Laser ablation ICP-MS of size-segregated atmospheric particles collected with a MOUDI cascade impactor : a proof of concept

A widely used instrument for collecting sizesegregated particles is the micro-orifice uniform deposit impactor (MOUDI). In this work, a 10-stage MOUDI (cutpoint diameter of 10 μm to 56 nm) was used to collect samples in Ljubljana, Slovenia, and Martinska, Croatia. Filters, collected with and without rotation, were cut in half and analyzed for nine elements (As, Cu, Fe, Ni, Mn, Pb, Sb, V, Zn) using laser ablation ICP-MS. Elemental image maps (created with ImageJ) were converted to concentrations using NIST SRM 2783. Statistical analysis of the elemental maps indicated that for submicron particles (stages 6–10), ablating 10 % of the filter (0.5 cm2, 20 min ablation time) was sufficient to give values in good agreement (±10 %) to analysis of larger parts of the filter and with good precision (RSE < 1 %). Excellent sensitivity was also observed (e.g., 20± 0.2 pg m−3 V). The novel use of LA-ICP-MS, together with image mapping, provided a fast and sensitive method for elemental analysis of size-segregated MOUDI filters, particularly for submicron particles.


Introduction
Inhalation of particle-bound metals in atmospheric particulate can negatively impact human health (Chen and Lippmann, 2009).Particle-bound Fe, Ni, and V can lead to oxidative stress, pulmonary inflammation, cardiac effects, and cardiovascular and respiratory illnesses (Aust et al., 2002;Bell et al., 2009;Campen et al., 2002).Particle size is also a factor.Submicron particles pose the greatest health risks (Davidson et al., 2005), and particle-bound metals from anthropogenic sources (e.g., fossil fuel combustion and ve-hicle emissions) commonly partition to these smaller sizes (< 1 µm) (Fang and Huang, 2011).A well-known instrument for the collection and speciation of size-segregated particles is the multi-orifice uniform deposit impactor (MOUDI) (Allen et al., 2001;Herner et al., 2006;Ntziachristos et al., 2007;Pekney et al., 2006;Singh et al., 2002).Particles are sorted by size into stages using cascade impaction and deposited on filters.Models with various flow rates (2-130 L min −1 ) and stages (cut-point diameter between 0.01 and 18 µm) are available, as well as rotating and non-rotating options (Marple, 1991;Marple et al., 2014).To determine elemental concentrations, filters are typically extracted with acid and analyzed by ICP-MS (Canepari et al., 2008;Herner et al., 2006;Li et al., 2012;Ntziachristos et al., 2007;Pekney et al., 2006) or ICP-AES (Fang and Huang, 2011), whereas in some cases direct analysis of the filters was performed by X-ray fluorescence spectrometry 11 or proton-induced X-ray emission (Brüggeman et al., 2009).Detection limits using these methods are challenging due to low particle mass and contamination risks; after corrections have been made for blanks, elemental concentrations are often below detection limits (Pekney et al., 2006).
A promising alternative to acid extraction involves laser ablation ICP-MS (LA-ICP-MS) (Aubriet and Carré, 2010).This method samples the filters via pulsed laser ablation, effectively removing the particles, thereby eliminating the need for harsh chemicals, offering faster sample preparation times, reducing contamination, and increasing sensitivity.Initial efforts using LA-ICP-MS have been promising (Gligorovski et al., 2008;Triglav et al., 2010), although problems associated with spatial inhomogeneity (Brown et al., 2009), matrixmatched standards (Chin et al., 1999;Tanaka et al., 1998), and laser instabilities have been reported.Hsieh et al. (2011) optimized the LA-ICP-MS parameters suitable for analysis of nanometer-and submicrometer-sized airborne particulate matter sampled by an electrical low-pressure impactor.In this work, we addressed these problems by using a MOUDI for particle collection and a highly stable excimer laser (193 nm ArF * ) for particle ablation.To our knowledge, we are the first to analyze MOUDI filters by LA-ICP-MS.Filters collected both with and without rotation were analyzed.Concentrations of nine elements from 10 MOUDI stages are reported with specific attention given to submicron particles (stages 6-10).

Laser ablation ICP-MS
A quadrupole ICP-MS (Agilent Technologies 7900, Palo Alto, USA) interfaced with a laser ablation system (193 nm ArF * excimer, Analyte G2, Teledyne Photon Machines, Inc.) was used to analyze nine nuclides ( 75 As, 63 Cu, 57 Fe, 55 Mn, 60 Ni, 208 Pb, 121 Sb, 51 V, 66 Zn).(See optimized parameters in Table 1.) Ablation took place in a HelEx two-volume cell applying a laser beam size of 150 µm (square mask), a scanning speed of 300 µm s −1 , a repetition rate of 10 Hz, and a fluence of 1.21 J cm −2 .Ablated materials were transported from the ablation cell to the plasma with helium; argon was added as the make-up gas before the ICP torch.Ions formed in the plasma were extracted and separated by their mass-to-charge (m/z) ratios.The mass spectrometer, in time-resolved analysis mode, measured one point per mass for the nine selected masses.The detection limit for each element was determined as 3 × SD of seven clean blanks (Table 2).

Filter preparation and ablation
All 10 filters (stages 1-10) were ablated in samples 1 and 2 (Ljubljana, rotation), eight filters (stages 3-10) were ablated in sample 3 (Martinska, non-rotation), and four filters (samples 6-9) were ablated in sample 4 (Martinksa, rotation).Filters were cut in half and four or five were co-mounted on a single glass slide with double-sided tape (Fig. 1).For the rotated filters (Fig. 1a), the laser raster pattern comprised parallel lines that spanned the width and ran the length of the co-mounted half filters.For the non-rotated filters (Fig. 1b), The laser beam energy was sufficient to remove several layers of the PTFE filter in each particle size range, and no particles were visible on the filters (100 × magnification) following ablation; hence, we assume that most particles were removed during ablation.However, deeply impacted particles may have remained embedded in the filters, particularly for the smaller particles in the case of MOUDI sampling without rotation and high loading.Elemental concentrations were determined via one-point calibration with the NIST standard.

Quality assurance and quality control
Filters from sample 1 (stages 6-10) were cut in half and analyzed by laser ablation and wet-chemical ICP-MS (Agilent Technologies 7900).For wet-chemical analysis, standards were prepared by diluting certified, traceable, inductively coupled plasma-grade, single-element standards.Filters were placed in metal-free HDPE vials containing 10 mL of an acid mixture (5 % HNO 3 and 2.5 % HCl, v/v), mixed by a rotary shaker for 12 h, and centrifuged.Extracts were measured without dilution.Recovery rates (85-105 %) were measured using NIST SRM 2783 as a reference standard.Good agreement was observed between wet-chemical and laser ablation ICP-MS for As, Fe, Mn, Pb, V, and Zn (R = 0.94-0.99;m = 0.83-1.52).Poor agreement was observed for Cu, Ni, and Sb, attributed in part to low concentration levels in the extracts (low µg L −1 ) (Table 3).
3 Results and discussion

Elemental image maps
Unlike wet-chemical ICP-MS (a bulk technique), LA-ICP-MS allows for microanalysis of solid samples.In this work, each laser pulse provided spatially resolved elemental data (in counts per second associated with a certain m/z value).Using ImageJ software (Schneider et al., 2012), these data were mapped into pixels.The pixels formed elemental image maps with pixel size P size (µm 2 ): where ss is the laser scanning speed (µm s −1 ), t acq is the ICP-MS acquisition time (s), and d is the laser beam dimension (µm).To obtain maps with square pixels we selected a scanning speed and an acquisition time so that ss × t acq = d; hence, P size = d 2 .In this work, P size = 150 × 150 µm 2 (ss = 300 µm s −1 , t acq = 0.5 s, d [square mask] = 150 µm).
For an analyzed area A anal (cm 2 ), the number of pixels P is given by 10 8 × A anal /d 2 .For our half filters, A anal was nominally 2.4 cm 2 or 10 677 pixels (1 cm 2 = 4444 pixels).Image maps for Pb are shown in Fig. 1a (sample 1, rotated filters of stages 7-10) and Fig. 1b (sample 3, non-rotated filters of stages 6-9).(See Figs.S1 and S2 in the Supplement for maps of other elements in samples 1 and 3, respectively.)The false image map colors in Fig. 1 are brightness values, where each pixel i with element intensity I i is converted by ImageJ to a brightness value B i (dimensionless) using 65 536 pseudocolors (or 65 536 levels of gray).

Visual inspection of elemental maps
Elemental image maps offer a robust tool for observing details about particle deposition not detectable by bulk methods.For example, the 2-D image maps in Fig. 1a illustrate the concentric circles created by MOUDI rotation.A clogged nozzle is apparent in stage 7, where minimal deposition is observed near the center of the half filter.Other images made apparent a thumb print, a scissors cut, and the edge of the mounting tape.The ability to "see" such errors made it straightforward to avoid these parts of the filter when selecting areas to analyze.The maps also offered insights into the deposition patterns of various elements.Most notable was Ni, which, unlike other elements, deposited in mounds even when collected with rotation (Fig. S3).
Additional information can be gleaned from 3-D image maps.For example, 3-D maps of non-rotated filters gave direct evidence for particle bounce (Marple et al., 2014).Relative intensities, measured with ImageJ, indicated that 12 % of the signal for Pb (stage 5) was located between spots (Fig. S4).Bounce was less pronounced in stages 6-10 (due to more nozzles).Three-dimensional maps also made apparent "spikes" in the data, defined as values more than twice the median.Spikes were observed for most elements in sample filters (rotated), filter blanks, and gas blanks.In each case, outliers were replaced with the median of the pixels in the surrounding area (2 × pixel size).This correction was made in all filters (including blanks) except for non-rotated filters, where the much higher concentrations in the spots masked the spikes.Spike removal is illustrated in Fig. S5 for Zn (stages 6-10).After spike removal, the average relative standard deviation of the mapped area decreased from 300 to 30 %.

Statistical analysis of elemental maps
The elemental brightness values associated with each pixel were analyzed statistically using ImageJ (after spike removal).First, we investigated how small an area could be ablated and still reproduce the mean half-filter value B half .We measured (for rotated filters) mean elemental areal brightness values, B areal (cm −2 ) = B i /A anal , for successively smaller areas A anal (generally rectangular shapes, in different sections of the half filter) and compared them to B half .For stages 6-10, 10 % of the total filter deposit area (0.5 cm 2 , 20 min ablation time) gave brightness values in good agreement (±10 %) to the half-filter values.Representative results for Pb (sample 1, stages 7-9) are shown in Fig. 2. For stages 1-5, 20 % of the filter deposit area (1 cm 2 , 40 min ablation time) gave similar results.A notable exception was Ni, which had irregular deposition (Fig. S3).For non-rotated filters (without spike removal), good agreement (±10 %) to B half was observed when (a) at least 15 % of the total deposit area was ablated in stages 7-10 and (b) areas containing 5 spots (of 40) and 8 spots (of 80), respectively, were ablated in stages 5 and 6.In stages 3 and 4, where individual spots were ablated, we measured only the relative standard deviations across spots: 25 % (n = 5) and 23 % (n = 9), respectively.

Elemental concentrations
Like wet-chemical ICP-MS, the ultimate goal of LA-ICP-MS is to measure elemental concentrations.Elemental image maps were converted to concentrations using the NIST standard.For rotated filters, B areal (cm −2 ) was converted to a mass density M areal (ng cm −2 ) using B areal,NIST and M areal,NIST (Eq.2).B areal values were blank corrected using the average elemental areal brightness value of seven clean blanks B areal,BL : Atmospheric elemental concentrations C air (ng m −3 ) were determined by multiplying M areal by the filter exposure area (4.91 cm 2 ) and dividing by the air volume (43.2 m 3 ).
Although half-filter areas were used in these calculations, as shown above, 10 % areas (stages 6-10) or 20 % areas (stages 1-5) gave comparable results (±10 %).A corresponding approach was used for non-rotated filters, except that spikes were not removed.Also, in stages 3 and 4, elemental brightness values were determined per spot rather than per area, then multiplied by the total number of spots per filter.Atmospheric elemental concentrations were highest in MOUDI stages 5-9 (1.0-0.1 µm); these concentrations are shown in Fig. 4a (Ljubljana, samples 1 and 2) and 4B (Martinska, samples 3 and 4).(See Tables S1 and S2 for all concentrations.)Elemental concentrations (ng m −3 ) are shown on the left; percentages (normalized to 100 %) are shown on the right.To facilitate comparison, the two Ljubljana samples (24 and 72 h) and the two Martinksa samples (both 24 h) are plotted side by side.Together, these graphs illustrate both the magnitude and relative contributions of the nine elements in each stage.Several trends are worth noting.First, at both sites, the largest concentrations were observed in stage 5 (cut point = 1 µm).In Ljubljana, the 24 h concentrations were generally greater than the 72 h values (except for stage 5), but the relative percentages in each stage were quite similar.These trends suggest a common major source for the elements, but one that varies in magnitude from day to day.Consistent with previous works (Grgić et al., 2009;Hitzenberger et al., 2006;Mirage, 1989;Pacyna and Pacyna, 2001), we attribute this source to traffic emissions.Fe and Zn were the major elements comprising 85 % of the total elemental mass in sample 1 (101 ng m −3 Fe; 44 ng m −3 Zn) and 90 % of the total elemental mass in sample 2 (127 ng m −3 Fe; 27 ng m −3 Zn).The other trace elements (e.g., Cu, Pb, V, and Mn) were also consistent with traffic emissions and vehicle exhaust or fossil fuel or biomass combustion (Mirage, 1989;Pacyna and Pacyna, 2001).We note the excellent sensitivity that was observed in detecting trace metals with 24 h concentrations as low as 20 (±0.2), 22 (±0.2), and 26 (±0.1) pg m −3 for V (stage 9), Mn (stage 10), and As (stage 10), respectively.
In Martinska (Fig. 4b), the total elemental concentrations were lower than in Ljubljana by roughly a factor of 2. The highest values were in stage 5 (36 ng m −3 ), predominated by Fe (92 %) with smaller amounts of Zn (3 %), Mn (2 %), Pb (1 %), and V (1 %).In general, higher total concentrations were observed in sample 4 (day 2), and there was more variability in composition between the 2 days than in Ljubljana.The largest variability was observed for V in stage 8; concentrations varied from 1.92 (39 %) in sample 3 (day 1) to 1.36 ng m −3 (14 %) in sample 4 (day 2).Vanadium has been observed previously in marine aerosols (Turšič et al., 2006) and is attributed to continental pollution from oil combustion (Tolocka et al., 2004).The variability in the direction of continental winds on the 2 days of sampling may have influenced this signal.

Conclusions
In this proof-of-concept paper, we have demonstrated the usefulness of LA-ICP-MS as a tool for analyzing the elemental composition of size-segregated atmospheric particles collected on filter-based media.Previous problems associated with LA-ICP-MS were addressed: (1) MOUDI rotation sampling overcomes the lack of uniformity in particle deposition, creating a sample highly suitable for LA-ICP-MS 2-D mapping; (2) the 2-D mapping mode yields results which show a high degree of accuracy when larger areas are ablated and superior detection limits; and (3) quantification problems due to non-matrix matched standards are circumvented by ablating through the filter or obliterating the particles on the filters, warranting the reliable use of one-point calibrating on NIST SRM 2783.Together, these improvements allowed for an efficient and sensitive measurement of elemental composition.Although half filters were analyzed in much of this work, we showed that comparable results could be obtained by ablating only 1 cm 2 of filter or less.The ability to analyze a filter in roughly 40 min of instrument time makes feasible routine measurements of size-segregated par- ticles.Compositional graphs of these particles, such as those shown in Fig. 4 for Ljubljana and Martinska, will be useful to the atmospheric community by allowing comparison of elemental profiles of particulate collected at diverse sites (e.g., urban industrial centers to remote background locations).Such profiles can be compared over days, months, or even years; short-term and long-term compositional changes can be used to monitor atmospheric changes such as a new pollution source, the impacts of pollution remediation, and the effects of climate change.A key limitation to this approach is the lack of a size-segregated reference standard; hence, measurement of absolute elemental concentrations is not yet feasible.Nonetheless, much can be learned from relative changes in elemental composition, which are easily measured by this technique.

Figure 1 .
Figure 1.Setups for LA-ICP-MS of rotated (a, sample 1, stages 7-10) and non-rotated (b, sample 3, stages 6-9) filters; both optical and elemental images (pseudocolored image maps of Pb) for each filter are shown.Lab blanks (LB) and the NIST 2783 standard (STD) were analyzed next to the filters.

Figure 2 .
Figure 2. Comparison of mean brightness values for smaller ablation areas (B fraction ) to the half-filter value (B half ) for Pb in sample 1 (stages 7-9).The blue bar shows that ablation of smaller areas agreed with the half-filter value to ±10 %.

Figure 3 .
Figure 3. Relative standard errors (RSE = RSD / √ P ) as a fraction of the number of pixels P (= 21 822 pixels, associated with a total filter deposit area A tot of 4.91 cm 2 ).Solid lines are theoretical values for RSD = 30, 40, and 60 %.Markers are experimental values for Pb in stages 4 (filled squares), 5 (filled triangles), and 7 and 8 (open squares and circles).

Table 1 .
Optimized operating conditions for the laser ablation ICP-MS system.No. of line scans/mapping 70-150 (0.56-4.38 min sequence line scan −1 ) Isotopes measured 75 As, 63 Cu, 57 Fe, 55 Mn, 60 Ni, 208Pb, 121 Sb, 51 V, 66 Zn Table2.Detection limits (DLs; 3 × SD) were determined from the mean standard deviations of seven clean blanks with ablation areas of 1.1 cm 2 .DLs were converted to concentrations using the NIST standard and a theoretical air sampling period of 24 h at 30 L min −1 .All values were spike corrected.Units are in ng m −3 .

Table 3 .
Correlation coefficient (R) and slope (m in y = mx) for concentrations measured by laser ablation ICP-MS (x) and wetchemical ICP-MS (y) in MOUDI stages 6-10.Reasonable agreement was observed for the first six elements.