whitby.l Coarse Mode: The diameters were as,,follows: average, 7.4 pm, 8.1 pm; exceptional, 17.4 pm, 23.6 pm; average range, 6.8-8.4 pm; Whitby’s range, 5-30 pm. The geometric standard deviations were average, 2.8, 2.9, 3.2; average range, 2.3-3.8; Whitby’s range, 2-3. Accumulation Mode: The diameters were as follows: average, 0.43 pm, 0.44 pm; exceptional, 0.71 pm; average range, 0.37-0.46 pm; Whitby’s range, 0.15-0.5 pm. The geometric standard deviations were as follows: average, 1.68, 1.8, 1.79; average range, 1.4-2.2; Whitby’s range, 1.6-2.2.

References and Notes Whitby, K. T. Atmos. Environ. 1978, 72, 135-159. Friedlander, S.K. Atmos. Environ. 1978, 72, 187-195. Chuan, R. L. “Rapid Measurements of Particulate Size Distribution”, in “Find Particles”; Liu, 8. Y. H., Ed.; Academic Press: New York, 1976. Berner, A.; Lurzer, Ch.; Pohl, F.; Preining, 0.;Wagner, P. Scl. Total Environ. I n press. Berner, A. In “Aerosol Measurements in the Submicron Sire Range”; EPA-600/2-79-105; Washington, D.C., 1979. Lurrer, Ch. “Messung trimodaler Massengrossenvertellungenurbanen Aerosols mit Kaskadenlmpaktoren”, GAF-Konferenz 7, 1979. Puxbaum, H. Z. Anal. Cbem. In press. Puxbaum, H. Z. Anal. Cbem. In press. Lurzer, Ch.; Berner, A. J. Aerosol. Scl. 1979, 70, 231. Berner, A. Chem. Ing. Tech. 1978, 50, 399. Berner, A. ”Praktlsche Erfahrungen mA einem 20-Stufen-Impaktor”; Stabu-Reinh. Luft 32, 1972; p 315. Berner, A. “Dle Messung der Verteilungsfunktion von Stiiuben mHtels vielstufiger Kaskadenlmpaktoren”, EGKS-Report, 1970. Cohen, J. J. Am. Ind. Hyg. Assoc. J . 1967, 28, 95. Strasser, J. J. Aerosol. Sci. 1079, 10, 236.

COMMUNICATIONS TO THE EDITOR Comment on “Adsorption of Alcohols on Alumina. 1. Gravimetric and Infrared Spectroscopic Investigation”

Sir: Infrared studies of the adsorption of alcohols on alumina have shown that at least two types of chemisorbed species occur: alkoxide species, resulting from a dissociative chemisorption, and carboxylate species, formed at higher temperatures. Knozinger et a1.l consider that a third species may alelobe formed by coordinative chemisorption onto Lewis acid sites. This could be a precursor of dissociative adsorption. In a recent paper, Knozinger and Stubner2reported infrared results on the adsorption of isobutyl alcohol on1 v-A1203from which they concluded that an alcohol molecule is indeed coordinated to an anion vacancy. This result was mainly deduced from the study of (CH3)2CHCD20Hadsorption; the adsorption shifts the v(CD2)vibrations toward lower frequencies, which is taken as evidence for the formation of the postulated coordinated species. We here report some results which rather indicate that the CH stretching frequencies increase when ethers 0022-3654/80/2084-2083$0 1 .OO/O

or alcohols are coordinated and decrease when dissociative adsorption of alcohols occurs. The observation of CH stretching frequencies of CHDz groups, first carried out by Saur, Lavalley, and Romanet in CH3CH2Xcompounds,3 has demonstrated that the CH bond in the plane of dimethyl ether is slightly stronger than the out-of-plane bonds.4 The low frequency associated with the latter has been explained by a participation of the lone pair to a a* CH orbital on the adjacent carbon atom.5 This effect has been evidenced through the disappearance of the weak bond frequencies when the lone pairs were involved in complex formation, such as (CH&0-A1Cl3 or (CH3)20-BF3.6 Figure 1 illustrates how coordination of CD30CD2Hwith A1C13 affects the v(CH) vibrations of the ether; the two v(CH) frequencies increase, whatever the position of the CH bond. This is in agreement with Derouault’s results obtained by force constant refinement of (CH3)20-AlX3complexes.’ Correspondingly, Yakerson et a1.8 noticed an increase of v(CH) frequencies when CH30CH3coordinates onto alumina. We confirm this result using CD30CD2H(Figure 1); admission of 40 0 1980 American Chemical Society