A Global Assessment of Copper, Zinc, and Lead Isotopes in Mineral Dust Sources and Aerosols

The stable isotope compositions of Cu and Zn in major geochemical reservoirs are increasingly studied with the aim to develop these isotope systems as tools to investigate the global biogeochemical cycles of these trace metals. The objectives of the present study were (i) to expand the range of Cu, Zn, and Pb isotope compositions of mineral dust by analyzing samples from major mineral dust sources in Asia and Africa (Chinese Loess Plateau, Chinese deserts, Thar desert, Sahel region) and (ii) to assess the potential impact of human activities on the isotope composition of aerosols by synthesizing published Cu and Zn isotope compositions in aerosols and natural and anthropogenic sources. For the newly analyzed mineral dust areas in Asia and Africa, δ65CuNIST−976 values range from −0.54 to +0.52‰, δ66ZnJMC−Lyon values from −0.07 to +0.57‰, and 206Pb/204Pb values from 18.522 to 19.696. We find a significant geographic control with samples from the Thar Desert having the heaviest isotopic compositions (δ65CuNIST−976 = +0.48 ± 0.06‰, δ66ZnJMC−Lyon = +0.49 ± 0.11‰) and samples from the Sahel and the Badain Jaran desert having the lightest Zn isotope composition (δ66ZnJMC−Lyon = +0.19 ± 0.15‰ and +0.07 ± 0.07‰, respectively). We find important variations in the isotope signatures between particle size fractions with heavier isotopic compositions in the smallest and largest particle size fractions and lighter isotopic compositions in the mid particle size fractions. Associations with the mineralogical composition are less clear. Newly analyzed aerosol samples for Beijing and Xi'an show δ65CuNIST−976 values of +0.29 ± 0.19‰ and +0.16 ± 0.04‰, δ66ZnJMC−Lyon values of −0.36 ± 0.04‰ and +0.02 ± 0.06‰, and 206Pb/204Pb values of 18.129 ± 0.003 and 18.031 ± 0.003, respectively. Based on a synthesis of published and novel data, we suggest improved ranges and mean values for the isotopic composition of mineral dust from selected locations in Asia and Africa and of anthropogenic sources such as non-exhaust traffic emissions, combustion, electroplating and galvanization. This should serve as a valuable reference for future studies using these isotope systems. This paper demonstrates univocally that human activity introduces a wide range of Zn isotope compositions into the atmospheric environment and, thus, impacts the biogeochemical cycle of Zn.


INTRODUCTION
Earth system processes and human activity influence the global biogeochemical cycles of copper (Cu) and zinc (Zn) in a variety of ways. For example, atmospheric long-range transport and deposition of mineral dust and particulate matter delivers these micronutrients to remote marine and terrestrial regions, controlling ecosystem development, and hence influencing the global carbon cycle (Mahowald et al., 2005). However, both metals are toxic at higher concentrations for oceanic microorganisms or when inhaled by humans, they induce adverse health effects. Zinc is associated with oxidative stress, which is a contributing factor in many chronic diseases (Prasad et al., 2004). Adamson et al. (2000) reported that the soluble Zn concentration in atmospheric dust determined the acute toxicity and caused inflammation of lung cells. Copper causes damage to molecules such as proteins and lipids (Brewer, 2010). Both Cu and Zn in urban atmospheric particles (Beijing) are present to a high percentage in easily leachable fractions and may therefore be environmentally mobile and bioavailable . Identifying their sources in atmospheric aerosols remains therefore critical for controlling emissions and improving air quality in urban environments (Moffet et al., 2008).
Trace metals in aerosols originate from anthropogenic and natural sources. Natural sources include soil and mineral dust particles, volcanic dust and gas, and biogenic particles (e.g., Moreno et al., 2009;Duan and Tan, 2013). Desert winds aerosolize several billion tons of soil-derived dust each year, including concentrated seasonal pulses from Africa and Asia (Kellogg and Griffin, 2006) and, consequently, natural reservoirs are important sources for aerosol particles. Even urban aerosols can be composed of a large proportion of mineral particles. In northern Chinese cities, including the capital Beijing, mineral dust particles play an important role especially in spring during so-called "Asian dust storms" (e.g., Mori et al., 2003;Xie et al., 2005;Schleicher et al., 2010Schleicher et al., , 2011Sun et al., 2010;Yu et al., 2011;Chen et al., 2016). Asian dust plays also an important role on a global scale. For example, Uno et al. (2009) showed that Asian dust was transported one full circle around the globe. Ferrat et al. (2013) calculated dust fluxes numerically and estimated annual emissions of 179 Mt/y from Chinese dust sources. Xuan and Sokolik (2002) calculated annual mean PM 10 emission rates that are ∼0.38 t/ha for the Taklimakan Desert, 0.24 t/ha for the Central Gobi Desert, and 0.05 t/ha for the deserts located on the Alxa Plateau. They calculated an annual mean dust emission of PM 10 of 8.4 million tons. More than half of the total annual dust is emitted in spring (Xuan and Sokolik, 2002).
Examples of anthropogenic aerosol sources include emissions from fossil fuel combustion (largely used in energy production), industrial processes, domestic heating and cooking, exhaust and non-exhaust traffic sources like brake and tire abrasion, re-suspension, construction sites, and biomass burning (e.g., Moreno et al., 2009;Colbeck and Lazaridis, 2010;Calvo et al., 2013;Fuzzi et al., 2015). The chemical complexity and variability of the inorganic urban aerosol cocktail implies that the air pollution signature for each city needs to be individually characterized (Moreno et al., 2006) and that we need appropriate tools to fully trace the respective sources. However, source identification remains a challenging task because the methods developed so far (largely based on multivariate statistics) are often inconclusive and expensive (computationally and analytically).
A promising alternative is to use the stable isotope signatures of metals as a source apportionment tool. Variations in the isotopic composition of elements are widely used to identify and quantify source contributions and geochemical processes, which in turn improves our understanding of their biogeochemical cycles. Limitations in mass spectrometer designs, however, restricted the application of this approach to radiogenic and light stable elements until the development of multiple collector inductively coupled plasma source mass spectrometry (MC-ICP-MS) (e.g., Wiederhold, 2015). With respect to Cu and Zn, work on exploring their isotope systematics has focussed in particular on biogeochemical cycling in the aqueous environment (Maréchal et al., 2000;Vance et al., 2008;Peel et al., 2009;Little et al., 2014;Zhao et al., 2014;Moynier et al., 2017). Zhu et al. (2002) studied different mass fractionation processes of Cu isotopes (e.g., at high temperatures, redoxrelated, and biological processes). Other studies on Cu isotopes focussed on sulfides and hydrothermal mineralization (e.g., Ehrlich et al., 2004;Mason et al., 2005;Markl et al., 2006) while Zn isotopic ratios were utilized to identify pollutant sources first assessing dust pollution around a mining site in Siberia (Dolgopolova et al., 2006) and later in the River Seine (Chen et al., 2008). Subsequent work analyzed Zn isotope compositions of atmospheric particulates to assess the influence of industrial processes on isotope fractionation Ochoa Gonzalez and Weiss, 2015;Dong et al., 2017). Central to the efforts of using non-traditional stable isotopes for anthropogenic source tracing is the isotopic characterization of potential source materials and an improved understanding of what controls isotope variability at the source and along the pathway to the sink. This includes the effects of mineralogy and particle size fraction (Feng et al., 2009;Ferrat et al., 2011), atmospheric processing (Rubasinghege et al., 2010) and admixing of other sources during atmospheric transport (Ferrat et al., 2013).
Anthropogenic emissions have surpassed natural emissions for many trace elements (Rauch and Pacyna, 2009). This has affected the global atmospheric cycles of these elements and led to the atmospheric deposition in even the most remote areas (Marina-Montes et al., 2020). Anthropogenic processing can induce redox changes for metals that usually occur only in one oxidation state in the environment (Wiederhold, 2015). For example, ore smelting and roasting, the production of elemental metals, or engineered nanoparticles for industrial purposes, and combustion processes can cause isotope fractionation due to redox effects (including Zn), which may be preserved if the process is incomplete and isotopically fractionated metal pools with different oxidation states are released into the environment Mattielli et al., 2009). The main fractionation processes affecting the Zn isotope composition of atmospheric emissions are likely evaporationcondensation processes (Cloquet et al., 2008). The isotopic characterization of anthropogenic atmospheric endmembers has been attempted in several studies, mostly for Zn but also for Cu.
The present study examines associations between observed isotopic compositions in urban aerosols and source reservoirs including mineral dust and anthropogenic materials. We first determine Cu, Zn, and Pb isotope compositions from a new set of major global mineral dust sources in Asia (Chinese Loess Plateau, Chinese deserts, and Thar desert) and Africa (Sahel region) to better characterize the range of isotopic compositions of this natural aerosol source and to assess the importance of geographic variability. We also present new isotope data of sizefractionated mineral dust samples to assess relevance of particle size to aerosol tracing. We then compile from published literature Zn and Cu isotope compositions of possible source materials in aerosols derived from human activity. The main emphasis is on Zn since more Zn isotope studies of anthropogenic materials are published. Finally, we compare isotopic compositions of urban aerosols (data from literature and novel data for two Chinese cities, Beijing and Xi'an) and of major natural and anthropogenic reservoirs to discuss implications for aerosol source tracing.

Location and Description of Samples
We studied a set of mineral dust collected from major global dust sources. Table 1 details the sampling sites of the mineral dust sources considered in this study: The Taklimakan desert of northwestern China (samples TK-074 and TK-103), the Tengger desert (TG-018), the Badain Jaran desert (BJ-024), and the Chinese Loess Plateau of northern China (CLS-20-90 and JY-1), the Thar desert of northwestern India (TDSD 1-3) and the Sahel region between Bamako and Timbuktu (SHD 10-32).
Four samples from the Chinese deserts (TG-018, BJ-024, TK-103, and TK-074) were separated into five size fractions by wet sieving (32-63 µm and >63 µm) and settling in solution [pipette method, see Lerman et al. (1974)] based on Stoke's Law (<4 µm, 4-16 µm, and 16-32 µm). Isotope ratios were determined in all fractions. Concentrations were determined in the lowest, highest and middle size fraction only due to sample limitations. Bulk isotopic compositions and concentrations of the samples were then calculated by mass balance with the concentrations of the missing size fractions approximated by taking the average concentrations of the size fractions above and below.
Two urban aerosol samples were collected in Xi'an and Beijing, China, during the summer of 2011. These samples were collected using the wet collection method [see Yan et al. (2015) for more details]. In Beijing, the collecting devices were placed on the roof (∼35 m above ground) of the Chinese Academy of Sciences, a building surrounded by mixed commercial and residential districts. In Xi'an, collecting devices were placed on the roof (∼45 m above ground) of a building in a residential district.

Sample Digestion of Mineral Dust and Aerosol Samples
All samples were digested in ISO 5 (class 100) laminar flow hoods in ISO 6 (class 1000) clean laboratories using previously developed methods (Ferrat et al., 2012b;Dong et al., 2013). All acids used were of supra-pure grade (HF) or purified by quartz distillation system (HNO 3 and HCl). Dilute solutions of acids were prepared with 18.2 MΩ cm −1 grade Millipore system (Bedford, USA). Laboratory equipment was cleaned with 6 M HCl, 15.4 M HNO 3 and distilled 0.5 M HNO 3 on a hot plate at 120 • C for 24 h. Approximately 50 mg of each sample was digested using 1 ml of 24 M HF and 0.25 ml of 15.4 M HNO 3 in PFA vessels (Savillex, USA). The samples were digested on a hot plate at 150 • C for 72 h. Every 24 h, the samples were put into an ultrasonic bath for 15 min. The samples were dried and redissolved twice in 0.5 ml 15.4 M HNO 3 and once in 0.5 ml 7 M HCl, to remove possible HF remaining in the digested samples. Samples were finally taken up in 3 ml 7 M HCl and separated into aliquots for concentration and isotope ratio analysis.

Determination of Cu, Zn, Pb, and Sc Concentrations and of Enrichment Factors
Concentrations of Cu, Zn, Pb, and Sc were determined by quadrupole inductively coupled plasma mass spectrometry (ICP-MS, Varian, USA; Ferrat et al., 2012b). Sample solutions were dried and re-dissolved in 10 ml 0.5 M HNO 3 . External calibration solutions (0.1-200 µgl −1 in 0.3 M HNO 3 ) were prepared by diluting multi-elemental standard solutions CCS-1 (Inorganic Ventures Inc.) containing 100 mgl −1 Sc and CCS-6 including 100 mgl −1 Pb, and single element Rb standard (VWR Aristar R 10,000 mgl −1 ). Internal standard solutions were prepared using 1,000 mgl −1 Rh and In (Alfa Aesar Specpure R ) standard solutions. The limits of quantification were calculated after each run based on the intensity and standard deviation measurements of the calibration standards and calibration blank and were generally at or below a few tens of ng g −1 for Pb, Sc, Cu, and Zn. These were at least two orders of magnitude below the lowest concentrations measured. Accuracy and precision were based on repeat measurements of the certified reference material USGS G-2 granite. Repeat measurements (n = 6) showed a precision of 10% at 2σ-level for all analyzed elements. Copper, Zn, Pb, and Sc concentration were ∼6% lower than the certified values. Elemental concentrations for Cu and Zn in TK-074 and TK-103 were published previously as a part of the validation process for our improved analytical procedure (Dong et al., 2013).
The elemental enrichment factor (EF) was calculated using where X is the element assessed, scandium (Sc) the reference element, and the upper continental crust (ucc) is the reference reservoir (Taylor and McLennan, 1985).

International Reference Standards and Standard Conversion
The variation of stable isotope ratios in nature is small. The small differences are expressed as delta values (δ) in per mill deviation  Figure 1 in (Ferrat et al., 2011) TDSD-02 Thar Desert Figure 1 in (Ferrat et al., 2011) TDSD-03 Thar Desert Figure 1 in (Ferrat et al., 2011) where R sample and R ref are the sample and reference isotope ratios, respectively. The referencing standards used during measurement sessions in individual laboratories are so called "in-house" or working standards and differ from laboratory to laboratory. Consequently, for inter-comparison, δ-values expressed relative to a working standard need to be converted and reported relative to an international reference standard. The δ-value for a sample (δ sample ) measured against a working standard (δ work ) on an international scale (δ ref ) [see Werner and Brand (2001)] is given by For Zn, four standards have to date been widely used and hence can essentially serve as an "international" reference standard: (i) the Johnston and Matthew JCM 3-0749L single element solution (called "JMC-Lyon") from the ENS Lyon (denoted thereafter as JMC-Lyon standard, Maréchal et al., 1999); (ii) the IRMM-3702 certified isotope standard produced by the Institute for Reference Materials and Measurements (IRMM) in Belgium (denoted as IRMM standard); (iii) the AA-ETH single element solution from the ETH Zürich (denoted as AA-ETH standard, Archer et al., 2017) and (iv) the SRM-683 standard from NIST. The International Atomic Energy Agency (IAEA) recommends using in-house working standard materials instead of international standards during measurements on a daily basis (Werner and Brand, 2001). Arguments including similar properties of the sample and the standard, homogeneity, available quantities, costs for which primary reference material to use as an international standard have been widely discussed (e.g., Brand et al., 2014;Archer et al., 2017;Moynier et al., 2017) and conversion values based on intercalibration efforts (Moeller et al., 2012;Archer et al., 2017) or statistical analysis of published values (Moynier et al., 2017) were proposed. To date, no complete summary of all measured Zn standard values has been presented so we compiled these published values for the different Zn standards in Table 2. We report δ-values relative to the Lyon standard (δ 66 Zn JMC−Lyon ). The rationale was that most of the data used in this study were originally measured against the Lyon standard, hence conversions were kept to a minimum. Furthermore, the majority of publications still report Zn isotope variations relative to the Lyon standard. We used the following conversions values for δ 66 Zn IRMM−3702 , δ 66 Zn AA−ETH , and δ 66 Zn SRM−683 :

Standard:
δ 66 Zn JMC-Lyon (‰): Reference: Average of 24 studies compiled in Table 2 AA-ETH +0.28 (Archer et al., 2017) SRM-683 +0.12 (Yang et al., 2018) Figure 1 displays the relative location of δ 66 Zn of the suggested standards (IRMM-3702, AA-ETH, SRM-683) vs. JMC-Lyon. Exemplarily, we added two generic samples A and B to this graph to illustrate the δ 66 Zn location relative to IRMM and Lyon (see blue solid and red dashed arrows), respectively, to avoid wrong conversions.  The widely used international reference material SRM-976 (elemental Cu) from NIST (denoted thereafter NIST-976) to report Cu isotope data is not anymore commercially available. Two European Reference Materials (ERM by the European Commission) standards (Cu nitrate solutions), ERM-AE-633 and ERM-AE-647 were recently proposed as new international standards (Moeller et al., 2012). The advantage of the ERM-AE-633 standard is that it has almost the same δ-value as the NIST-976 [variations of ±0.01 were reported by Moeller et al. (2012), Šillerová et al. (2017), Kříbek et al. (2018)] and it was proposed that no conversion between δ-values reported relative to AE-633 is necessary (Peréz Rodríguez et al., 2013;Kusonwiriyawong et al., 2017). We report all δ-values relative to the NIST-976 standard and use the following conversion values:

Isotope Ratio Determinations of Cu, Zn, and Pb
Prior to isotope ratio analysis, Cu, Zn, and Pb were separated from matrix elements using ion exchange chromatography procedures previously developed to account for small sample size, low concentrations, and complex matrices (Ferrat et al., 2012a;Dong et al., 2013). For Cu and Zn isotope ratio measurements, separation from elements such as Mg, V, Ti, Ni, Cr, and Ba is necessary because these elements generate interferences with Cu and Zn isotopes in argide form (Mg), (hydro)-oxide form (Cr, Ti, V), or double-charged species (Ba). A Bio-Rad AGMP-1 resin (100-200 mesh) was employed for the separation. To ensure complete separation from the matrix, the Cu fraction was passed through the ion exchange column twice and the Zn fraction once. We tested the yield for Zn and Cu during passage of ion exchange resin with an in-house granite (HRM-24) throughout the course of the study. The column recoveries for the HRM-24 granite were quantitative for both elements: 98 ± 5% for Zn (n = 3) and 96 ± 4% for Cu (n = 3). The Cu and Zn solutions were dried and re-dissolved in 0.1 ml 15.4 M HNO 3 and evaporated to remove the remaining chlorides and finally dissolved in 1 ml 0.1 M HNO 3 for isotope ratio analysis. For the Pb isotope ratio determinations, Pb was isolated with an EiChrom selective extraction Sr-resin following a previously developed procedure (Ferrat et al., 2012a). A Nu plasma source multiple-collector mass spectrometer (MC-ICP-MS, Wrexham, UK) connected to a Nu DSN-100 Desolvation Nebulizer System was used for the isotope analysis. Instrumental mass bias effects were corrected using standard sample bracketing for Cu and double spike for Zn (Arnold et al., 2010). Zinc isotope ratios are reported against the "JCM-Lyon"-standard JCM 3-0749L and Cu isotope composition against NIST-976 standard using the δ-notation: The total procedural blank for Cu and Zn after separation was ∼3.5 and 5 ng, respectively. Accuracy and precision were monitored by repeated measurements of the geological standards BCR-027 Blend Ore and USGS-G2 granite and the industrial single element standards Romil Cu and Romil Zn. Within error, the isotope ratios of the reference materials were identical to previously reported values (see Table S1). Precision was estimated by the average analytical error (±2 sigma) of the mineral dust samples measured in this study, i.e., 0.07‰ for δ 65 Cu-and 0.07‰ for δ 66 Zn-values. Isotope data for Cu and Zn in TK-074 and TK-103 were published previously as a part of the validation process for our improved analytical procedure (Dong et al., 2013).

Mineralogical Determinations (XRD)
The mineral compositions of the samples were determined using X-ray diffraction (XRD). X-ray powder diffraction data were collected using an Enraf-Nonius PDS120 diffractometer equipped with a primary Germanium (111) monochromator and an INEL 120 • curved position sensitive detector. The dust samples were ground in an agate mortar and loaded into a circular well-mount of 7 and 13.5 mm diameter with a depth of 1 mm. Operating conditions for the Cu source were 40 kV and 35 mA. The horizontal slit after the monochromator was set to 0.24 mm to confine the incident beam to pure Cu Kα1 radiation. The angular linearity of the detector was calibrated using the external standards silver behenate (AgC 22 H 43 O 2 ) and silicon (NIST SRM 640). The full 2-Theta linearization was performed with a least-squares cubic spline function. The whole pattern stripping method described in detail elsewhere (Cressey and Schofield, 1996) was used to derive the phase proportions. For mineral quantification, appropriate standards were selected from reference materials of the mineral collection at the Natural History Museum, London, and analyzed under identical run conditions as the samples. In some samples, total mineral composition was below 100 wt.%. This indicated the presence of amorphous components and minor amounts of unidentified phases where low-intense peaks could not be assigned to a phase. Quantification errors were assessed varying best-fit scale factors while inspecting the influence on the residual background. The estimated relative error for the mineral assemblages was 5 and 10% for phase proportions above and below 10 wt.%, respectively.

RESULTS
We present in this section the results of the new data set acquired for the samples collected from the Chinese, Indian and Saharan deserts.

Concentrations and Isotope Ratios of Bulk Mineral Dust
Elemental concentrations of mineral dust samples are lowest in the Sahel Region and Thar Desert and highest in the Chinese deserts (Tengger, Badain Jaran, Taklimakan, Gobi). Copper concentrations in the Thar Desert are up to 10 µg/g and in the Chinese deserts up to 69 µg/g, Zn concentrations in the Thar Desert are up to 35 µg/g and in the Chinese deserts up to 203 µg/g. By comparison, upper continental crust Cu and Zn concentrations are 28 and 67 µg/g, respectively (Rudnick and Gao, 2003). Lead concentrations in the Sahel Region are up to 16 µg/g and in the Chinese deserts up to 30 µg/g ( Table 3).
The isotopic compositions of mineral dust samples (bulk and individual size fractions) range between −0.07‰ and +0.57‰ for δ 66 Zn JMC−Lyon , between −0.54‰ and +0.52‰ for δ 65 Cu NIST−976 , and between 18.522 and 19.696 for 206 Pb/ 204 Pb ( Table 3). The isotope ratios of the mineral dust samples from the different regions are shown in Figure 2. The δ 66 Zn JMC−Lyon values from samples of the Chinese deserts range between +0.07‰ and +0.33‰, of the Chinese Loess Plateau between +0.22‰ and +0.49‰, of the Thar desert between +0.41‰ and +0.57‰ and of the Sahel region between −0.02‰ and +0.31‰. The δ 65 Cu NIST−976 values in the Chinese deserts range between 0.00‰ and +0.14‰, in the Chinese Loess Plateau between +0.16‰ and +0.38‰, in the Thar desert between +0.44‰ and +0.52‰ and in the Sahel region between −0.20‰ and +0.11‰.

Concentration and Isotope Ratios of Cu, Zn, and Pb in Particle Size Fractions
Copper concentrations decrease from the smallest to the largest fraction in samples from the Tengger desert (TG-018) and Badain Jaran desert (BJ-024) ( Table 3). The samples from the Taklimakan desert (TK-074 and TK-103) have the lowest Cu concentration in the 16-32 µm size fraction. Zinc concentrations correlate inversely with particle size in all samples, with a range between 206 and 389 µg/g of Zn in the <4 µm size  Previous studies showed significant variations in the radiogenic isotope compositions of Sr, Nd, and Pb between size fractions of natural mineral dust (Feng et al., 2009). Similarly, we find significant variations in the stable Cu and Zn isotope compositions of the different size fractions compared to analytical precision (Figure 3, Table 3). For Zn, these variations are up to 0.35‰, for Cu up to 0.97‰. For 206 Pb/ 204 Pb we observe variations up to 0.22. The larger variability found in Cu compared to Zn is in agreement with the greater variability found for the isotopic composition of Cu in rocks and minerals in general, likely partly controlled by the variable redox states of Cu in nature (Markl et al., 2006). The isotopic composition vs. particle size fraction plots (Figure 3), excluding the δ 65 Cu values of two samples (TK-074 and BJ-024, discussed below), are similar for Cu and Zn, i.e., a U-shaped curve with the most positive values predominantly in the smallest and largest particle size fractions and more negative values in the mid particle size fractions (Figure 3).

Mineralogical Composition
The mineral compositions were determined in selected bulk samples from the Gobi Desert (MG-09022), the Chinese Loess . Also shown are the maximum differences between highest and lowest values for each element and sample ( ). The error bars for Pb are within the symbol size. Copper and Zn isotope data of TK-074 and TK-103 were published previously as a part of the validation process for our improved analytical procedure (Dong et al., 2013).
The correlation between isotopic composition, concentrations and mineralogy was assessed using multivariate regression of the entire data set available (n = 17). Table 4 shows the results for the Pearson product-moment correlation coefficient, r; which varies from 1 (perfect positive correlation) to 0 (no correlation) to −1 (perfect negative correlation).

DISCUSSION
The discussion section is organized the following way: First, we discuss the range of Zn and Cu isotope compositions in natural reservoirs compiled from new and published data ( Table S3). Mineral dust is an important natural aerosol source since several billion tons of soil-derived dust is aerosolized each year, including seasonal pulses from Africa and Asia (Kellogg and Griffin, 2006). Second, we complile and discuss isotopic compositions of potential aerosol source materials derived from anthropogenic activity reported in the literature (Table S3). Anthropogenic Zn and Cu sources are important since anthropogenic atmospheric emissions have surpassed natural ones (Rauch and Pacyna, 2009) and anthropogenic sources are manifold, especially in urban areas. Third, we compile δ 66 Zn JMC−Lyon and δ 65 Cu NIST−976 isotope values of urban aerosols from literature and add new data from two Chinese cities. We compare the isotopic composition of aerosols from different cities and discuss the possible influence of the major natural and anthropogenic reservoirs.

Zinc and Cu Isotope Signatures in Mineral Dust and Other Natural Reservoirs
We observe important geographical variations in the isotopic composition of mineral dust (Figure 2). The samples from the Thar desert display the heaviest Zn and Cu isotope compositions with δ 66 Zn JMC−Lyon = +0.49 ± 0.11‰ and δ 65 Cu NIST−976 = +0.48 ± 0.06‰ (n = 2). In contrast, the samples from the Sahel region and the Badain Jaran desert exhibit the lightest Zn isotope compositions (δ 66 Zn JMC−Lyon = +0.19 ± 0.15‰ (n = 3), δ 66 Zn JMC−Lyon = +0.07‰ (n = 1), respectively). The mineral dust samples of the Chinese Loess Plateau are heavier with respect to their Zn and Cu isotope compositions than those of Chinese deserts. While the isotopic variability of the different size fractions results in a significant spread (Figure 2, Taklimakan, Tengger, and Badain Jaran), the bulk values of the samples collected from each geographical region remain within a small range of 66 Zn max−min = 0.23‰ and 65 Cu max−min = 0.14‰, as shown in Figure 2. The 206 Pb/ 204 Pb values for the Chinese deserts and the Loess Plateau fit well in the range of 206 Pb/ 204 Pb values of 18.648-18.786 derived from a peat bog and proposed to serve as "Asian natural background" by Ferrat et al. (2012b).
Calculated enrichment factors for Zn and Cu are below five for all samples (see Table 3), and, thus, we suggest that deposition of metals from anthropogenic emissions from long range transport in these regions are negligible. There are slight differences between the dust source regions: Chinese dust samples show slightly higher EFs with average values of 2.2 and 1.3 for Cu and Zn, respectively, in the Chinese deserts, and 1.2 and 1.1 for the Chinese Loess Plateau. Both, the Sahel region (average of 0.7 and 0.9 for Zn and Cu, respectively) and the Thar desert (average of 0.5 and 0.4 for Zn and Cu, respectively) have EFs below 1 and, thus again, no anthropogenic influence is assumed. Figure 4 shows bivariate plots for the isotope ratios of the three elements (Zn, Cu, and Pb) measured. The Zn vs. Pb and Cu vs. Pb plots separate the samples from the Thar desert well. The samples from the Sahel region, from the Chinese desert and the Chinese Loess Plateau tend to be clustered together, while those from the Thar desert plot toward the more radiogenic Pb.
We do not find a clear mineralogical control on the Zn and Cu isotope compositions of mineral dust. A link between the enrichment of light isotopes and clay minerals was observed for Cu and Zn in altered and unaltered hydrothermal deposits (Markl et al., 2006) and in soils (Bigalke et al., 2010). However, we observe similar Zn and Cu signatures for samples with dominant or lesser contributions of clay minerals, such as the samples from the Chinese deserts and the Thar desert. Overall, the multivariate regression analysis of our limited data set suggests no significant associations between mineralogy and Zn or Cu isotope composition (Tables 4 and S2). By contrast, Pb isotope compositions seem interrelated with quartz and smectite abundance, suggesting a possible link between the mineralogy and radiogenic isotope composition of mineral dust.
We observe significant isotope variations between the different size fractions of mineral dust as described in section Concentration and isotope ratios of Cu, Zn, and Pb in particle size fractions. Most samples (except δ 65 Cu NIST−976 values of TK-074 and BJ-024) display a U-shaped curve (Figure 3) with the highest δ-values in the smallest and largest particle size fractions and lower values in the mid particle size fractions. The different size fractions for Taklimakan reveal systematic changes for quartz (i.e., an increase from the fine to the coarse fraction) and clay minerals (i.e., a decrease from the fine to the coarse fraction) but not for the other fractions.
Enrichment factors for all mineral dust samples are low (<5) and we do not assume an influence from anthropogenic sources. However, we caution that any minor anthropogenic influence may disproportionately impact the smaller size fractions, due to the potential for long-range transport of fine particles. Both Taklimakan samples show the highest EFs for the smallest size fraction (<4 µm), lower for the medium size fraction (16-32 µm), and smallest for the largest size fraction (>64 µm).
The differences in isotopic ratios between size fractions may have important implications for source apportionment studies as long-range transport of small particles may cause shifts in the isotope signatures of dust in the atmosphere. Mori et al. (2003) showed that the particle sizes changed during long-range transport from China to Japan but that the chemical composition of crustal elements remained nearly constant. But if Cu and Zn isotope compositions vary between different size fractions of the source material, as our new data suggests, then we can expect changes in the isotopic composition along a transect from the source to the sink due to particle size (in addition to fractionation processes during atmospheric transport).
For the Cu isotope compositions in natural reservoirs as shown in Figure 5 (right panel), it is important to note that we include only the products of "primary mineralization." The complete natural range is considerably broader, with products of secondary mineralisation (including products formed during low temperature supergene processes in ore deposits or during metamorphic processes) spanning a considerably broader range of −16 to +10‰ (Mathur et al., 2009;Moynier et al., 2017). Mineral dust δ 65 Cu NIST−976 values are very similar to those of most common parent rocks and, thus, the Cu isotope composition of mineral dust looks like that of igneous/clastic sedimentary rocks. Figure 6 compiles the Cu and Zn isotope compositions of anthropogenic reservoirs. More details can be found in Table S3. We divided the anthropogenic reservoirs into non-exhaust traffic derived materials (e.g., brakes and tires; filled circles in Figure 6), materials derived from industrial activities (combustion, galvanization, electroplating; open circles in Figure 6), and other sources (e.g., contaminated soils, fertilizers, and cement; gray circles in Figure 6). To the authors' knowledge, no Cu or Zn isotope compositions of other important aerosol source materials, such as biomass combustion, or train, airplane, and shipping emissions have been studied to date.
iii) Some anthropogenic materials are isotopically lighter or heavier than the sphalerite range. These materials were processed either by high-temperature processes/evaporation such as coal combustion or by electroplating and galvanization, thus most likely products from unidirectional fractionation processes. Pulverized coal and a mixture of coal/pet coke combustion led to significant isotope fractionation between the byproducts [the coal fly ash, with δ 66 Zn JMC−Lyon =1.23-1.95‰ (n = 6), and the bottom ash, with δ 66 Zn JMC−Lyon = 0.55-1.13‰ (n = 3)] with respect to the starting materials (δ 66 Zn JMC−Lyon = 1.02-1.47‰, n = 5) (Ochoa Gonzalez and . Electroplating causes large isotope effects due to redox reaction (Kavner et al., 2008). Electroplated materials may constitute isotopically distinct metal sources released to the environment as metal wastes and mobilized via corrosion processes. Kavner et al. (2008)   We make the following observations for the Cu isotope compositions of anthropogenic reservoirs (Figure 6, right panel): i) δ 65 Cu NIST−976 values for anthropogenic reservoirs range from −3.6 to +1.8‰. This range is smaller than the considerably broader natural range of −16 to +10‰ (e.g., Mathur et al., 2009;Moynier et al., 2017) if secondary mineralization (formed during low temperature supergene processes in ore deposits or during metamorphic processes) is included. ii) In comparison to Zn, human activity seems not increase the range of Cu isotope compositions. Possible reason for this might be the higher variability of Cu isotopic compositions in natural reservoirs. iii) Traffic-related sources have δ 65 Cu NIST−976 values ranging from −0.18 to +0.71‰, with a median value of +0.34‰. iv) Contaminated soils spread over a wide range of Cu isotope compositions. For aerosols, soils can be seen as a source (resuspension) or sink (deposition). Most Zn or Cu isotope studies focussed on soils close to (active or historic) mining, refining and/or smelting sites (Figure 6, Table S3) or for Cu also in vineyard soils. Figure 7 shows the δ-values for Zn and Cu in urban aerosols. Studies were conducted in several cities, including Barcelona (Spain, Ochoa Gonzalez et al., 2016), London (UK, Ochoa Gonzalez et al., 2016;Dong et al., 2017), Zaragoza (Martín et al., 2018), and Metz (France, Cloquet et al., 2006). Gioia et al. (2008) as well as Souto-Oliveira et al. (2018 reported Cu and Zn isotope compositions from São Paulo, Brazil. To the authors' knowledge, our data from Beijing and Xi'an provides the first Zn and Cu isotope data for an Asian city. For the Zn isotope compositions of urban aerosols (Figure 7, left panel), we make the following observations:

Comparison of Zn and Cu Isotope Compositions in Urban Aerosols With Natural and Anthropogenic Reservoirs
i) The Zn isotope compositions of aerosols spread over a wide range from δ 66 Zn JMC−Lyon of −1.36 to +0.78‰. This range is considerably broader than the range observed for mineral dust or natural unpolluted soils and is more similar to that of the anthropogenic sources discussed in section Zinc and Cu isotope signatures of anthropogenic reservoirs. Generally, δ 66 Zn values in aerosols tend toward isotopically values even lighter than those identified for the anthropogenic sources (section Comparison of Zn and Cu isotope compositions in urban aerosols with natural and anthropogenic reservoirs). ii) The Zn isotope compositions of aerosols can vary considerably within one city. The highest variability ( 66 Zn max−min = 1.5‰, n = 62) to date was found in São Paulo , which the authors attribute to highly variable source contributions and frequent changing meteorological conditions. In London, smaller temporal and spatial variation were observed but significant differences in particle sizes were also noted (Ochoa Gonzalez et al., 2016). iii) The two samples from Beijing and Xi'an in China fit well into the observed ranges from other cities. The δ 66 Zn JMC−Lyon value from Beijing of −0.36 ± 0.04‰ is lower than the δ 66 Zn JMC−Lyon of 0.02 ± 0.04‰ from Xi'an collected at the same time (summer 2011). Beijing is situated further from the dust sources than Xi'an. The isotopically lighter composition in Beijing would be in line with a stronger influence from anthropogenic activities in this city (more industry in Beijing and a mixed commercial-residential area around the sampling site in Beijing, compared to a residential area in Xi'an).
We make the following observations for Cu isotope compositions of urban aerosols (Figure 7, right panel): i) The δ 65 Cu NIST−976 values spread over a wide range from −0.01 to +0.97‰. One sample from Barcelona is even isotopically lighter with a δ 65 Cu NIST−976 value of −0.43‰. ii) Aerosol Cu isotope composition can vary considerably depending on the season. For example, isotopically heavy Cu were detected during the winter season in coarse particles (PM 2.5−80 ) in London (δ 65 Cu NIST−976 ranging between +0.63 and +0.97‰), along with higher EFs and Cu/Sb ratios (Ochoa Gonzalez et al., 2016). The authors attributed this to increased contribution from fossil fuel combustion. iii) The lighter δ 65 Cu values in urban PM, combined with lower Cu/Sb ratios, can be used for source tracing of brake particles (Ochoa Gonzalez et al., 2016;Dong et al., 2017).

SUMMARY AND CONCLUSIONS
1 Based on our synthesis of new and published data, we propose the following isotopic ranges (and means if statistically reasonable) of Zn and Cu in anthropogenic and natural reservoirs (see Table S3 and Figures 5, 6 for references): a Mineral dust derived from: 2 We present new Cu, Zn, and Pb isotope compositions for mineral dust from major deserts in Asia and Africa and from aerosols collected in Beijing and Xi'an, China. 3 We show that human activities introduce a wide range of Zn isotope compositions into the environment, highlighting that humans are changing the global geochemical cycles of Zn isotopes. 4 We do not find Zn isotope fractionation between sphalerite (natural reservoir) and ore concentrate (anthropogenic reservoir), while significant Zn isotope fractionation does occur through anthropogenic activities. We also suggest that Zn isotope fingerprints are a promising tracer to differentiate between combustion vs. non-combustion sources in aerosols and can play a major contribution to air quality assessments in urban environments. 5 Copper isotopes seem either not fractionate to the same significant extent during combustion and other industrial  Mattielli et al., 2009, 4. Shiel et al., 2010, 5. Gelly et al., 2019, 6. Thapalia et al., 2010, 7. Dong et al., 2017, 8. Fekiacova et al., 2015, 9. Cloquet et al., 2006, 10. Ochoa Gonzalez et al., 2016, 11. Gioia et al., 2008.
processes or could reflect the wide isotope range of natural reservoirs.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
NS led analysis and compilation of data and wrote the paper and made all the figures. The paper manuscript was subsequently revised by all authors, each of whom contributed to the intellectual content. SD performed most of the laboratory work, including concentration and isotope analysis of the new samples presented in this paper. YS provided all mineral dust and aerosol samples from China including the size fractionated samples. JN performed XRD analyses and led the interpretation of the mineralogical data.