Click Here

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, D01203, doi:10.1029/2009JD012701, 2010

for

Full Article

Self-attenuation artifacts and correction factors of light element measurements by X-ray analysis: Implication for mineral dust composition studies P. Formenti,1 S. Nava,2,3 P. Prati,4,5 S. Chevaillier,1 A. Klaver,1 S. Lafon,1 F. Mazzei,4,5 G. Calzolai,2,3 and M. Chiari2,3 Received 19 June 2009; revised 28 September 2009; accepted 8 October 2009; published 15 January 2010.

[1] On a global scale, mineral dust is one of the major components of atmospheric

aerosols and has important effects on the radiative budget of the atmosphere and thus on climate forcing. An accurate measurement of the concentration of crustal elements, namely Na, Mg, Al, Si, K, Ca, Ti, and Fe, is mandatory for the study of desert aerosols. The concentration of light elements, when measured by X-ray emission techniques such as X-ray fluorescence (XRF) and particle-induced X-ray emission (PIXE), can be underestimated owing to self-absorption of the emitted soft X-rays inside aerosol particles. In this work, we analyzed dust samples collected in field campaigns and samples produced in the laboratory using dust of known composition. Measurements have been conducted with PIXE and energy-dispersive XRF (ED-XRF), together with an attenuation-free technique such as particle-induced gamma-ray emission (PIGE) and attenuation corrected wavelength-dispersive XRF (WD-XRF) by internal standard calibration. We focus on the determination of Al and present results of a PIXE versus PIGE intercomparison. Aluminum concentration was measured with both techniques in dust samples collected by aircraft sampling over western Africa during winter 2006 and summer 2007. An underestimation of the Al concentration determined by PIXE was observed (up to 40%), and it was compared with the results of a simple calculation using basic physics and the size distribution of the collected aerosol. Similar attenuation was observed for Mg, Al, and Si in the laboratory samples analyzed by ED-XRF and WD-XRF. In order to use concentration ratios involving light elements as tracers of the region of emission of the sampled dust, these artifacts (i.e., underestimation of the concentration of light elements) induced by self-attenuation should be properly considered and corrected. Citation: Formenti, P., S. Nava, P. Prati, S. Chevaillier, A. Klaver, S. Lafon, F. Mazzei, G. Calzolai, and M. Chiari (2010), Self-attenuation artifacts and correction factors of light element measurements by X-ray analysis: Implication for mineral dust composition studies, J. Geophys. Res., 115, D01203, doi:10.1029/2009JD012701.

1. Introduction [2] Atmospheric aerosols are a major unknown in climate research since they present a large variability in terms of origin, concentration, and properties [Forster et al., 2007]. [3] Amongst the various aerosol species, mineral dust plays a special role. One of the most abundant aerosol species in the atmosphere, mineral dust accounts for 40% of the suspended particulate matter (PM) at the global scale 1 Laboratoire Interuniversitaire des Syste`mes Atmosphe´riques, Universite´s Paris VII and XII, CNRS, Creteil, France. 2 Department of Physics, University of Florence, Florence, Italy. 3 INFN, Florence, Italy. 4 Department of Physics, University of Genoa, Genoa, Italy. 5 INFN, Genoa, Italy.

Copyright 2010 by the American Geophysical Union. 0148-0227/10/2009JD012701$09.00

[Forster et al., 2007], the largest contribution being from arid and semiarid regions of North and West Africa [Laurent et al., 2008]. Dust has multiple impacts on climate, environment, and human health [Sokolik et al., 2001]. This is related to the extension of its size distribution. Dust particles have diameters ranging from fractions to tens of microns, changing in proportion during transport as a result of dry and wet deposition. Various measurements [Schu¨tz, 1987; Formenti et al., 2001; Maring et al., 2003; Garrett et al., 2003] have shown how the coarse fraction gets depleted during transport of mineral dust from Africa across the Atlantic Ocean. [4] Besides size, the impact of mineral depends on composition. Mineral dust is made of various minerals whose proportions vary not only according to size but also according to the mineralogy of the source region of emission, to the erosion conditions, and to the distance after transport [Sokolik et al., 2001]. The major components are,

D01203

1 of 8

D01203

FORMENTI ET AL.: DUST CORRECTION BY X-RAY ANALYSIS

in order of mean abundance, alumino-silicates in the form of clays (illite, kaolinite, smectite, and montmorillonite) or feldspar, quartz (silicon dioxide), calcium carbonates (calcite and dolomite), and iron oxides (hematite and goethite) [Pye, 1987]. All these mineral phases have a specific importance in terms of dust impact as they have spectrally different refractive indices in the visible and infrared spectra (relevant to the impact on radiation) and/ or different hygroscopicity and solubility properties (relevant to the impact on cloud microphysics and biogeochemistry). Therefore their relative concentrations have to be estimated in order to make predictions. However, the direct quantification of the dust mineralogical composition by classical soil science methods such as X-ray diffraction is not an easy task [Caquineau et al., 2002]. Even when performed under conditions optimized for aerosol analysis, the limit of the detection of X-ray diffraction remains elevated, of the order of 800 mg of the total collected mass [Caquineau et al., 1997]. At concentrations typical of airborne dust outside storm events, this lower limit is often unachievable, especially when sampling is performed on aircraft platforms, where the duration of sampling exposure and the flow rate are often technically limited. Nevertheless, aircraft sampling is unavoidable to characterize mineral dust at the global and regional scales because of the fact that transport of dust plumes, especially when intercontinental, occurs mostly above the boundary layer, where it cannot be documented by ground-based stations. [ 5 ] Furthermore, the conversion of the peak area corresponding to a certain mineral in the X-ray diffraction spectra to its mass concentration is not straightforward as it depends on the chemical form under which a certain mineral is present in the sample [Caquineau et al., 1997]. This is particularly true in the case of clays, where substitution elements such as Fe, K, Mg, and Ca are found in the crystalline structure perturbing the nominal distance amongst reticular planes. Furthermore, the comparison of mineralogical concentration of different samples might be false if the ratio of crystallized to amorphous mass differs. [6] Because of these difficulties, the mineralogical composition of mineral dust is often inferred by its elemental composition, which can be measured with higher accuracy. With the exception of carbon and oxygen, Al, Si, Na, Mg, K, Ca, and Fe are the most abundant and typical tracers of mineral dust [Pye, 1987]. Because their relative proportions in the corresponding minerals differ, the interelemental ratios are used as a proxy of the relative abundance of one mineral with respect to another [Pye, 1987]. By extension, the interelemental ratios measured at a particular receptor site located downwind of dust emission have been often used in conjunction with back trajectory analysis to retrace the origin of the dust-laden air masses impacting at the site and possibly to differentiate source regions. Comparing elemental ratios and not elemental concentrations also allows us to encompass the difficulty of comparing rapidly varying concentration levels which are typical of mineral dust, for example, during sandstorms where total aerosol concentrations might vary by several order of magnitudes in a few tenths of minutes. [7] With regard to the bulk elemental composition, the most popular ratios which have been used in the past are those to Al [Pye, 1987; Bergametti et al., 1989; Chiapello et

D01203

al., 1997; Formenti et al., 2001, 2003, 2008; Arimoto et al., 2002; Lafon et al., 2006; Shen et al., 2007; Cao et al., 2008; Yuan et al., 2008]. As an example, based on the Si/Al, Ca/Al, and Fe/Al ratios, Chiapello et al. [1997] differentiated three major sectors of transport to the Cape Verde Islands, just offshore of Senegal, of mineral dust which had been emitted from source regions at different latitudes over North and West Africa. [8] Despite being less abundant than Si, the reason for choosing Al as a tracer of mineral dust resides in the fact that Al can be dosed on a larger selection of collection media (polycarbonate filters and glass membranes) and by a larger number of analytical techniques than Si [Chow, 1995]. In these studies, Al concentration values had been obtained with nuclear activation analysis (NAA) or with ion beam analysis (IBA) methods. [9] IBA techniques are often chosen as they are fast, nondestructive, and do not necessitate any sample preparation. This is true, in particular, for PIXE [Johansson and Campbell, 1988], which is sensitive to all the crustal elements but oxygen and carbon. However, the quantitative detection of Al by PIXE is prone to artifacts owing to the self-absorption of the fluorescence X-rays of light elements (Z  15) inside the particle grains themselves [Boni et al., 1990; Jex et al., 1990; D’Alessandro et al., 2003]. Of course, the same problem is suffered by another widely used technique, the XRF. [10] This artifact depends on particle size and can reach up to a few tens of percent of the measured concentrations in PM10 particles (PM10 is defined as the particulate matter with aerodynamic diameter smaller than 10 mm). Estimating a correction factor as a function of particle size is therefore necessary when studying the mineral dust cycle and impact. Without a proper correction, self-attenuation-induced artifacts can significantly alter the typical concentration ratios discussed above and prevent a firm identification of the region of origin of the analyzed aerosol. [11] In a previous experiment this problem has been faced in two ways. In some cases [e.g., Formenti et al., 2003], aerosol samples have been analyzed by nuclear techniques such as NAA. NAA exploits the gamma ray emission of nuclei irradiated with thermal neutrons; gamma rays being more energetic than X-rays, they do not suffer significant attenuation when passing through particles with dimensions on the order of a few microns. Correction factors for the attenuation effect in mineral dust have been deduced on the basis of an intercomparison between NAA and PIXE [Cornille et al., 1990; Salma et al., 1997]. Nevertheless, the number of analyzed samples was quite limited and the samples had been collected not far from Damascus (Syria), so the aerosol composition could be significantly affected by anthropogenic sources. Furthermore, NAA is a very sensitive technique, but its widespread use as well as the intercomparison with other techniques is prevented by the difficulties in having available a proper neutron flow (actually to get access to a nuclear facility properly equipped). In other cases [e.g., Lafon et al., 2006; Formenti et al., 2008], attenuation artifacts are taken into account by using calibration standards made from a soil matrix whose composition and grain size distribution approach those expected for the aerosol samples. The grain size distribution is controlled by crunching. Although sound, this procedure

2 of 8

D01203

FORMENTI ET AL.: DUST CORRECTION BY X-RAY ANALYSIS

presents practically some difficulties in achieving a uniform deposit on a filter medium. [12] In this paper, we present a direct quantification of self-absorption of Al X-rays in dust aerosols estimated from field samples. The latter had been collected onboard the UK community BAe-146 research aircraft of the Facility for Airborne Atmospheric Measurements (FAAM) over western Africa, in the framework of the African Monsoon Multidisciplinary Analyses (AMMA) program and the Geostationary Earth Radiation Budget Intercomparisons of Longwave and Shortwave (GERBILS) experiment. [13] The quantification is based on the contemporary analysis by PIXE and by another IBA-type technique, PIGE [Johansson and Campbell, 1988]. This technique, based on the detection of gamma rays as NAA, is unaffected by selfattenuation artifacts. Nonetheless, this technique is not routinely employed for aerosol analysis of large data sets, such as those issued from field campaigns, because of two main reasons: PIGE is sensitive only to light elements (Li, B, F, Na, Mg, Al, and Si), and it does not have truly multielemental capabilities (only one or a few elements at a time can be detected, depending on the analytical settings). In this work Al was chosen due to its particular importance in the study of desert dust. The self-attenuation effect of other light elements was estimated with other techniques on dust samples generated in the laboratory by crunching dust of known composition.

2. Experimental Details 2.1. Sample Collection and Preparation [14] Field aerosol samples from the Sahel were collected onboard the UK community BAe-146 research aircraft of FAAM operated from Niger during two different field campaigns: the Special Observation Period of the AMMA project Dust and Biomass Experiment (SOP0/DABEX) in January – February 2006 [Haywood et al., 2008] and the GERBILS experiment in June 2007 [Christopher et al., 2009]. In wintertime, the aircraft flew over Niger, Benin, and Nigeria. The collected dust originated over northern Africa and was transported to the sampling areas in welldefined plumes within the boundary layer [Osborne et al., 2008]. In summertime, the aircraft traveled over Niger, Mali, and Mauritania. Dust layers encountered during this field campaign were at times transported from the Sahara within the Saharan Air Layer (SAL), but more frequently dust was generated locally over the Sahel by erosion due to large mesoscale convective systems [Marsham et al., 2008; Christopher et al., 2009]. As a consequence, the dust mean size distributions are expected to be different, summertime dust being possibly more enriched in coarse particles than wintertime dust. [15] Aerosol particles were sampled by filtration on 47 and 90 mm Nuclepore filters (nominal pore size of 0.4 mm) according to the procedure described in detail by Formenti et al. [2003, 2008]. Samples were collected only during horizontal flight legs lasting not less than 20– 30 min in order to guarantee sufficient loading of the filter samples. Under the typical sampling conditions the air volume that flowed through the filters was about 1 – 2 m3 with a deposition thickness ranging from 0.01 to 12.6 mg cm2. Blank samples were collected on every flight by placing

D01203

filters in the sampling line as if they were actual samples and exposing them to the air stream for a few seconds. The two data sets considered for this work consist of 38 samples (winter 2006, AMMA SOP0/DABEX) and 32 samples (summer 2007, GERBILS). [16] Laboratory samples were prepared in the Laser Interferometer Space Antenna (LISA) using a granite geostandard (GSN) of known and certified composition. In this case the geostandard powder was first crunched, and a weighted fraction was put in a known volume of a filtered ethanol solution. Then the solution was homogenized with ultrasonic waves and a vortex and then refiltered onto AOX 47 mm diameter Nuclepore membranes. 2.2. Compositional Analyses [17] PIXE-PIGE analyses have been performed by a proton beam at the 3 MV Tandetron accelerator of the LABEC laboratory of INFN (Florence), with the external beam setup described by Calzolai et al. [2006]. Each sample has been irradiated for about 1000 s with a beam intensity ranging from 5 to 30 nA, depending on the sample load, over a spot of 2 mm2. During irradiation the filter was moved in front of the beam so that most of the area of the deposit was analyzed. PIXE spectra were fitted using the GUPIX code [Maxwell et al., 1995] in the thin target approximation, and elemental concentrations (mg cm2) were obtained via a calibration curve from a set of thin standards of areal density known within 5% (Micromatter Inc.). PIGE has been run simultaneously to PIXE for the Al determination, exploiting the 1013 keV gamma rays produced by the 27Al(p,p0g)27Al reaction. PIGE is based on the detection of energetic gamma rays (Eg  1 MeV, in this case) and thus is not affected by artifacts due to the X-ray self-attenuation in the aerosol grains. Nevertheless, uncertainties in the quantitative analysis could come from the variation of the proton energy through the sample. A proper beam energy (3060 keV in vacuum before the exit window, corresponding to about 2900 keV on the target) has been selected in order to obtain a cross section for the reaction considered that remains constant despite the energy loss of the protons in the target (maximum 100 keV for aerosol grains of about 10 mm) [Marino et al., 2008]. [18] XRF analyses were used as well on the laboratory standards mentioned earlier. Measurements were carried on with different techniques in two different laboratories. In both the cases a pressure of the order of 0.1 mbar was maintained in the vacuum chamber hosting the samples in analysis. [19] ED-XRF analyses have been performed at the Physics Department of the Genoa University using an ED-2000 spectrometer by Oxford Instruments. With ED-2000, excitation X-rays are produced by a Coolidge tube (Imax = 1 mA, Vmax = 50 kV) with an Ag anode; the primary X-ray spectrum can be controlled by inserting filters (made of Al, Cu, and Ag) between the anode and the sample. Two measuring conditions were fixed to optimize the sensitivity for groups of elements: runs with V = 15 kV, I = 100 mA, no primary filter, and live time = 1000 s gave ‘‘low Z’’ elements (from Na to P) while the ‘‘medium-high Z’’ elements (from S to Pb) were measured setting V = 30 kV, I = 500 mA, thin Ag primary filter, and live time = 3000 s. X-ray spectra were fitted for 24 elements (Mg, Al, Si, P, S,

3 of 8

D01203

FORMENTI ET AL.: DUST CORRECTION BY X-RAY ANALYSIS

Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, As, Se, Br, Sr, Zr, Mo, Ba, and Pb) using AXIL software package [Van Espen et al., 1977]. The elemental concentrations (mg cm2) were obtained by comparing the sample yields with a sensitivity curve measured in the same geometry on a set of thin standards certified within 5% (Micromatter Inc.). As in the case of PIXE, with this technique, X-rays emitted by light elements can be attenuated if the sample thickness and/or the particle size are large enough. The calibration via a set of certified thin standards cannot take into account this effect which must be corrected in the data reduction [D’Alessandro et al., 2003], as explained in section 4. [20] WD-XRF analyses were performed at the LISA in Cre´teil using a PW-2404 spectrometer by Panalytical. With PW-2404, excitation X-rays are produced by a Coolidge tube (Imax = 125 mA, Vmax = 60 kV) with an Rh anode; the primary X-ray spectrum can be controlled by inserting filters (Al, at different thickness) between the anode and the sample. Each element was analyzed 3 times, with specific conditions (voltage, tube filter, collimator, analyzing crystal, and detector), lasting from 8 to 10 s. Data were collected for nine elements (Na, Mg, Al, Si, P, K, Ca, Ti, and Fe) using SuperQ software (by Panalytical, see http:// www.panalytical.com). In this case, the elemental concentrations (mg cm2) were obtained by comparing the sample yields with a sensitivity curve measured in the same geometry on a set of geostandards deposited on filters, characterized by a composition and a size distribution similar to those of real mineral dust samples. In this way, results are already self-attenuation corrected; however, the accuracy of the procedure strongly depends on the level of similarity between the standards and the real samples, and it is not trivial to achieve such a good level in the analysis of real-world aerosol samples of unknown composition and size distribution. 2.3. Dimensional Analyses [21] Electron microscopy analysis was used to investigate the aerosol size distribution on the field samples. We used a combination of scanning and transmission electron microscopes, both equipped with an energy dispersive X-ray detection system. This allows investigation of the 2-D structure of particles both in the fine and in the coarse size fractions. [22] Analytical scanning electron microscopy (SEM) was performed with an instrument type JEOL 6301F equipped with an X-ray energy-dispersive spectrometer (Oxford Link Pentafet Detector and Link ISIS analyzer, Oxford Instruments, UK). SEM allows obtaining ‘‘bulk’’ images of aerosol particles, i.e., 2-D images with some information regarding their volume. This is particularly useful in the investigation of mineral dust, which generally consists of complex aggregates in the coarse fraction. SEM has a large dynamic range of magnification, as it allows us to image particles of diameter from fractions to tens of micron. [23] Analytical transmission electron microscopy (TEM) was performed with an instrument type JEOL 100CXII equipped with an X-ray energy-dispersive spectrometer (PGT Prism 2000 Si(Li) detector and Avalon analyzer, Princeton Gamma-Tech). TEM provides a better resolution for particles smaller than 0.5 mm in diameter. However, only

D01203

2-D images of particles are provided without any further information on their bulk structure. [24] Full details of data acquisition (SEM and TEM) and analysis are presented by Chou et al. [2008]. Images were analyzed using the Histolab counting program (Microvision Instruments, France). The particle geometric diameter Dp was calculated from the projected area Aproj assumed to be the equivalent area of a spherical particle rffiffiffiffiffiffiffiffiffi  2 Dp Aproj Aproj ¼ p ; i:e:; Dp ¼ 2 : 2 p

ð1Þ

[25] The image analysis provided us with the total number of particles per image (Ni,TOT) as well as their number size distribution dNi(Dp). The particle number size distribution in the atmosphere dN(Dp) was then calculated as     X     Sfilter 1  ; dNi Dp  dN Dp ¼ STOT Vair

ð2Þ

where Sfilter is the surface of the filter, STOT is the total analyzed surface, and Vair is the air volume sampled through the filter. [26] In the following, the number size distribution is presented as dN(Dp)/dlogDp normalized to the total particle number. [27] Because of the large number density of particles, the size distribution of the GSN laboratory standard could not be analyzed by electron microscopy. Instead, we used an FPIA-3000 Flow Particle Image Analyzer. The system uses a charge-coupled device (CCD) camera to capture pictures of the particles as they pass in a controlled manner through a flow cell. The images are captured in real time and are analyzed in terms of their morphological parameters (26 parameters available) to produce corresponding distribution. An aliquot (volume of 0.5 –5 mL) of the ethanol solution used to produce the laboratory standards was analyzed. Results were expressed in terms of number and volume distribution as dN(Dp)/dlogDp and dV(Dp)/dlogDp, respectively. [28] To estimate the relative behavior of the various size distributions, we calculated the effective diameter Dp,eff as Z Dp;eff ¼

dN d log Dp d log Dp : dN D2p d log Dp d log Dp D3p

ð3Þ

3. Results [29] Results of the simultaneous PIXE and PIGE Al concentrations on the field samples are given in Figure 1 as scatterplots of values obtained by the two techniques for the two series of data separately. The AMMA SOP0/ DABEX data (Figure 1a) show lower Al values, high PIXE versus PIGE correlation (R2 = 0.98), and an average attenuation factor (PIXE to PIGE ratio) of 0.88. GERBILS data (Figure 1b) present higher Al levels, a more scattered

4 of 8

D01203

FORMENTI ET AL.: DUST CORRECTION BY X-RAY ANALYSIS

D01203

particle. Some of the samples have been analyzed by SEM to measure the size distribution of the collected particles. Winter samples turned out to be very homogeneous also in this respect; as a synthetic figure the average effective diameter is 3.0 ± 0.3 mm [Chou et al., 2008]. The size distribution of the summer samples was firmly measured on two samples only since the average high loading turned out in a frequent overlapping of particles in the SEM images. The average effective diameter deduced was 5.5 ± 1.0 mm. The SEM size distributions are compared in Figure 2. It should be noted that alumino-silicate minerals being the most abundant species in the aerosol [Chou et al., 2008], the size distribution of Al can be assimilated to that of the bulk aerosol. [31] With a simple attenuation model assuming single spherical particles, attenuation factors (AF) can be calculated as a function of the geometric particle diameter, D [Holynska and Markowicz, 1981], AF ¼

Figure 1. Regression line between Al concentration values measured by PIXE and PIGE on samples collected in the Sahel area: (a) 38 samples from winter campaign in 2006 and (b) 32 samples from summer campaign in 2007. The black line corresponds to the average attenuation coefficient, while the 45° dashed line simply indicates the condition of no attenuation. distribution (R2 = 0.89), and an average attenuation factor as low as 0.65. This result can be explained in terms of the differences in the origin and mean size distribution of the sampled aerosol (as discussed previously, summertime dust is possibly more enriched in coarse particles than wintertime dust). As a matter of fact, while winter data look more homogeneous, with attenuation in the range of 0 – 25% (10 – 90th percentile), the larger dispersion of PIXE/PIGE concentration ratios observed in the summer samples, with attenuation in the range of 15– 50% (10 – 90th percentile), makes less significant the figure of a unique attenuation factor. [30] The attenuation of Al X-rays could be due, in principle, to the layer of PM deposited on the Nuclepore filters. Actually, the PM thickness on the filters was deduced by a standard gravimetric analysis and ranged between 5 and 100 mg cm2: with a density of the deposited PM on the order of 2 – 3 g cm3, this corresponds to an average layer thickness of 0.3 –0.4 mm maximum and thus to a negligible X-ray attenuation. In other words, a significant X-ray self-attenuation can only occur inside each PM

1  expða  DÞ ; aD

ð4Þ

2 with a =  m  r, where r is the particle mass density and 3 m is the mass attenuation coefficient (which depends on the particle composition and on the X-ray energy). Using this model, we calculated attenuation factors for Na, Mg, Al, Si, and K and for typical desert dust minerals, namely K-feldspar, kaolinite, and illite (Figure 3). The chemical formulae for K-feldspar and kaolinite are KAlSi3O8 and Al2Si2O9H4, respectively. For illite, we used an average composition expressed in term of percent of oxides as SiO2 53%, Al2O3 29%, K2O 9.31%, Fe2O3 0.38%, CaO 0.38%, MgO 1.11%, TiO2 0.11%, Na2O 0.28%, and LOI 5.44%. [32] As can be seen from Figure 3, the attenuation of Al X-rays is almost the same in the three different minerals: we found that the three curves can be well fitted by a single function described by an analytical expression like that reported in equation (4) with a = 0.2 mm1. We thus used this expression to calculate the Al attenuation for the samples in which the size dimensional distribution had been measured by SEM (Figure 2). We obtained typical attenu-

Figure 2. Size distribution of winter (circles) and summer (crosses) data deduced by the SEM analyses on field samples.

5 of 8

D01203

FORMENTI ET AL.: DUST CORRECTION BY X-RAY ANALYSIS

D01203

(see section 2.2). We therefore adopted a different approach based on the analysis by WD-XRF (self-attenuation corrected; see section 2.2 on laboratory analyses) and ED-XRF (sensitive to self-attenuation) of samples prepared in the laboratory with crunched dust of known composition. The comparison among the results obtained with the analysis of three samples (all with a deposited thickness of around 100 mg cm2) indicated that in the crunched dust the concentration of Mg, Al, and Si obtained by ED-XRF was attenuated by an average factor of 0.64 ± 0.05 (basically the same value for the three elements within the experimental uncertainty) while from K onward, no significant effect was observed. A further group of three samples was analyzed to control the reproducibility of the preparation procedure (in particular, the crunching) in terms of sample composition. The latter turned out to be constant within the 10% analytical uncertainty. The size distribution of the dust used to prepare the laboratory samples was measured with the optical method described above (see section 2.3): the size of the deposited dust ranged from about 2 to about 10 mm, with the mean value being around 6.5 mm. Following the same approach adopted for the field samples, we calculated the expected attenuation for Mg, Al, and Si concentration values (using equation (4) with aMg = 0.32, aAl = 0.20, and aSi = 0.21) and obtained attenuation factors equal to 0.42, 0.56, and 0.55, respectively (Na was not present in detectable concentrations in the laboratory samples). The calculation is in a quite fair agreement with the measured attenuations in the case of Al and Si, while it looks overestimated for Mg.

4. Discussion

Figure 3. Attenuation of Na, Mg, Al, and Si X-rays inside a spherical particle grain calculated in (a) illite, (b) kaolinite, and (c) K-feldspar. ation factors of about 0.8 and 0.6 for the winter and summer sets, respectively. The result is in fair agreement with the direct measurement of the attenuation based on the PIXE/ PIGE Al concentration ratios given above (Figure 1) and shows that this simple model can provide an estimate of the attenuation of X-ray emitted by light elements provided the size distribution is known. [33] For what concerns the other light elements, in the minerals considered and for particles with diameters from 5 to 1 mm, the attenuation factor for Si is in the 0.60 – 0.90 range; very similar to that of Al, the attenuation factors of Mg and Na are in the 0.50– 0.85 and 0.40 – 0.80 ranges, respectively, while the attenuation factor of K is always greater than about 0.85. A check of the attenuation calculated with the model for these elements was not possible by direct measurement of the PIXE/PIGE concentration ratio

[34] The self-attenuation of soft X-rays emitted by light elements, if not corrected, can affect the possibility of distinguishing the origin of dust on the basis of the characteristic concentration ratios as discussed in section 1. In a recent paper, Marenco et al. [2006] reported on several episodes of Saharan dust long-range transport which have been detected and characterized in the frame of a large campaign at the Global Atmospheric Watch station of Mount Cimone (Italy). The analysis of the samples collected in this experiment by ED-XRF not corrected by any attenuation factor led to singling out of a geochemical signature of the composition of transported dust in terms of the Ca/Al concentration ratio (Ca/Al = 0.90 ± 0.23 [Marenco et al., 2006]). The authors compared their figure with the Ca/Al measured during a previous campaign at Capo Verde (Ca/Al = 1.68 ± 0.78 [Chiappello et al., 1997]; see section 1): they observed that this was the sole concentration ratio not in agreement in the two experiments (other typical concentration ratios had been measured by the two groups as Si/Al, Fe/Ca, K/Ca, Ti/Ca, and Ti/Fe). The 50% discrepancy in the Ca/Al ratios is of the same order as the attenuation factor measured for Al in this work and could be easily eliminated by a proper correction. The Al/Si concentration ratio should be used with caution. As a matter of fact, the mean attenuation calculated from the various minerals is almost equal for both elements, and the attenuation effect cancels out. However, the attenuation is different when looking at the individual minerals separately (Figure 3). Because of the different size distribution of these elements

6 of 8

FORMENTI ET AL.: DUST CORRECTION BY X-RAY ANALYSIS

D01203

Table 1. Typical Concentration Ratios for Some Elements Contained in Some Dust Geostandards Used in This Worka Geostandard

Al/Fe

Si/Fe

Ca/Fe

Ti/Fe

TUMB CUBB CLOB

2.3 2.7 1.6

33.3 18.2 6.7

2.3 0.8 1.9

0.03 0.06 0.15

a See text for the meaning of the labels. A 5% uncertainty should be considered on the reported ratios.

(feldspar and quartz being characterized by larger mean distribution than illite and kaolinite [Pye, 1987]), it is expected that the correction on the Si/Al ratio should progressively lose importance with time after emission. [35] When a safe correction is not possible, we recommend to rely on concentration ratios of elements with atomic number Z greater than 19 (K), like the Ca/Fe ratio. As an example one could consider the geochemical signature of some of the geostandard used in this work. We used dust with different origin: dust labeled TUMB from the Mount Maouna (Tunisia), CUBB from the Ulan-Bush desert (China), and CLOB from the Loess upland (northwestern China). We report in Table 1 some typical concentration ratios for these dust samples. Looking at the Al/Fe ratios it is clear that a 30% attenuation of the Al concentration, if not corrected, prevents the possibility of distinguishing the origin of the dust. The span of the Si/Fe ratios is much larger and a possible underestimation of Si concentration values could be less important from a practical point of view. The concentrations ratios between elements unaffected by the attenuation artifacts (i.e., Ca/Fe and Ti/Fe in Table 1) can be safely used to distinguish the three kinds of dust.

5. Conclusions [36] In the atmosphere, and even in layers above the boundary layer well downwind source regions, the typical dimension of desert dust grains collected on filters are such that they produce significant attenuation in the X-rays emitted by light elements as Na, Mg, Al, and Si. For some of these elements, true concentration values could be obtained by analyzing the samples with attenuation free techniques (i.e., PIGE or, in some cases, inductively coupled plasma spectroscopy [Boumans, 1987a, 1987b]), but in a PIXE and/or XRF analysis a correction is necessary. [37] Provided that the size distribution of the collected aerosol is known/measured, a correction factor can be deduced by a simple calculation assuming spherical particles. The tests presented in this paper indicate that this approach is in quite good agreement with the measured attenuation for Al and Si. In the case of Na and Mg, the composition of the analyzed samples and the level of the experimental uncertainties prevent any firm conclusion, even if according to our results a more refined model seems necessary. [38] In this perspective, a refinement of the present results would be to calculate size-resolved correction factors by analyzing samples collected by multistage cascade impactors rather than bulk filter samples. In this case, particular care should be devoted to the choice of equipment and sampling conditions to avoid the pileup of particle grains. Some devices, such as the rotating MOUDI impactor

D01203

[Marple et al., 1991], have been developed to minimize such effect. [39] Acknowledgments. This work has been financially supported by the France-Italy cooperation project Galileo 2006/2007. On the basis of a French initiative, AMMA was built by an international scientific group and is currently funded by a large number of agencies, especially from France, the United Kingdom, the United States, and Africa. It has been the beneficiary of a major financial contribution from the European Community’s Sixth Framework Research Programme. Detailed information on scientific coordination and funding is available on the AMMA International Web site at www.amma-international.org. P. Formenti acknowledges financial support from the API-AMMA national program and the UK Met Office for participating in the AMMA SOP0/DABEX and GERBILS campaigns. Jim Haywood, PI of both campaigns, is thanked for organizing the campaigns and providing invitations.

References Arimoto, R., W. Balsam, and C. Schloesslin (2002), Visible spectroscopy of aerosol particles collected on filters: Iron-oxide minerals, Atmos. Environ., 36, 89 – 96. Bergametti, G., L. Gomes, G. Coude-Gaussen, P. Rognon, and M.-N. Le Coustumer (1989), African dust observed over Canary Islands: Sourceregions identification and transport pattern for some summer situations, J. Geophys. Res., 94, 14,855 – 14,864. Boni, C., A. Caridi, E. Cereda, and G. M. Braga Marcazzan (1990), A PIXE-PIGE setup for the analysis of thin samples, Nucl. Instrum. Methods Phys. Res. Sect. B, 47, 133 – 142. Boumans, P. W. J. M. (1987a), Inductively Coupled Plasma-Emission Spectroscopy, Part 1, Methodology, Instrumentation and Performance, John Wiley, Hoboken, N. J. Boumans, P. W. J. M. (1987b), Inductively Coupled Plasma-Emission Spectroscopy, Part 2. Application and Fundamentals, John Wiley, Hoboken, N. J. Calzolai, G., M. Chiari, I. Garcıa Orellana, F. Lucarelli, A. Migliori, S. Nava, and F. Taccetti (2006), The new external beam facility for environmental studies at the Tandetron accelerator of LABEC, Nucl. Instrum. Methods Phys. Res. Sect. B, 249, 928 – 931. Cao, J. J., S. C. Lee, X. Y. Zhang, J. C. Chow, Z. S. An, K. F. Ho, J. G. Watson, K. Fung, Y. Q. Wang, and Z. X. Shen (2008), Characterization of airborne carbonate over a site near Asian dust source regions during spring 2002 and its climatic and environmental significance, J. Geophys. Res., 110, D03203, doi:10.1029/2004JD005244. Caquineau, S., M.-C. Magonthier, A. Gaudichet, and L. Gomes (1997), An improved procedure for the X-ray diffraction analysis of low-mass atmospheric dust samples, Eur. J. Mineral., 9, 157 – 166. Caquineau, S., A. Gaudichet, L. Gomes, and M. Legrand (2002), Mineralogy of Saharan dust transported over northwestern tropical Atlantic Ocean in relation to source regions, J. Geophys. Res., 107(D15), 4251, doi:10.1029/2000JD000247. Chiapello, I., G. Bergametti, B. Chatenet, P. Bousquet, F. Dulac, and E. Santos Soares (1997), Origins of African dust transported over the northeastern tropical Atlantic, J. Geophys. Res., 102, 13,701 – 13,709. Chou, C., P. Formenti, M. Maille, P. Ausset, G. Helas, M. Harrison, and S. Osborne (2008), Size distribution, shape, and composition of mineral dust aerosols collected during the African Monsoon Multidisciplinary Analysis Special Observation Period 0: Dust and Biomass-Burning Experiment field campaign in Niger, January 2006, J. Geophys. Res., 113, D00C10, doi:10.1029/2008JD009897. Chow, J. C. (1995), Critical review: Measurement methods to determine compliance with ambient air quality standards for suspended particles, J. Air Waste Manage. Assoc., 45, 320 – 382. Christopher, S. A., B. Johnson, T. A. Jones, and J. Haywood (2009), Vertical and spatial distribution of dust from aircraft and satellite measurements during the GERBILS field campaign, Geophys. Res. Lett., 36, L06806, doi:10.1029/2008GL037033. Cornille, P., W. Maenhaut, and J. M. Pacyna (1990), Sources and characteristics of the atmospheric aerosol near Damascus, Syria, Atmos. Environ., Part A, 24, 1083 – 1093. D’Alessandro, A., F. Lucarelli, P. A. Mando`, G. Marcazzan, S. Nava, P. Prati, G. Valli, R. Vecchi, and A. Zucchiatti (2003), Hourly elemental composition and sources identification of fine and coarse PM10 particulate matter in four Italian towns, J. Aerosol Sci., 34, 243 – 259. Formenti, P., M. O. Andreae, J. Cafmeyer, W. Maenhaut, B. N. Holben, L. Lange, G. Roberts, P. Artaxo, and J. Lelieveld (2001), Saharan dust in Brazil and Suriname during the Large-Scale Biosphere-Atmosphere Experiment in Amazonia (LBA) – Cooperative LBA Regional Experiment (CLAIRE) in March 1998, J. Geophys. Res., 106, 14,919 – 14,934.

7 of 8

D01203

FORMENTI ET AL.: DUST CORRECTION BY X-RAY ANALYSIS

Formenti, P., W. Elbert, W. Maenhaut, J. Haywood, and M. O. Andreae (2003), Chemical composition of mineral dust aerosol during the Saharan Dust Experiment (SHADE) airborne campaign in the Cape Verde region, September 2000, J. Geophys. Res., 108(D18), 8576, doi:10.1029/ 2002JD002648. Formenti, P., et al. (2008), Regional variability of the composition of mineral dust from western Africa: Results from the AMMA SOP0/ DABEX and DODO field campaigns, J. Geophys. Res., 113, D00C13, doi:10.1029/2008JD009903. Forster, P., et al. (2007), Changes in atmospheric constituents and in radiative forcing, in Climate Change 2007: The Physical Science. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon et al., pp. 131 – 234, Cambridge Univ. Press, Cambridge, U. K. Garrett, T. J., L. M. Russell, V. Ramaswamy, S. F. Maria, and B. J. Huebert (2003), Microphysical and radiative evolution of aerosol plumes over the tropical North Atlantic Ocean, J. Geophys. Res., 108(D1), 4022, doi:10.1029/2002JD002228. Haywood, J. M., et al. (2008), Overview of the Dust and Biomass-burning Experiment and African Monsoon Multidisciplinary Analysis Special Observing Period-0, J. Geophys. Res., 113, D00C17, doi:10.1029/ 2008JD010077. Holynska, B., and A. Markowicz (1981), Experimental evaluation of the Rhodes-Hunter model for the particle size effect in X-ray fluorescence analysis of thin samples, X Ray Spectrom., 10, 61 – 63. Jex, D. G., M. W. Hill, and N. F. Mangelson (1990), Proton induced X-ray emission of spherical particles: Corrections for X-ray attenuation, Nucl. Instrum. Methods Phys. Res. Sect. B, 49, 141 – 145. Johansson, S. A., and J. L. Campbell (1988), PIXE: A Novel Technique for Elemental Analysis, John Wiley, Hoboken, N. J. Lafon, S., I. N. Sokolik, J. L. Rajot, S. Caquineau, and A. Gaudichet (2006), Characterization of iron oxides in mineral dust aerosols: Implications for light absorption, J. Geophys. Res., 111, D21207, doi:10.1029/ 2005JD007016. Laurent, B., B. Marticorena, G. Bergametti, J. F. Le´on, and N. M. Mahowald (2008), Modeling mineral dust emissions from the Sahara desert using new surface properties and soil database, J. Geophys. Res., 113, D14218, doi:10.1029/2007JD009484. Marenco, F., et al. (2006), Characterization of atmospheric aerosols at Monte Cimone, Italy, during summer 2004: Source apportionment and transport mechanisms, J. Geophys. Res., 111, D24202, doi:10.1029/2006JD007145. Maring, H., D. L. Savoie, M. A. Izaguirre, L. Custals, and J. S. Reid (2003), Mineral dust aerosol size distribution change during atmospheric transport, J. Geophys. Res., 108(D19), 8592, doi:10.1029/2002JD002536. Marino, F., G. Calzolai, S. Caporali, E. Castellano, M. Chiari, F. Lucarelli, V. Maggi, S. Nava, M. Sala, and R. Udisti (2008), PIXE and PIGE techniques for the analysis of Antarctic ice dust and continental sediments, Nucl. Instrum. Methods Phys. Res. Sect. B, 266, 2396 – 2400.

D01203

Marple, V. A., K. L. Rubow, and S. M. Behm (1991), A Microorifice Uniform Deposit Impactor (MOUDI): Description, calibration, and use, Aerosol Sci. Technol., 14, 434 – 446. Marsham, J. H., D. J. Parker, C. M. Grams, C. M. Taylor, and J. M. Haywood (2008), Uplift of Saharan dust south of the intertropical discontinuity, J. Geophys. Res., 113, D21102, doi:10.1029/2008JD009844. Maxwell, J. A., W. J. Teesdale, and J. L. Campbell (1995), The Guelph PIXE package II, Nucl. Instrum. Methods Phys. Res. Sect. B, 95, 407 – 421. Osborne, S. R., B. T. Johnson, J. M. Haywood, A. J. Baran, M. A. J. Harrison, and C. L. McConnell (2008), Physical and optical properties of mineral dust aerosol during the dust and biomass-burning experiment, J. Geophys. Res., 113, D00C03, doi:10.1029/2007JD009551. Pye, K. (1987), Aeolian Dust and Dust Deposits, Academic Press, New York. Salma, I., W. Maenhaut, H. J. Annegarn, M. O. Andreae, F. X. Meixner, and M. Garstang (1997), Combined application of INAA and PIXE for studying the regional aerosol composition in Southern Africa, J. Radioanal. Nucl. Chem., 216, 143 – 148. Schu¨tz, L. (1987), Sahara dust transport over the North Atlantic Ocean – model calculations and measurements, in Saharan Dust, SCOPE Ser., vol. 14, edited by C. Morales, pp. 267 – 277, John Wiley, Chichester, U. K. Shen, Z. X., J. J. Cao, R. Arimoto, R. J. Zhang, D. M. Jie, S. X. Liu, and C. S. Zhu (2007), Chemical composition and source characterization of spring aerosol over Horqin sand land in northeastern China, J. Geophys. Res., 112, D14315, doi:10.1029/2006JD007991. Sokolik, I. N., D. M. Winker, G. Bergametti, D. A. Gillette, G. Carmichael, Y. J. Kaufman, L. Gomes, L. Schuetz, and J. E. Penner (2001), Introduction to special section: Outstanding problems in quantifying the radiative impact of mineral dust, J. Geophys. Res., 106, 18,015 – 18,027. Van Espen, P., H. Nullens, and F. Adams (1977), A computer analysis of X-ray fluorescence spectra, Nucl. Instrum. Methods Phys. Res., 142, 243 – 250. Yuan, X., Z. Xie, J. Zheng, X. Tian, and Z. Yang (2008), Effects of water table dynamics on regional climate: A case study over east Asian monsoon area, J. Geophys. Res., 113, D21112, doi:10.1029/2008JD010180. 

S. Chevaillier, P. Formenti, A. Klaver, and S. Lafon, Laboratoire Interuniversitaire des Syste`mes Atmosphe´riques, UMR 7583, Universite´s Paris VII and XII, CNRS, 61 Avenue du Ge´ne´ral de Gaulle, F-94010 Cre´teil CEDEX, France. G. Calzolai, M. Chiari, and S. Nava, Department of Physics, University of Florence, Via G. Sansone 1, I-50019 Sesto Fiorentino (Firenze), Italy. F. Mazzei and P. Prati, Department of Physics, University of Genoa, Via Dodocaneso 33, I-16146 Genova, Italy. ([email protected])

8 of 8