Sr-Nd-Hf Isotopic Analysis of Reference Materials and Natural and Anthropogenic Particulate Matter Sources: Implications for Accurately Tracing North African Dust in Complex Urban Atmospheres

Abstract

We present novel chemical separation protocols for isotopic analysis of low mass aliquots (0.3 mg and 25 mg) of several reference materials and real-world samples of relevance to urban airborne particulate matter (PM) investigations. A high-yielding gravity flow column chromatography scheme was developed for facile and quantitative separation of Sr, Nd, and Hf prior to multi collector – inductively coupled plasma – mass spectrometry (MC-ICP-MS). Because we are interested in isolating and accurately quantitating individual anthropogenic and natural aerosol sources in complex industrial/metropolitan atmospheric environments, laboratory protocols were optimized using National Institute of Standards and Technology Standard Reference Material (SRM) 1648a (urban atmospheric PM), SRM 1633b (coal fly ash), and European Commission standards BCR-723 (vehicular road dust), and BCR-2 (basalt rock standard). Sr, Nd, and Hf procedural blanks from column chromatography were low (averaging only 37 pg, 17 pg, 11 pg, respectively) and recoveries were high (averaging 95%, 82%, and 92%, respectively). A volume-adjustment protocol was established using isotope reference solutions SRM 987 (SrCO₃), JNdi (Nd₂O₃), and in-house Hf standards to dilute the dried samples prior to MC-ICP-MS based on projected uncertainties for low sample masses. ⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, and ¹⁷⁶Hf/¹⁷⁷Hf isotopic ratios in SRM 1648a, BCR-723, and SRM 1633b are reported for the first time that can serve as provisional reference values. The novel method was used to characterize isotopic ratios and elemental abundances in two anthropogenic urban aerosol sources, namely motor vehicles and petroleum refining using airborne fine PM collected in a vehicular tunnel and fluidized-bed catalytic cracking catalysts, respectively. Two other important mineral-rich urban PM sources, namely soil (i.e., resuspended crustal material) and concrete/cement dust (i.e., construction activity) were also characterized. These are the first isotopic measurements in these environmental compartments and were compared with literature data for long-range transported North African dust, which is a prominent summertime PM source in urban regions in southeastern United States. We demonstrate the capability of coupled Sr-Nd-Hf isotopes to uniquely trace different mineral dust sources with overlapping elemental composition (Sahara-Sahel region, local soil, and concrete/cement) and accurately isolate various urban PM sources demonstrating the superiority of isotopic markers over elemental tracers.

Graphical Abstract

1. INTRODUCTION

The Sahara-Sahel region emits an estimated 400 – 2,200 × 10¹² g yr⁻¹ of dust [1, 2], which travels long distances to prominently affect marine and terrestrial nutrient fluxes, cloud formation, Earth’s radiation budget and climate, as well as airborne particulate matter (PM) concentrations in large, industrialized cities in far-flung populated regions of Europe, North and South America, and sub-Saharan Africa [2–6]. Regulatory compliance with air quality standards in cities and strategies to better protect public health require identifying sources and quantifying their individual contributions to ambient PM. Importantly, governments provide exceptions to “exceptional events” that increase PM levels but are beyond regulatory control making it critical to accurately apportion North African dust in urban atmospheres. To date, North African dust has largely been isolated from urban PM mixtures via (i) receptor modeling using elemental analysis [3, 4, 7] and (ii) subtracting background spikes from PM concentrations at populated receptor sites [7–10]. The first approach introduces significant uncertainties because long-range transported North African dust (crustal and distal), local resuspended soil and road dust (crustal and local), concrete dust (non-crustal and local), coal fly ash (non-crustal and local), and other urban mineral-rich dust sources are similar in composition in terms of Ca, Si, K, Ti, Al, Fe, and/or rare earth elements [8, 11, 12]. The second approach is also fraught with difficulties since background PM itself can be influenced by local sources, is often highly variable, and may not always represent a true control [7]. Consequently, existing approaches can discriminate North African dust input only approximately and only when it is substantially higher than the background [8] introducing errors when we quantitatively estimate contributions of various sources to PM levels in metropolitan locations.

An alternate and increasingly popular approach to accurately isolate aerosol sources with overlapping elemental characteristics is to employ isotopes, which has been successful in identifying anthropogenic emitters [13–18]. Analogously, we hypothesize herein that limitations posed by using major and rare earth elements to isolate mineral dust sources dust sources (North African dust, local soil, and construction activities) in urban environments can be overcome by measuring coupled Sr-Nd-Hf isotopes [19]. Additionally, isotopic analysis is expected to accurately apportion long-range transported dust even during low impact days (e.g., during leading and trailing edges of North African dust incursions).

To date, Sr, Nd, Pb, and Hf isotopes have been used for fingerprinting and tracking trans-Atlantic dust principally in remote or rural locations [20–26]. While ⁸⁷Sr/⁸⁶Sr and ¹⁴³Nd/¹⁴⁴Nd ratios have been primarily used to trace the potential source areas (PSA) [27] of dust [24, 28, 29], ¹⁷⁶Hf/¹⁷⁷Hf and ¹⁴³Nd/¹⁴⁴Nd have lately evolved as promising tracers for long-range transported dust with higher content of radiogenic Hf, which has been attributed to variations in source bedrock and loss of zircon component of dust during transport [21–23]. Hence, coupling Hf isotopes to Sr and Nd also is likely to enhance the accuracy of isolating long-range transported natural dust sources. However, only a few investigations have measured Sr-Nd-Hf isotopes simultaneously, and those that have were also in remote environments such as ice sheets and offshore sites [19, 21–23, 29–31]. To our knowledge, coupled Sr-Nd-Hf isotopes have never been characterized in a complex urban environment where aerosols emanate from multiple anthropogenic and natural sources and consequently their ability to trace crustal PM in such environments is unclear. Therefore, we focus on ⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, and ¹⁷⁶Hf/¹⁷⁷Hf, which are strongly influenced by crustal sources [4, 12]. However, since isotopic tracers are only useful if there are measurable discrepancies between local rock/soil and long-range transported dust, quantifying these isotopes in urban soils is also important. Only by demonstrating contrasting isotopic characteristics of urban Sr, Nd, and Hf sources can we establish their isotopes as dust tracers in industrialized environments. However, an optimized analytical method for accurate and precise quantitation of Sr-Nd-Hf isotopes in ambient PM (and some of its sources) is not yet available.

Typically, only sub-milligram masses of aerosols (e.g., PM_2.5 and PM₁₀) are collected for analysis introducing heterogeneities and analytical uncertainties in laboratory analysis including isotopic ratios, reducing the accuracy of speciation and modeled source contribution estimates necessary to improve air quality [10, 32–35]. It is therefore critical to develop protocols that maximize elemental and isotope recovery and concurrently remove interferences to the highest extent possible from sub-mg samples.

Existing chromatographic processes employed to extract and separate Sr, Nd, and Hf from dust and geological samples are complex [36, 37]. They typically require a pressurized flow setup and commonly used TODGA resins require elution acid to be maintained at elevated temperatures (60–70°C) to completely recover Hf [38]. Paradoxically, current methods used to extract Sr, Nd, and Hf from dust have been optimized and verified only for rock reference materials (e.g. BHVO-1, BHVO-2 AGV-1, AGV-2, USNM3529, and BCR-2) [37, 39] or loess [39]. However, the low mass coupled with the complex and distinctive matrix of urban PM might shift chromatographic peaks and elution curves [14, 40], necessitating the rigorous documentation of the applicability of new sample preparation and instrumental procedures that target aerosols. Additionally, efficiencies of existing methods have not been reported for sub-mg samples.

The objectives of this research targeting Sr-Nd-Hf isotopic ratios are to (i) develop a convenient gravity flow chromatography method to preconcentrate samples prior to multi collector – inductively coupled plasma – mass spectrometry (MC-ICP-MS), (ii) maximize precision and accuracy by optimizing both chromatography and mass spectrometry for a ~100-fold change in equivalent sample mass (0.3 mg and 25 mg), and (iii) evaluate performance of the novel method against multiple reference materials and environmental samples with a wide range of matrix chemistries relevant to urban PM. In addition, we fingerprint the isotopic ratios of several major urban PM sources by analyzing the Sr-Nd-Hf isotopic composition of collected bulk samples to establish their ability to isolate North African dust. We characterize the isotopic and elemental abundances of (i) local soil in the vicinity of a large oil refinery that copiously emits Sr, Nd, and Hf, (ii) airborne PM_2.5 from inside an underwater roadway tunnel representing vehicular contributions, (iii) concrete dust (primarily emitted from construction activities and is enriched in Ca and Sr) from two different commercially available dry mixes, and (iv) four different fluid catalytic cracking (FCC) catalysts that are enriched in all rare earths, especially Nd and represent primary PM emitted from petroleum refineries [41, 42].

2. METHODS

2.1. Microwave-assisted sample dissolution

Samples (see Appendix Table A.5 for more details including mass employed) were dissolved in a microwave oven (CEM MARS 6, Matthews, North Carolina) in two stages (first with HNO₃ and HF then followed by H₃BO₃) each at 200 °C and 20 minutes [42]. This procedure solubilized 46 elements: 31 from groups 1–16 (Na, Mg, Al, Si, K, Ca, Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Se, Rb, Sr, Zr, Mo, Cd, Sn, Sb, Cs, Ba, Pb, Th, and U) and 15 rare earths (Y, La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu) without leaving any residue. The resulting solution was divided into two aliquots at 1:15 volume ratio. The smaller aliquot was diluted in ultrapure water to obtain a 2 % nitric acid matrix and directly used for elemental analysis (section 2.2). The larger aliquot was dried down in a Class 100 cleanroom on a hot plate at 80 °C to remove matrix acids, and the solid residue was redissolved in the appropriate acid matrix (3.5 M HNO₃ in this case). The resulting solution was employed for elemental separation prior to isotopic analysis, which is further discussed in section 2.3. All Certified Reference Material (CRM) samples were duplicated by dividing into five batches of 25 mg each before acid dissolution in different vessels following the procedure explained in section 2.2 from which aliquots equivalent to 0.3 mg were pipetted for calibration of low mass samples explained in section 2.3.

2.2. Elemental analysis of aerosol source samples and CRMs with q-ICP-MS

Prior to isotopic analyses by MC-ICP-MS, samples were analyzed by quadrupole ICP-MS equipped with a collision cell pressurized with NH₃ (PerkinElmer® NexION® 300). Instrumental setup and operating parameters are summarized in Appendix Table A.1. Internal standardization was achieved using ⁷⁴Ge (for Na, Mg, K, Mn, and Co), ¹¹⁵In (for Al, Si, Ca, Sc, Ti, V, Cr, Fe, Ni, Cu, Zn, Ga, As, Se, Rb, Sr, Zr, Mo, Cd, Sn, Sb, Cs, Ba, Th, U, and rare earths), and ²⁰⁹Bi (for Pb). Following our earlier aerosol work [42], ²⁷Al, ⁵¹V, ⁵²Cr, ⁵⁷Fe, ⁶⁰Ni, ⁶⁵Cu, and ⁶⁶Zn were analyzed in Dynamic Reaction Cell (DRC) mode using ammonia as the cell gas. The isotopes analyzed with ICP-MS in standard (i.e., with no cell gas) and collision cell mode are summarized in Appendix Table A.9. Complete sample dissolution in the microwave oven and quality control in our instrumental analysis was confirmed by obtaining quantitative recoveries of certified elements (84–113%) in National Institute of Standards and Technology (NIST) SRM 1648a (urban particulate matter) and SRM 1633b (coal fly ash) as well as Community Bureau of Reference BCR-723 (road dust) and United States Geological Survey (USGS) BCR-2 (natural basaltic rock standard) as summarized in Appendix Table A.2. Recoveries of Sr, Nd, and Hf lie in the range of 86–109%, 92–96% and 89–98% respectively. The choice of these reference materials was driven by their close relevance to our research. SRM 1648a is airborne particulate matter collected in an urban area (near St. Louis, MO) with many certified elements. SRM 1633b represents anthropogenic particles emitted from a typical industrial source with high trace metals and rare earth content. BCR-723 is a proxy for particles emitted from near-road surfaces through vehicular and aeolian resuspension. BCR-2 is a geological reference material that has certified Sr-Nd-Hf isotopic ratios making it invaluable for this work. Also note that SRM 1648a, SRM 1633b, and BCR-723 have been extensively used in validating the elemental composition of urban PM e.g., [4, 14, 16, 32, 43].

2.3. Optimization of sample preparation via liquid chromatography

The four rock/dust reference materials mentioned in the previous section were used to evaluate column performance; BCR-2, SRM 1648a, BCR-723, and SRM 1633b. Elemental elution curves with resin chromatography depend on the sample loading flux [40]. Hence, the column diameter needs to be scaled based on the sample size to obtain comparable elution curves for all sample masses. Since the mass of airborne particulate matter (PM) samples vary based on local aerosol concentrations and there is a lower bound for column diameter to avoid wall effects and short circuiting [44], it is not always feasible to obtain columns of necessary diameter to maintain a constant loading flux. Therefore, we developed a column separation scheme that can be universally applied to maximize recovery and well-separate interfering elements. Column separation behavior was monitored for typically low and high PM masses. The minimum mass was 0.3 mg, chosen to simulate 12 μg m⁻³ of PM_2.5 (the current United States’ primary annual average National Ambient Air Quality Standard) collected over 24 hours period at 1 m³ h⁻¹ (i.e., the Federal Reference Method [10]). The maximum sample mass was arbitrarily chosen at 25 mg representing a nearly 100-fold increase that has been used to measure reference materials [45].

Column separation chemistry was evaluated by eluting suitable reagents based on literature data for the separation of elements of interest and their respective interfering species (Appendix Table A.4). Aliquots of the elutant were collected at intervals determined by the required data resolution. Samples were dried and redissolved in the appropriate reagent before loading onto the columns and after separation, but prior to analysis to remove residual acid matrix from the previous chemical processing step. A detailed description of the column separation process along with column cleaning, reagent purification, and apparatus used has been documented in Appendix section A.2, Table A.6 (whole procedure) and Table A.7 (reagent description and purity). All procedures were conducted on a Class 100 workbench inside a Class 1000 certified room. The final chromatography process was performed using four columns as mentioned in subsections 2.3.1 – 2.3.4. Elements of interest in liquid chromatography (Rb, Sr, Ti, Hf, Lu, Yb, Nd, Sm) were analyzed in aliquots collected from column separation processes to aid in column optimization using a Thermo Scientific ICP-MS (iCAP RQ ICP-MS).

2.3.1. Sr cleanup with Sr-Spec resin (Column I)

Sr-Spec resin (50–100 μm size, Eichrom Technologies) was used to extract Sr from the sample aliquot and ensure complete separation of Rb which causes isobaric interference on ⁸⁷Sr. Columns of 0.25 mL volume were made with heat shrink PTFE with 3.5 mm inner diameter, 27.5 mm resin bed height, and 2 mL reservoir volume. Expecting the Sr-Spec resins to selectively retain Sr and Pb in HNO₃ medium [46, 47], a dried sample cut was dissolved overnight in 0.5 mL of 3.5 M HNO₃ solution and passed through the column until Rb was washed out. Sr was eluted with ultrapure water. The Sr cut had Rb/Sr ratio < 2×10⁻⁴ in all our samples and ⁸⁷Rb only contributed < 0.08% on signal for m/z = 87 (Appendix Table A.10). Pb is reported to be eluted by passing ≥ 6 M HCl [38, 46], hence 7 M HCl was passed until complete elution. Sr-Spec resins were discarded after each use. Cleaning, conditioning and loading procedure for this step is detailed in Appendix section A.2.1.

2.3.2. 50W-X8 Cation exchange resins for Sr-Nd-Hf separation (Column II)

The primary column separation employed Dowex 50W-X8 in its hydrogen form, 200–400 μm mesh size resins [48]. Two types of columns were evaluated because these resins expand in relatively low concentration acids [48]; (i) 3 mL heat shrink PTFE with 6.4 mm inner diameter, 9 cm resin bed height, and 15 mL reservoir volume and (ii) 2.5 mL polyethylene with 7.1 mm inner diameter, 6 cm resin bed height, and 20 mL reservoir volume. The sample was dissolved in 1 M HCl + 0.1 M HF. The acid volume varied from 1.5 mL for low sample mass (< 1 mg) to 5 mL for high sample mass (10 mg – 25 mg). The elution process was optimized with 1 M HCl + 0.1 M HF, 1.5 M HCl, 2.5 M HCl, and 7 M HCl based on the separation efficiency, elution time, and elution volume. Eluted fractions were dried down before proceeding to next chromatography step and 30 mL of acid wash (25 mL of 7 M HCl + 5 mL of 2 M HF) cleaned the resins for each subsequent use.

2.3.3. Nd-cleanup with α-HIBA and AG 50W-X4 resins (Column III)

Since Sm and Nd are not expected to be separated in either of the above two column processes [38, 48], we evaluated AG 50W-X4 (100–200 mesh size, hydrogen form, Bio-Rad) resins and α-hydroxyisobutyric acid (HIBA) as the eluting agent for this purpose. A 1.5 mm inner diameter and 24 cm long glass column was prepared with 0.3 mL resin volume. Resins were equilibrated by submerging in 0.225 M α-HIBA overnight (same value used for elution) and maintaining the pH at 4.7 using NH₄OH. Because rare earths are chelated by negatively charged ligands generated upon α-HIBA dissociation [49], a 0.200–0.225 M concentration at pH=4.6–4.7 was selected [49–51] wherein > 85% acid will be dissociated. The pH remained constant after passing through the column indicating sufficient buffer capacity. The pH was monitored before and after every column separation as its close control is required to obtain the desired elution pattern [50]. Resins were discarded after each use and columns were cleaned with concentrated HNO₃.

2.3.4. Hf-cleanup with Ln-Spec resins (Column IV)

Isotope ¹⁷⁶Hf has direct isobaric interference from ¹⁷⁶Lu and ¹⁷⁶Yb, both of which have been reported to be removed with 50W-X8 cation resin [48]. However, elements like Ti are likely to enter the Hf-cut [48], which can affect the transmission of Hf into the mass spectrometer [52], so an additional cleaning step was performed with Ln-Spec resins (50–100 mesh size, Eichrom Technologies). Ln-Spec resins retain both Ti and Hf in absence of HF but Ti elutes in presence of H₂O₂ in HNO₃ or HCl [48, 53] Therefore, the dried down Hf cut from the initial cation column step was dissolved in 3 M HCl overnight and loaded onto an Ln-Spec filled column (7.1 mm inner diameter and 3.5 cm resin bed height). An additional 7 M HCl was passed after loading to further remove Na and any traces of Lu and Yb in the Hf cut. Ti removal was tested with 3 M HNO₃ + 1 % H₂O₂. Loss of high field strength elements (HFSE, Ti, Zr, Nb, Hf, and Ta) has been reported when HCl and H₂O₂ mixture was passed through Ln-Spec resins [53]. Hence, the column was washed with 2 mL of ultrapure water twice. Hf was eluted by passing 1 M HNO₃ + 0.3 M HF and 1 M HNO₃ + 0.5 M HF solutions. Columns were cleaned with a 5 mL wash of 2 M HF followed by a 10 mL wash of 7 M HCl. Bubble formation due to H₂O₂ was minimized by pairing it with HNO₃ instead of HCl and > 7.1 mm diameter columns were used to achieve good separation even in the unlikely case that bubbles appeared.

2.4. MC-ICP-MS setup and data corrections

Purified cuts of Sr, Nd, and Hf were analyzed using a Neptune™ Plus MC-ICP-MS, equipped with nine Faraday cups, each connected to current amplifiers using 10¹¹ Ohm resistors. A desolvation sample introduction system (ESI Apex Ω) enhanced sensitivity by 7 to 10 times over the conventional Scott-type spray chamber. For ¹⁴³Nd/¹⁴⁴Nd measurement, N₂ gas was introduced in the sample introduction passage to reduce oxide formation and analyte loss. Standard solutions were used for tuning and analyzing MC-ICP-MS performance (Sr isotopic standard SRM 987 for ⁸⁷Sr/⁸⁶Sr and JNdi-1 for ¹⁴³Nd/¹⁴⁴Nd). Due to the lack of commercially available solution-based reference material for Hf isotopes, a large batch of in-house reference standard was prepared whose isotopic ratio was measured as ¹⁷⁶Hf/¹⁷⁷Hf = 0.282141 ± 0.000015 (see Appendix Figure A.9) calculated from 12 measurements alongside BCR-2 (¹⁷⁶Hf/¹⁷⁷Hf = 0.282850 ± 0.000008) with identical instrumental settings. These solution-based standards were also used to model uncertainty equations, which were later used to optimize sample volume. BCR-2 (for Sr-Nd-Hf) and BCR-723 (for Sr-Nd) were used to verify the accuracy of our measurements, post column separation. Bias due to fractionation was corrected by normalizing ratios using Russell’s law [54, 55]. ⁸⁷Sr/⁸⁶Sr was normalized with ⁸⁸Sr/⁸⁶Sr = 8.375209, ¹⁴³Nd/¹⁴⁴Nd with ¹⁴⁶Nd/¹⁴⁴Nd = 0.7219; ¹⁷⁶Hf/¹⁷⁷Hf with ¹⁷⁹Hf/¹⁷⁷Hf = 0.7325. Appendix Tables A.2 and A.3 provides more details of setup and analysis methods. Temporal drifts were corrected by applying the standard-sample bracketing (SSB) method using ⁸⁷Sr/⁸⁶Sr = 0.710255 for SRM 987, ¹⁴³Nd/¹⁴⁴Nd = 0.512102 for JNdi-1 and ¹⁷⁶Hf/¹⁷⁷Hf = 0.282141 for the in-house standard. For better comparison between source samples, Nd and Hf isotopic ratios are reported as deviation from ratios measured for chondritic uniform reservoir (CHUR, ¹⁷⁶Hf/¹⁷⁷Hf = 0.282785 and ¹⁴³Nd/¹⁴⁴Nd = 0.512630).

2.5. Uncertainty modelling and sample volume optimization

Sample dilution after chromatographic chemical separation was carefully considered to maximize precision by balancing MC-ICP-MS signal strength and number of individual measurements for each element. Cumulative error propagation during isotope ratio measurements was mathematically modeled by incorporating three major instrumental uncertainty sources [55–57] in Equation 1; electronic noise and fractionation correction (f(r)), blank (P_bk), and long term (external) bias (P_ex) to obtain total uncertainty in measured ratio (R) (see Appendix A.3 for full details):

$$u(R)=u\left(P_{ex}\right)+\sqrt{{(u(f(r)))}^2+{\left(u\left(P_{bk}\right)\right)}^2}$$ (1)

where u(x) represents uncertainty associated with x. An empirical equation for u(f(r)) was obtained from multiple measurements (reported in Figure 2) of reference standard solutions at different concentrations:

$$u(f(r))=\sqrt{{\left(\frac{k_1}{\sqrt{N}}\right)}^2+{\left(\frac{k_2}{N}\right)}^2}$$ (2)

where N is the signal strength, and k₁ and k₂ are empirical parameters obtained by fitting the above equation for each element (see Figure 2 for data fitting). Uncertainty due to blank correction and intermediate precision bias were also calculated [56, 57] and detailed in the Appendix (Table A.11 and Section A.3). This approach helped in optimizing the final volume and dilution factor for each sample to maximize precision allowing 0.5 mL of sample for uptake, signal stabilization, and a margin of safety to prevent sample tube drying. Thus, volume $V=\frac{M}{C}=V_0+F\cdot t$, where F is the sample flow rate (0.1 mL min⁻¹), t is the total analysis time (min), V₀ is the fixed volume (0.5 mL), M is the mass of element of interest (ng), and C is concentration (μg L⁻¹). The voltage signal for a reference isotope was expressed as N = S · M/V, where S is sensitivity. The signal variance stabilized after 3 minutes of analysis and the standard error was proportional to $\frac{1}{\sqrt{t}}$ thereafter, which directly governs the uncertainty (u) of a measurement as a function of total sample analysis time (Appendix Equation A11, Figure A.10). Since both signal strength N and measurement time t can be described as a function of volume V and uncertainty is a function of N and t (Appendix Equations A.3, A.8 and A.10), we can represent uncertainty (u) as a function of volume V. The optimum volume was then found by setting $\frac{\partial u(V)}{\partial V}=0$.

Figure 2.

Measurement uncertainties for ⁸⁷Sr/⁸⁶Sr (left), ¹⁴³Nd/¹⁴⁴Nd (center), and ¹⁷⁶Hf/¹⁷⁷Hf (right) isotope ratios calculated from observed standard deviation. Uncertainty was given by 95% confidence interval (C.I.) = 2×standard deviation/ square root of the number of readings in a single measurement. k₁ and k₂ are empirical constants estimated from best-fit curves and used in Equation 2. Data presented here were generated using analysis of SRM 987 for Sr, JNdi-1 for Nd, and an in-house prepared standard for Hf.

2.6. Samples for characterizing PM sources

Based on previous studies on Houston, Texas [4, 32, 41], four major local sources were identified as potential influencers for Sr, Nd and/or Hf in ambient PM – (i) soil and road dust resuspension, (ii) motor vehicle emission, (iii) FCC catalysts emission from petroleum refineries, and (iv) concrete dust from construction activity. Soil samples were collected at Clinton Drive (latitude 29.73372; longitude - 95.25759), a hyper-industrialized section of Houston, Texas where the United States Environmental Protection Agency Air Quality Index is often in the unhealthy range, and consequently is the target of many investigations e.g., [4, 42]. We collected the soil samples using plastic spatulas and nylon brushes after digging to ~5 cm depth (to remove contamination from very recent surface activities), oven-dried (105 °C for 24 hours in a loosely covered Teflon vessel) and sieved with 40 μm mesh (to avoid large clumps and vegetative matter that are unlikely to represent crustal material in ambient PM) [41] before dissolution. No further attempts were made to characterize the soil samples.

Fine PM emitted by the ~20,000 predominantly gasoline-driven light-duty vehicles traversing the Washburn Tunnel (29.72829; −95.21192) daily was collected along with background PM_2.5 from the ventilation (blower) room [58]. Two sets were obtained (total 4 samples): one from January 3 – February 1, 2013 and the other from February 2 – February 14, 2013 on 47 mm Teflon filters using Rupprecht & Patashnick 2025 Partisol samplers at flow rate of 1 m³ h⁻¹. Tunnel samplers were placed on a raised catwalk sufficiently away from the north entrance/exit (~44 m) to nearly eliminate sampling of ambient/outside air. Background air was sampled from the ventilation room housing a blower that force-convects air into the tunnel. This approach provided realistic estimates of both tailpipe and non-tailpipe automotive emissions under actual driving conditions accounting for heterogeneities in vehicle/engine types, driving habits, maintenance histories, etc. thereby capturing numerous real-world sources of variability. Concentrations measured inside the tunnel were subtracted from the respective ventilation room values yielding the cumulative emission profile of the vehicular fleet traversing the tunnel over 43-days [58].

Three different fresh zeolite catalysts used in FCC units of petroleum refineries, an important economic activity and pollution source in Houston, were obtained from Grace Davison, Columbia, MD (designated as FCC1–FCC3). Due to possibility of catalyst poisoning during cracking [41], we also analyzed a spent or equilibrium catalyst removed by cyclone separators from the FCC unit’s product stream after its prolonged use from a global refining company with local operations (see also Appendix Table A.5). The fresh and spent FCC catalysts are depicted separately due to differences in their elemental composition.

Representative concrete dust was collected from commercially available concrete mix from two different manufacturers (Quikrete and CTS). All bulk samples were sifted through 40 μm polyethylene sieves before dissolution. Aerosol samples were weighed and dissolved as collected on standard 2 μm pore sized Teflon filters commonly used in airborne PM investigations (Whatman Catalog number: 7592–104, PM_2.5 Air Monitoring PTFE) [3, 4, 13, 32, 41–43, 58].

3. RESULTS AND DISCUSSION

3.1. Column chemistry

In contrast to previous cation exchange results with large (≥ 100 mg) rock samples e.g., [48, 59–61], Sr and Nd could not be separated in one step in our environmental samples (Appendix Figure A.1). Even though specific reasons for this result are unknown, it could potentially arise from matrix effects and lower mass loading [14]. Nevertheless, it explicitly demonstrates the need for a new analytical method necessitating initial Sr extraction before separating Hf, Nd, and matrix elements. The manuscript only depicts low equivalent mass (0.3 mg) results for SRM 1648a as a representative example since it most closely simulates our target mass of ambient PM. Results for other matrices (BCR-2, BCR-723, and SRM 1633b) and high mass (25 mg) samples are shown in Appendix Figures A.2–A.5.

3.1.1. Step 1: Sr-Spec columns for initial Sr extraction (Column I)

As shown in Figure 3(a), Sr and Pb were completely retained on the resin when passed with 3.5 M HNO₃ as expected from previous studies [46, 47] whereas Hf, Ti, Fe, rare earths, Rb, and Ca, and other matrix elements were washed out in the first 2 mL. Sr fully eluted with as little as 1 mL of ultrapure water separating from the Nd-Hf fraction. Ba co-eluted with Sr thereby reducing oxide interference for Nd [59]. Pb did not wash out until 7 M HCl was passed, which was not further pursued herein because it is predominantly of anthropogenic origin [4, 13] and is less likely to be a useful crustal dust tracer in urban environments. However, due to the importance of Pb in urban PM investigations, this fraction can be further analyzed when performing research to distinguish its anthropogenic and crustal origins [4, 13, 16, 28, 38]. Similar results were obtained with other samples (Appendix Figures A.2–A.5) demonstrating that Sr-Spec resins comprehensively separated Sr in all cases.

Figure 3.

Elution curves of elements from 0.3 mg equivalent SRM 1648a when passed in sequence through columns (a) Sr- Spec resin, mesh 50–100 μm, (b) 50W-X8 resin, 200–400 μm mesh, hydrogen form, (c) AG 50W-X4 resin, 100–200 μm mesh, hydrogen form with 0.225 M α-HIBA at pH=4.7, and (d) Ln-Spec resin. (See Appendix Figures A.2 – A.5 for facile chromatographic separation of Sr, Nd, and Hf from all other samples).

3.1.2. Step 2: Cation exchange for Hf, Nd, and matrix separation (Column II)

The first 2 mL eluted in step 1 (section 3.1.1) was subjected to cation exchange to separate Nd and Hf. As shown in Figure 3(b), the Hf fraction was completely eluted (but along with Ti) in the first 10 mL of passing 1 M HCl + 0.1 M HF without any detectable Lu and Yb. Several trace elements (e.g., Zn, Cu, Co, Ni, Mn), which are tracers for anthropogenic urban PM sources [15, 16] also eluted in the first 10 mL of 2.5 M HCl rinse but were not directly pursued herein. Another 15 mL of 2.5 M HCl ensured the complete removal of Ca and Fe (eliminating its interference for Hf [48]) and > 90 % removal of Rb. The high mass (25 mg) SRM 1633b sample eluted ~ 25 % of Lu, Yb, Tm, and ~18% of other lanthanides in this initial fraction (Appendix Figure A.5). For all other samples > 95 % of rare earths were eluted in the next 25 mL of 7 M HCl (see Appendix Figures A.2, A.4, and A.5). For high mass samples, all rare earths were eluted in no more than 15 mL whereas low mass samples exhibited a wider tailing until a maximum of 25 mL (cumulative 60 mL in Figure 3(b)). However, in this step, Ti was not separated from Hf, which may bias MC-ICP-MS measurements [48, 62] and Sm co-eluted with Nd necessitating additional steps as described next.

3.1.3. Step 3: α-HIBA chemistry to separate Sm and Nd (Column III)

The isobaric interference of ¹⁴⁴Sm on ¹⁴⁴Nd can be mathematically corrected during MC-ICP-MS [38, 63]. However, since propagated uncertainties can amplify and significantly reduce accuracy for low concentrations as in our case, a third step to chemically separate Sm from Nd was implemented using the final cut from step 2 (35 – 60 mL range in Figure 3). All heavy rare earths and Sm were eluted in the first 3 mL of α-HIBA (Figure 3(c)). Even though the subsequent 4 mL eluted > 99 % of Nd, we collected an additional 1 mL ensuring complete Nd recovery, noting that Pr does not pose any interferences in our measurements. No significant differences were observed in peak locations between different reference materials as no matrix elements were present at this point. Any α-HIBA remaining in the Nd cut was removed by adding Aqua Regia and heating overnight to 105 °C in a closed Teflon vial [64].

3.1.4. Step 4: Ln-Spec resins to separate Ti and Hf (Column IV)

Both Ti and Fe cause inaccurate ¹⁷⁶Hf/¹⁷⁷Hf isotope measurements [48, 52]. Since Fe was removed in step 2, the primary focus in this step was to remove Ti from Hf. This step also removes any remaining traces of the main interferents (Lu and Yb), and a few matrix elements (e.g., Na, Mg, and Al) from slipping into the Hf-cut [48], which may induce measurement bias [38, 59]. The Hf-cut from cation exchange (first 10 mL from step 2 and Figure 3(b)) was initially dissolved in 3 M HCl, passed through Ln-Spec resins, and followed by 7 M HCl to remove the interfering elements. As shown in Figure 3(d), Ti and Sb eluted by passing 1% H₂O₂ + 3 M HNO₃. We did not observe typical orange/yellow-colored Ti-peroxide complexes for the low equivalent mass (0.3 mg) sample, which was however prominent in 25 mg samples and has sometimes used as a visual indicator for Ti elution [48, 53]. Washout (1 mL + 1 mL) with ultrapure water is recommended between reagents to prevent Hf elution before passing 1 M HNO₃ + 0.5 M HF. Although 0.3 M HF is sufficient to elute Hf, a sharper curve was achieved with 0.5 M HF. As an aside, Zr was shown to co-elute with Hf in Figure 3(d), which does not affect Hf isotopes measurements in MC-ICP-MS [48, 62] but severely biases Hf in Thermal Ionization Mass Spectrometry (TIMS) [48, 65].

The entire scheme for separating Sr, Nd, and Hf fractions based on the above data is shown in Figure 1 for convenience and described in more detail in Appendix A (Table A.6). The measured procedural blanks after passing through columns and the acid blanks (not passed through columns) are summarized in Table 1. Blank values in this study are on the lower end of literature reports [14, 20, 31, 36–39, 48, 60, 66] and are 140–1,800 times lower than elemental masses in even the smallest sample mass analyzed (0.3 mg) demonstrating reliable measurements. Analysis of samples post-purification showed significant reduction of isobaric interferences: ⁸⁶Kr, ⁸⁷Rb, and ¹⁴⁴Sm contributed < 0.2 % of the total signal on ⁸⁶Sr, ⁸⁷Sr, and ¹⁴⁴Nd respectively (see example of SRM 1648a in Appendix Table A.10) and ¹⁷⁶Lu and ¹⁷⁶Yb interfered < 0.01 % on ¹⁷⁶Hf, which was crucial to avoid substantial positive errors in ¹⁷⁶Hf/¹⁷⁷Hf analysis [48]. Additionally, Ce/Nd ratio was observed to be less than 0.015 in the Nd cut thus ensuring negligible effect of Ce on ¹⁴³Nd/¹⁴⁴Nd measurements. Recovery of Sr, Nd, and Hf for all CRM samples was greater than 94.5%, 82.0% and 91.8% respectively (see Table 2 for metals recovery from individual CRMs), demonstrating high efficiency of the proposed scheme. These near quantitative recoveries also demonstrate only minor on-column fractionation effects, which is likely to affect lighter elements like Sr and Nd and was corrected by Russell’s correction [67] as discussed in Section 2.4. Elemental analysis of the filter blanks used in PM collection is summarized in Appendix Table A.9.

Figure 1.

The complete flow chart demonstrating the proposed column-resin separation scheme to measure isotopic ratios. The column separation characteristics are discussed in Section 3. The dimensions of each resin column have been presented with parameters D – inner diameter and H – resin height. The pure elution cuts of Sr, Nd and Hf that are used for isotope ratio analysis are highlighted in grey text boxes. Acid matrix of each eluted cut are presented in parenthesis next to important elements in the respective cut.

Table 1.

Average (± standard deviation) procedural blank (dissolution + column chemistry) values measured and comparison with our lower bound sample mass. Data from other studies reporting procedural blanks for Sr, Nd, or Hf are also shown for comparison.

Element	Blank values measured only from acids (pg) (n=3)	Procedural blanks measured (pg) (n=3)	Elemental mass in 0.3 mg SRM 1648a (pg)	Literature reported blank range (pg)	References
Sr	5±1	37±5	65,000	29–400	[14, 20, 31, 38, 39, 60, 66]
Nd	1±0.5	17±3	7,500	1–225	[14, 20, 31, 36, 38, 39, 60, 66]
Hf	3±1	11±2	1,500	10–59	[37, 38, 48, 60, 66]

Table 2.

Recovery (in %) of elements measured post chromatography procedure proposed in this study.

Samples	Mass of sample dissolved (mg)	Recovery in %
		Sr	Nd	Hf
BCR-2	0.3	98.0	95.1	92.2
	25	98.6	96.5	95.1
SRM 1648a	0.3	94.5	96.5	93.9
	25	96.5	96.5	93.1
BCR-723	0.3	95.2	91.3	95.1
	25	99.5	97.0	93.2
SRM 1633b	0.3	98.1	95.9	91.8
	25	99.5	81.5	93.6

3.2. Isotope values measured in reference materials

As shown in Figure 4, isotope ratios measured in BCR-2 were ⁸⁷Sr/⁸⁶Sr = 0.705014 ± 0.000012, ¹⁴³Nd/¹⁴⁴Nd = 0.512633 ± 0.000031, and ¹⁷⁶Hf/¹⁷⁷Hf = 0.282850 ± 0.000008, which agreed with reported values [68] demonstrating the accuracy and precision of the methodology. Note that Sr and Nd isotopic ratios are not certified in BCR-723, but our ¹⁴³Nd/¹⁴⁴Nd ratio (0.512223 ± 0.000014) overlapped with a recently published value (0.512222 ± 0.000016) but not ⁸⁷Sr/⁸⁶Sr [14]. This deviation in Sr ratios cannot be interpreted at this juncture but points to the need for more analyses of BCR-723 to establish an acceptable range of values. Nevertheless, our ⁸⁷Sr/⁸⁶Sr measurements agree with other reference materials (BCR-2 (Figure 4) and NIST 987 (Figure A.9)) [68] demonstrating good quality control.

Figure 4.

Compilation of five duplicate measurements of ⁸⁷Sr/⁸⁶Sr and ¹⁴³Nd/¹⁴⁴Nd for all reference materials (BCR-2, BCR-723, SRM 1648a, SRM 1633b) analyzed in this study. The gray area represents the mean ± 95% confidence interval. Solid vertical lines represent the average depicted in the associated text along with the 95% confidence interval. Dashed vertical lines for BCR-2 represent isotopic data (mean ± standard deviation) from the GeoRem database [68]) and those for BCR-723 are from [14]. These isotopic ratios for 1648a and 1633b and ¹⁷⁶Hf/¹⁷⁷Hf in BCR-723 have not been reported earlier. Difference of means ≤ expanded uncertainties were regarded as not significantly different as per ERM (Application Note 1) [69]. Data for the above graphs are given in Appendix Table A.13.

⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, and ¹⁷⁶Hf/¹⁷⁷Hf ratios were also measured for the first time in SRM 1648a and SRM 1633b as 0.712578 ± 0.000010, 0.511740 ± 0.000021, 0.282395 ± 0.000017; and 0.703045 ± 0.0000011, 0.512591 ± 0.000021, 0.282454 ± 0.000015, respectively. Individual measurements for all four reference materials are depicted graphically in Figure 4 and validate our procedures.

3.3. Isotopic ratios to distinguish crustal and anthropogenic sources in urban settings

Measured isotopic ratios for four sources relevant to greater Houston [4, 32, 42] are shown in Figure 5 (corresponding elemental ratios are in Appendix Table A.8). Since this research was motivated by the need to isolate and quantify North African dust in complex urban atmospheres, these matrices are superimposed on North African aerosol signatures in the Caribbean and North Atlantic [22, 23, 28]. As seen in Figure 5(a), ⁸⁷Sr/⁸⁶Sr well-distinguished PM from the targeted urban sources; (i) local soil (0.712001 ± 0.00008), (ii) motor vehicles (0.7150 ± 0.0006), (iii) FCC catalysts (0.7110 ± 0.0002) and (iv) concrete (0.70805 ± 0.00002). Importantly, Sr ratios vary widely in long-range transported North African dust (0.7117 – 0.7174) presumably due to differences in potential North African source regions [25], sea spray influences [70], and changes in particle size during transport [21, 22]. Consequently, Sr isotopic ratios in North African dust in Barbados overlapped significantly with vehicular emissions and soil and was close to FCC catalysts pointing to difficulties in separating these four aerosol sources solely via Sr isotopes. In contrast, Sr from concrete dust was easily distinguished because it is extremely non-radiogenic (0.70805 ± 0.00002). Measurements for concrete dust agree with literature data for cement (0.7076 – 0.7103) [13, 71].

Figure 5.

Sr-Nd variation (top panel, (a)) and Hf-Nd variation (bottom panel, (b)) in several urban source samples measured in this study. Anthropogenic sources (petroleum refining, motor vehicles, and concrete) and soil resuspension are compared with trans-Atlantic North African dust composition obtained from the literature [22, 23, 28]. Hf and Nd variation are expressed in relative deviation with respect to CHUR (¹⁷⁶Hf/¹⁷⁷Hf = 0.282785 and ¹⁴³Nd/¹⁴⁴Nd = 0.512630). Data for the above graph can be found in Appendix Table A.12.

Further, as shown in Figure 5a, the Nd isotopic composition alone is unlikely to distinguish North African dust, soil, concrete, and motor vehicles because they all exhibited relatively similar ε_Nd. Other research studies have reported ε_Nd in the range of −10.9 to −8.2 for aerosols collected near a densely trafficked site, which bounds our narrower range of −9.5 to 9.0. Similarly, a European study reported a single value of ε_Nd in cement (= 7.8) [13], which is within our measured range of −8.2 to −7.7. FCC catalysts’ Nd isotopic signature was highly non-radiogenic (−28 ≤ ε_Nd ≤ −22) and substantially different from all other matrices analyzed. Ours are the first reports of Nd isotopic ratios in petroleum refineries’ primary emissions, which depict them to be slightly less radiogenic than steel plant emissions (−20 ≤ ε_Nd ≤ −18) [72, 73]. While fresh catalysts (−28.4 ≤ ε_Nd ≤ −25.9) were tightly clustered, the spent catalyst (ε_Nd = −21.8 ± 0.7) migrated towards more radiogenic Nd, attributed to its poisoning [41], which was confirmed by measuring enrichment of V, Co, Ni, Co, Mo, and Sb in comparison to fresh catalysts (see Appendix figure A.11 and figure A.9). This approach demonstrates that Sr and Nd isotopes together can accurately distinguish aerosols emanating from construction (i.e., cement/concrete) and crude oil cracking operations from long-range transported dust, but not North African dust from soil and automobiles. These results motivate the need to monitor another isotope to better separate sources that are enriched in typically crustal elements. We chose Hf for this purpose, which also captures sea spray that tags North African dust as it transports across the Atlantic Ocean, depicted as “seawater array” in blue color on the top left corner of Figure 5b [74] lying far above the North African dust region.

In contrast to the Sr-Nd relationship in Figure 5a, concrete dust (ε_Hf = −18 to −16.5) and vehicular PM (ε_Hf = −20.7 to −20.3) lie closer to one another in Figure 5b appearing below the zircon-bearing sediment array and separating from FCC catalysts (ε_Hf = −14.3 to −12.5). Hf in vehicular aerosols were least radiogenic in comparison to other sources, attributed to its anthropogenic origins from brake linings and autocatalysts as they were also enriched in Zr (26 ± 9 μg g⁻¹), which is strongly chemically associated with Hf [58, 75]. Soil had more radiogenic Hf (ε_Hf = −13 to −12) than concrete dust and vehicular PM but overlapped with FCC catalysts. As seen in Figure 5b, the Nd-Hf combination better distinguishes motor vehicles from North African dust (that scatters around the zircon-free sediment array [21–23]) compared with Sr-Nd in Figure 5a. The higher content of radiogenic Hf in long-range transported North African dust (−11 ≤ ε_Hf ≤ −6) allows its facile separation from all other sources considered that were less radiogenic (−21 ≤ ε_Hf ≤ −13), indicating its ability to well-trace North African dust in urban settings even in the presence of myriad other sources.

3.4. Evaluation of elemental tracers supplemental to isotopic ratios for anthropogenic sources

In addition to using isotopic ratios to distinguish dust sources of interest, we fingerprinted PM source samples for their elemental composition and calculated corresponding enrichment factors with respect to the upper continental crust (UCC) with Ti as the reference to assess potential elemental markers for the sources of interest. As seen in Figure 6, vehicular emissions sampled in the Washburn Tunnel showed enrichment factors > 100 for Cu, Sb, Cd, Mo, Sn, and Se and > 50 for As demonstrating that these seven elements can potentially be used to isolate PM emanating from automobiles. However, Sn has several industrial interferences [32] making it unsuitable as a vehicular tracer and therefore its elemental ratios were not further pursued. Additionally, As and Se were omitted because they are co-emitted by construction activities and coal combustion [32] and hence are not expected to be useful to uniquely track vehicular emissions or concrete dust. This was corroborated using the sample collected in the forced ventilation room i.e., tunnel background representing PM downstream of several industrial sources [58], which was highly enriched in As and Se (enrichment factors > 100 as shown in Figure A.12) demonstrating their non-vehicular anthropogenic sources. In contrast, as summarized in Appendix Table A.8, dominant elemental ratios of Cu, Sb, Cd with respect to both Ti (crustal element reference) and Zn (anthropogenic element reference) validate their ability to uniquely track vehicular emissions [32]. Note that although FCC catalysts exhibited a higher Mo/Zn ratio than motor vehicles the Mo/Ti ratio was substantially enriched suggesting Mo could possibly be a good tracer of vehicular PM_2.5 as we recently reported [32].

Figure 6.

Average enrichment factors (EFs) of elements measured in this study with respect to average upper continental crust (UCC) [78] with Ti as reference for various PM source samples (Houston soil, n=4; vehicular PM, n=2; fresh FCC catalysts, n=3; used FCC catalyst, n=1; and concrete dust, n=2) with DRC-q-ICP-MS. The solid grey horizontal line represents EF = 1 and dashed horizontal blue line represents EF=10 above which enrichment was considered significant. The blue data points show 50<EF<100 which were tagged as being “moderately enriched” whereas the red data points show EF >100 which were tagged as being “strongly enriched” and consequently potential tracers for the respective sources. Supplementary elemental abundance data for this graph has been plotted in Appendix Figure A.6 for reference.

Concrete dust was enriched in eight metals Ca, Cu, Zn, As, Se, Sr, Mo, and Cd as seen in Figure 6. As explained above, Cu, Mo, and Cd were assigned to automobile emissions since they had 10-fold higher enrichment factors in PM collected within the vehicular tunnel (~1,000 compared to ~100 in concrete) and is consistent with other literature reports [58]. Additionally, Zn, As, and Se are emitted by myriad anthropogenic sources negating their use as unique markers for any source [32]. Hence, Ca is the only element that may strongly separate concrete dust and identify anthropogenic Sr from construction activities because Sr is only relatively weakly enriched (~50) in this source.

FCC catalysts are enriched lighter lanthanides (La-Sm and Gd), which are typically used to isolate any influence of refining operations on ambient atmospheric rare earth concentrations [41, 42]. Ratios of La to other light rare earths (Appendix Table A.8) were highest for FCC validating them as strong tracers of Nd originating from refineries, and when coupled with ¹⁴³Nd/¹⁴⁴Nd can effectively remove anthropogenic influences on ambient airborne lanthanoids. The local soil only shows marginal enrichment of vehicular marker metals (Mo, Cd, Pb, Sn, and Sb) and concrete marker metal (Ca) reflective of “contamination” from the 22.7 billion vehicle miles driven by Texans just in November 2019 [76, 77] and the strong construction industry in Houston ($7.3 billion in approved building permits in the year 2019). While Si/Al has been suggested as a potential tracer to isolate local soil from North African dust [4], elemental ratios (Appendix Table A.8) show Si/Al is higher in concrete dust than local soil making the ratio ineffective to trace local soil in our case and potentially in other urban/industrialized environments as well. A comparison between Sr-Nd-Hf isotopic characteristics (see Appendix Figure A.7) and elemental characteristics of Fe, Al, Si, and Ti (see Appendix Figure A.8) of mineral sources (this study) and key potential source areas (PSAs) in Africa additionally show soil is less likely to differentiate itself from other crustal sources solely using elemental measurements. In summary, even though isotopic ratios are able to separate these sources, including elemental analysis is likely to further differentiate and measure the influence of anthropogenic PM emission.

4. CONCLUSIONS

The newly developed ion-exchange chromatography preconcentration method requires a relatively simple laboratory setup and efficiently extracts Sr, Nd, and Hf from low mass samples displaying its applicability to a range of environmental matrices. This study not only demonstrates a novel chromatography scheme but also measured coupled ⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, and ¹⁷⁶Hf/¹⁷⁷Hf isotope ratios for the first time in three reference materials (BCR-723, SRM 1633b and SRM 1648a). Uncertainty equations were modeled with solution-based reference materials to calculate long-term error propagation and establish the volume to which the samples need to be diluted to maximize precision. Sr-Nd-Hf isotopic characteristics of several urban PM sources including motor vehicles, construction, and petroleum refining were also measured for the first time along with local soil. Isotopic characteristics are demonstrated to well-differentiate long-range transported crustal dust (i.e., from the Sahara-Sahel region of North Africa) from local mineral sources of PM (i.e., construction activity and local soil). For example, concrete/cement materials are highly non-radiogenic in Sr allowing them to be distinguished from North African dust. FCC catalysts are extremely non-radiogenic in Nd separating primary petroleum refining emissions from other aerosol sources. Hf-Nd coupling on the other hand separated North African dust from vehicular sources and soil from FCC catalysts which otherwise had overlapping ⁸⁷Sr/⁸⁶Sr ratios. Importantly, we show that while the elemental composition of crustal dust sources (North African dust and local soil) is not easily distinguishable, their Sr-Nd-Hf isotopic compositions can be used to explicitly isolate them. However, inclusion of elemental tracers is more likely to distinguish metal rich anthropogenic sources. Thus, we recommend both elemental and isotopic measurements to better isolate myriad natural and anthropogenic sources critical for accurately apportioning primary ambient PM in complex atmospheres.

Reported elution characteristics facilitate quantitation of PM emitters that closely overlap in major and trace elemental composition (e.g., North African dust, construction activities, and local soil that are dominated by Ca, Si, Al, Fe, Mg, and Ti as well as North African dust and soil that share identical rare earth signatures) even during periods of low dust activity. Isotopic analysis enables rigorous differentiation of long-range transported dust from otherwise colinear local sources enabling accurate source apportionment thereby allowing environmental agencies to refine and implement compliance strategies with respect to air quality standards. For example, regulatory exceptions can be negotiated in case local PM levels exceed federal or state standards under the “exceptional circumstances” category of the United States’ Clean Air Act by subtracting apportioned North African dust from measured total PM concentrations (and analogously in Asia and Europe).

Table 3.

A comprehensive list of tracers established in this study as well as tracers reported in the literature [4, 22, 23, 41, 58].

Sources	Isotopic tracers			Elemental tracers
	87Sr/86Sr	εNd	εHf
Houston soil	0.71201 ± 0.00008	−10.3 to −11.5	−13 to −12	Si, Al, Ca, Mg, Ti Fe, K
Motor vehicles†			−20.7 to −20.3	Sb, Cd, Mo, As, Cu
Petroleum refining FCC units		−28 to −22		La, Ce, Pr, Nd, Sm, Gd
Concrete dust	0.70805 ± 0.00002			Ca, Sr, Al, Si, As, Se, Cd
North African dust (literature)	0.7117 – 0.7174	−9 to −13	−11 to −6	Si, Al, Ca, Mg, Ti, Fe, K

^†Note that rhodium, palladium, and platinum are excellent elemental tracers for motor vehicles but have been omitted since they were not measured in this study.

HIGHLIGHTS.

• Optimized column chromatography for Sr-Nd-Hf recovery

• Novel protocols for accurate/precise isotopic measurements from low-mass samples

• First isotopic analysis and fingerprinting of 7 reference materials and urban dust sources

• Sr-Nd-Hf isotopes accurately isolate numerous crustal/mineral dust sources