Magnetic and geochemical characterization of iron pollution in subway dusts in Shanghai, China



[1] Dust samples collected from subway platforms in Shanghai, China, have been examined using magnetic measurements and geochemical analysis. Our results indicate that the Shanghai subway platform dusts have extremely strong magnetic signatures. These results, combined with X-ray diffraction analysis and scanning and transmission electron microscope examinations, indicate that the magnetic mineralogy of the dust is dominated by iron scraps due to wheel-rail mechanical abrasion and spherules rich in magnetite from fossil fuel combustion. Although the magnetic particles are primarily micrometer sized, fine submicron magnetic grains are also evident in the dust. The underground platform dusts have a much higher iron flake abundance and magnetic susceptibility than those from aboveground platforms because the latter ones are diluted by inputs of magnetically weaker ambient aerosols with a higher proportion of magnetite spherules. Geochemical analysis indicates that underground platform dusts have elevated Fe and Mn, but lower Al and Ti contents relative to aboveground subway dust. This is consistent with the closed nature of underground platforms, which therefore reduces exposure to soil-derived dust. Since the adverse environmental effects of subway particles may be linked to higher contents of iron and other metals, our results demonstrate that magnetic measurements provide a novel and effective approach for characterizing iron mineralogy and grain size in subway dusts.

1. Introduction

[2] Subway systems are key components of mass transportation networks in many large cities by the virtue of their higher passenger carrying capacity and speed over short distances. With a growing number of subway systems in operation and increasing passenger numbers, possible adverse health effects of subway systems have drawn considerable attention [Nieuwenhuijsen et al., 2007]. A range of pollutants associated with subways has been suggested [Nieuwenhuijsen et al., 2007], among which particulate pollution has been one of the major concerns because of its association with excess morbidity and mortality [Karlsson et al., 2005, 2008; Nieuwenhuijsen et al., 2007]. Elevated aerosol concentrations and significant quantities of heavy metals (e.g., Fe, Mn and Cr) compared to ambient aerosols have been reported for many subway systems, which are caused by frictional erosion of wheels, rails, brakes, etc. [Sitzmann et al., 1999; Chillrud et al., 2004; Salma et al., 2007; Kang et al., 2008]. There are claims that subway airborne dust particles are more genotoxic than street particles due to their higher Fe content [Karlsson et al., 2005, 2008]. The potential health effects of particles on transit workers and the commuting public, although not clearly determined, nevertheless calls for careful examination and management of such particles [Bachoual et al., 2007; Gustavsson et al., 2008].

[3] Previous studies indicate that magnetic measurements can be used to characterize aerosol or dust pollution in urban areas [Thompson and Oldfield, 1986; Maher and Thompson, 1999; Evans and Heller, 2003]. Relatively elevated concentrations of magnetic particles, derived from industrial emissions, vehicle exhausts and abrasion products, can lead to stronger magnetic signals in materials into which they are incorporated [Thompson and Oldfield, 1986; Maher and Thompson, 1999; Evans and Heller, 2003]. So far, the magnetic properties of urban aerosols, dusts, soils and tree leaves have been examined to monitor urban pollution worldwide [e.g., Shu et al., 2001; Muxworthy et al., 2002; Hanesch et al., 2003; Lu and Bai, 2008; Maher et al., 2008; Szönyi et al., 2008; Kim et al., 2009; Sagnotti et al., 2009]. Particles emitted from subway systems contain abundant iron metal or iron oxides [Sitzmann et al., 1999; Chillrud et al., 2004; Salma et al., 2007; Kang et al., 2008], with strong magnetic signatures [Thompson and Oldfield, 1986; Dunlop and Özdemir, 1997]; therefore, it is expected that magnetic measurements may be well suited to the study of subway dust pollution. This type of study has not been previously reported.

[4] In this study, we characterize the magnetic properties and geochemical composition of dusts collected from subway platforms in Shanghai, China, and discuss the source of the magnetic particles, their spatial variations and underlying controlling factors. Considering the rapid development of subway networks in developing countries like China, this magnetic approach can provide a novel and effective way of monitoring subway particle pollution.

2. Samples and Methods

[5] Shanghai is the largest city in China with a population more than 22 million. Subway systems began in early 1990 and have developed rapidly in recent years. There are 11 lines now in operation with a total length of 420 km and 1.9 billion passengers each year. Dust samples were collected from the platforms of subway Lines 1, 2, and 3 in Shanghai in 2008 (Figure 1). These three lines were the first lines in operation and are the busiest ones because they run through the downtown area. The platforms can be grouped into two types, namely, underground and aboveground (elevated or at ground level) in terms of their position relative to the ground surface. Underground platforms are dominant on Lines 1 and 2, while Line 3 is mainly composed of elevated aboveground stations. To ensure comparability of samples collected from different platforms, we used the organic plastic roofs of newsstands (about 2 m in height) on each platform to collect dust samples. The total number of sampling stations is 56, among which 29 are from underground platforms and 27 aboveground. At each station we obtained one sample (about 5∼10 g). For comparison purposes, ten dust samples from residential areas were also collected (Figure 1). All samples were collected using a clean toothbrush and were sealed in plastic bags.

Figure 1.

Schematic map of the study area with the sampled subway lines and stations in Shanghai. Open circles, solid circles, and stars represent underground subway dust, aboveground subway dust, and residential dust, respectively.

[6] Both low- (0.47 kHz) and high- (4.7 kHz) frequency susceptibility (χlf and χhf, respectively) were measured using a Bartington Instruments MS2B magnetic susceptibility meter. Frequency-dependent susceptibility (χfd) was calculated as χfd = (χlfχhf). The value for χfd% was obtained by expressing χfd as a percentage of χlf. An anhysteretic remanent magnetization (ARM) was acquired in a 0.04 mT direct current field, which was superimposed on a peak alternating field (AF) of 100 mT and expressed as susceptibility of ARM (χARM). An isothermal remanent magnetization (IRM) was imparted at 1 T and a backfield measurement was made using a field of −300 mT. These magnetizations are referred to as SIRM and IRM−300mT, respectively. Hard IRM (HIRM) was defined as HIRM = 0.5 × (SIRM + IRM−300 mT). S−300 was calculated as S−300 = (SIRM − IRM−300mT)/(2 × SIRM) × 100 [Bloemendal et al., 1992]. Thirty-one subway dust samples (16 and 15 from underground and aboveground, respectively) and two residential dust samples were selected for magnetic hysteresis measurements and to determine the IRM coercivity spectrum using a variable-field translational balance (VFTB). The following parameters were then derived: saturation remanent magnetization (Mrs), saturation magnetization (Ms), coercivity of remanence (Bcr) and coercivity (Bc). Sixteen samples were subjected to thermomagnetic analysis in air to determine their Curie temperatures (Tc) using the VFTB with a field of 36 mT.

[7] Magnetic extracts were obtained from 8 samples that reflect the full range of magnetic variability. A scanning electron microscope (SEM, JSM-5610) and transmission electron microscope (TEM, JEM-2100F) equipped with an energy dispersive X-ray spectroscopy (EDS) analysis system were used to examine the morphology and elemental composition of the magnetic particles. X-ray diffraction (XRD) analysis was also carried out on the extracted magnetic particles (Philips X′Pert-Pro MPD) with Cu- radiation.

[8] Particle sizes were analyzed with a laser size analyzer (Coulter LQ-100Q), after treatment with 5% H2O2 and 0.2 M HCl to dissolve organic matter and carbonate. We added 0.5 M sodium hexametaphosphate ((NaPO3)6) and used ultrasonic dispersion to ensure complete disaggregation before analysis [Ru, 2000].

[9] The samples were also subjected to geochemical characterization. Elemental concentrations (Al, Fe, Ca, Na, Ti, Mn, Cr, Cu and Ni) were determined using inductively coupled plasma atomic emission spectrometry (ICP-AES, IRIS Intrepid II XSP) after a mixed HF-HNO3-HClO4 digestion [Ru, 2000]. The China national reference material GSD9 was included for quality control and the analytical precision is better than 10% for all elements.

3. Results

3.1. Particle Size

[10] The underground subway dusts vary considerably in particle size, with median size ranging from 10.3 to 50.9 μm. The aboveground dusts have a narrow median size range from 15.5 to 25.2 μm. The dust collected from the residential area is finest (Table 1), with median size ranging from 14.9 to 20.1 μm. In general, the size distribution curves are negatively skewed, with a prominent mode at 20∼40 μm, and subordinate modes at 60–200 μm and 2∼10 μm (Figure 2). It suggests a mixture of three distinct dust components, with the finer components possibly reflecting ambient air input and the coarser component from local input.

Figure 2.

Typical particle size distribution curves for dusts collected from subway platforms and residential areas in Shanghai.

Table 1. Summary of Magnetic Properties and Median Particle Size of Dust Samples Collected From Subway Platforms and Residential Areas in Shanghai
 Underground (N = 29)Aboveground (N = 27)Residential (N = 10)
MedianMean ± SDMedianMean ± SDMedianMean ± SD
χ (10−7 m3 kg−1)1000920 ± 670100110 ± 325461 ± 36
χfd (10−7 m3 kg−1)9.716.7 ± ± ± 0.6
χfd (%)1.92.1 ± ± ± 0.7
χARM (10−7 m3 kg−1)960980 ± 680150150 ± 468593 ± 50
SIRM (10−3 A m2 kg−1)10001100 ± 770130140 ± 487384 ± 51
HIRM (10−3 A m2 kg−1)18.327.5 ± ± ± 2.1
S−300 (%)98.097.9 ± 1.497.797.6 ± 1.696.096.0 ± 0.9
SIRM/χ (kA m−1)12.413.2 ± 4.013.313.4 ± 1.813.713.8 ± 1.1
χARM/SIRM (10−5 m A−1)9.79.4 ± ± 1.711.411.7 ± 1.5
χARM/χ1.21.2 ± ± ± 0.2
Median size (μm)18.723.1 ± 11.320.220.0 ± 2.315.716.7 ± 2.2

3.2. Magnetic Properties of the Dust

[11] Concentration related parameters (χ, χfd, χARM, SIRM and HIRM) are much more variable than mineralogy and grain size indicators (S−300, χfd%, χARM/SIRM and χARM/χ) (Table 1). The χ of subway dust ranges from 55 × 10−7 m3 kg−1 to 2800 × 10−7 m3 kg−1, with a mean value of 530 × 10−7 m3 kg−1. For the two types of subway station, the mean value of χ for underground dust is about 8 times that of the aboveground stations (Table 1). Compared to the dust collected from residential areas, the aboveground subway dust is clearly magnetically enhanced, although the difference is not large (Table 1).

[12] SIRM also generally reflects the concentration of magnetic minerals. Compared to χ, SIRM generally reflects the concentration of magnetic remanence-carrying minerals [Thompson and Oldfield, 1986]. On the whole, SIRM has a significant positive relationship with χ (r = 0.98, p < 0.01; Figure 3a), which implies that χ is mainly dominated by ferromagnetic and ferrimagnetic particles (e.g., iron and magnetite) [Thompson and Oldfield, 1986]. Underground and aboveground dusts can be clearly distinguished, with the former having higher SIRM and χ values, but slightly lower SIRM/χ ratios (Figure 3a and Table 1). The aboveground subway dust is similar to dust collected from residential areas (Figure 3a).

Figure 3.

Relationships between (a) χ versus SIRM, (b) χ versus χARM, (c) SIRM versus Fe, and (d) Al versus Ti. The significant positive relationship between χ and SIRM in Figure 3a suggests that χ is dominated by ferro(i)magnetic particles. Undergound subway dusts have higher SIRM to Fe ratios (Figure 3c), which suggests a higher proportion of more strongly magnetic (ferromagnetic) particles. All of the dusts have similar Al to Ti ratios (Figure 3d), which suggest common natural sources. Open circles, solid circles, and stars represent underground subway dust, aboveground subway dust, and residential dust, respectively.

[13] S−300 serves as a measure of the relative importance of higher coercivity minerals (e.g., hematite and goethite) with respect to lower coercivity components (e.g., iron and magnetite) in the total assemblage [Bloemendal and Liu, 2005]. The subway dust samples have a mean S−300 value of about 97.7%, which suggests the dominance of ferromagnetic/ferrimagnetic particles as magnetic carriers. Thermomagnetic analysis (Figure 4) indicates that all samples are characterized by two or three distinct Curie temperatures at around 550°C, 730°C and 760°C, respectively. The first Curie temperature is close to that of magnetite (580°C), and the latter two are slightly lower than or close to that of iron (765°C) [Dunlop and Özdemir, 1997]. Minor substitution of other elements can decrease the Curie temperature of magnetite [Dunlop and Özdemir, 1997]. Likewise, the slightly lower iron Curie temperatures can be characteristic of iron alloys [e.g., Li et al., 2002]. In general, the heating curves of aboveground station samples (Figures 4a and 4b) are dominated by the Tc of 550°C, while the underground stations also have a higher Tc > 700°C component (Figures 4c and 4d). The cooling curve in Figure 4a is above the heating curve with a Tc of 580°C, indicating magnetite generation during thermal treatment. In contrast, the cooling curve in Figure 4c is below the heating curve, suggesting the oxidation of iron alloy into weaker magnetic minerals during the heating process. The heating and cooling curves in Figure 4b are more reversible.

Figure 4.

Thermomagnetic curves for selected samples, with the bold and thin lines representing heating and cooling curves, respectively. The arrows indicate the Curie temperatures (Tc) at around 550, 730 and 760°C. Note the difference between (a and b) aboveground and (c) underground stations. (d) The heating curve above 700°C is enlarged to more clearly illustrate the two Tc > 700°C components. The samples were measured in air using a field of 36 mT.

[14] χfd reflects the presence of fine viscous superparamagnetic (SP) grains close to the SP/single domain (SD) boundary (e.g., ∼0.02 μm for Fe3O4), while χARM is sensitive to SD (e.g., ∼0.02–0.06 μm for Fe3O4) grains [Maher, 1988]. The χfd does not correlation with χ, while χARM has higher correlation coefficients (r = 0.95, p < 0.01) (Figure 3b). The χfd% ranges from 0.0% to 6.2%, with the mean of 2.0%, indicating a minor viscous SP contribution to the total magnetic susceptibility [Thompson and Oldfield, 1986]. It should be noted that such a frequency-dependent susceptibility method only detect larger SP grains, and other methods such as low-temperature magnetic measurement should be used to detect smaller SP grains [Sagnotti et al., 2009]. Nevertheless, the mass-specific χfd values (Table 1) are significantly higher than those for typical soils (e.g., mean ± standard deviation of χfd for 1176 UK soil samples is 4.5 ± 13.5 × 10−8 m3 kg−1 [Dearing et al., 1996]). Submicrometer particles around the SP/SD threshold size have higher surface reactivity and they possibly pose a greater danger to human health [Nieuwenhuijsen et al., 2007], such a magnetic method is valuable in dust pollution studies. Extremely fine magnetic grains are normally present in the finest particle size fraction [cf. Oldfield et al., 2009], in which case the magnetic results here imply that the potential health effects associated with fine subway dust particles cannot be ignored.

[15] HIRM provides an estimate of higher coercivity mineral concentrations if HIRM is independent of the L ratio [Liu et al., 2007]. Its correlation with χ (r = 0.79, p < 0.01) and SIRM (r = 0.81, p < 0.01) suggests that higher coercivity minerals are closely associated with ferromagnetic/ferrimagnetic particles in the subway dust.

[16] The χARM/SIRM is commonly used as a grain size indicator of magnetic particles, peaking in the SD range and decreasing with increasing grain size [Maher, 1988]. The χARM/χ is also a grain size indicator, but the presence of SP grains can reduce its value [Maher, 1988]. Mean χARM/SIRM and χARM/χ values of 10.2 × 10−5 m A−1 and 1.6, respectively, confirm that the magnetic grains are predominantly pseudosingle domain (PSD) and multidomain (MD) in size [Oldfield, 1994]. The rather small variability of these parameters (Table 1 and Figure 3b) suggests that the grain size distribution of magnetic minerals is rather similar in the dust samples. On a plot of Mrs/Ms versus Bcr/Bc (Figure 5), most samples fall in the PSD region, lying between SD-MD and SD-SP theoretical mixing trends and closer to the MD region [Day et al., 1977; Dunlop, 2002], which is consistent with the inferred coarse nature of the magnetic grains with a contribution from SP grains. On the whole, samples from the aboveground stations are generally magnetically finer than those from the underground stations (Figure 5).

Figure 5.

Plot of Mrs/Ms versus Bcr/Bc [Day et al., 1977; Dunlop, 2002] for selected samples, which fall in the PSD region and lie between the theoretical mixing curves for SD-MD and SD-SP grains [Dunlop, 2002]. Samples from aboveground subway platforms generally have higher Mrs/Ms but lower Bcr/Bc values, which suggests a finer average magnetic grain size. Open circles, solid circles, and stars represent underground subway dust, aboveground subway dust, and residential dust, respectively.

3.3. XRD and SEM/TEM Analysis

[17] XRD analysis clearly confirms the presence of iron and magnetite in the magnetic extracts, with iron more abundant in underground samples (Figure 6). SEM analysis reveals two types of magnetic particles. Angular sheet-like particulates dominate the samples, with sizes generally <100 μm in length, <50 μm in width and about 1–2 μm in thickness. In addition, spherical particles are also present and are more abundant from ground level or elevated platforms (Figures 7a7d). The diameters of such particles are generally less than 50 μm. Elemental composition analysis indicates that the sheet-like particles are primarily composed of Fe with minor Cr, and the spherical particles are composed of Fe and O (Figures 7e and 7f). TEM analysis confirms the presence of clumped submicrometer-sized iron-rich particles (Figures 7g7k), which is consistent with the presence of SP and SD grains detected by magnetic methods.

Figure 6.

XRD spectra for selected samples from aboveground station Tonghexincun on subway Line 1 and underground station Weining Road on subway Line 2. M, magnetite; Fe, iron; H, hematite; and Q, quartz.

Figure 7.

SEM and energy-dispersive X-ray spectra (EDS) of extracted magnetic particles from (a) aboveground station Tonghexincun on subway Line 1 and (b) underground station Loushanguan Road on subway Line 2. (c) Iron scrap and (d) iron oxide spherules with (e and f) their representative EDS spectra. (g and h) TEM images of submicrometer-sized particles with (i–k) iron-rich features. The C and Cu peaks originate from the C-coated Cu TEM grids.

3.4. Geochemical Characterization of the Dust

[18] On average, the Fe and Mn contents follow the sequence: underground subway dust > aboveground subway dust > residential dust (Figures 8a and 8b). In contrast, Al, Ti and Ca concentrations follow the sequence: underground subway dust < aboveground subway dust < residential dust (Figures 8c8e). Na content is on average lower in underground subway dust (Figure 8f). The residential dust has major elemental compositions similar to previously reported values for street dust in Shanghai (Table 2) [Tanner et al., 2008]. On average, Cr contents are higher in the aboveground stations and in the residential areas, while Cu and Ni contents are higher in the underground stations (Table 2).

Figure 8.

Box-whisker plots with comparison of major elements (a) Fe, (b) Mn, (c) Al, (d) Ti, (e) Ca, and (f) Na in the dusts. Labels in Figure 8a apply to all the other panels. A, B, and C represent underground subway dust, aboveground subway dust, and residential dust, respectively.

Table 2. Summary of Geochemical Compositions of Dust Samples Collected From Subway Platforms and Residential Areas in Shanghai
 Underground (N = 29)Aboveground (N = 27)Residential (N = 10)Shanghai Soil Backgrounda MeanShanghai Street Dustb Mean ± SD
MedianMean ± SDMedianMean ± SDMedianMean ± SD
Fe (%)9.2812.68 ± 10.344.464.98 ± 1.853.834.21 ± 1.773.184.65 ± 1.05
Al (%)1.681.99 ± 0.993.723.54 ± 0.854.444.46 ± 0.726.994.52 ± 1.019
Ca (%)3.284.42 ± 2.167.567.31 ± 1.959.248.62 ± 2.940.92 
Na (%)0.410.67 ± 0.340.930.82 ± 0.320.820.87 ± ± 0.63
Mn (mg kg−1)10001400 ± 9509001000 ± 460820900 ± 370506.0904 ± 218
Ti (mg kg−1)13001500 ± 64028002600 ± 57035003400 ± 59044003200 ± 500
Cr (mg kg−1)163220 ± 140370800 ± 1300220660 ± 110073.8242 ± 121
Cu (mg kg−1)680700 ± 200330350 ± 170190440 ± 80027.7141 ± 56
Ni (mg kg−1)6195 ± 618085 ± 526880 ± 4029.8 

[19] Enrichment factors (EF), which are commonly defined as the observed element to Al ratio in the sample of interest, divided by the element/Al ratio in the natural background material, can be used to assess the degree of anthropogenic influence [e.g., Zhang and Liu, 2002]. The use of Al for normalization reflects the fact that Al is one of the most abundant elements on earth and that it is rarely a component of anthropogenic contamination. We take the Shanghai soil background values for reference (Table 2) [China National Environmental Monitoring Centre (CNEMC), 1990]. An EF > 1.5 suggests that a significant portion of the element is delivered from noncrustal materials [Zhang and Liu, 2002]. Overall, Ti has EF values < 1.5, which suggests that it is primarily natural in origin (Table 3). Na has EF values < 1.5 in the aboveground and residential dusts, and ∼2 for underground dusts. The remaining elements have EF values > 2, which indicates that anthropogenic input is present in the dusts. For the underground subway dusts, median EF values for elements are arranged in the following order: Cu > Ca > Fe > Cr > Ni > Mn > Na. For the aboveground subway dusts and residential dusts, the order is as follows: Cu > Ca > Cr > Ni > Mn > Fe > Na (Table 3).

Table 3. Summary of Enrichment Factors for Elements in Dust Samples Collected From Subway Platforms and Residential Areas in Shanghai
 Underground (N = 29)Aboveground (N = 27)Residential (N = 10)
MedianMean ± SDMedianMean ± SDMedianMean ± SD
Fe13.714.1 ± ± ± 1.0
Mn9.69.6 ± ± ± 1.3
Ti1.31.3 ± ± ± 0.1
Na2.12.2 ± ± ± 0.2
Ca16.717.0 ± 2.115.415.8 ± 2.915.715.6 ± 6.7
Cu101.1114.6 ± ± 16.611.126.1 ± 47.3
Cr11.111.1 ± 4.98.323.2 ± 41.14.615.2 ± 25.0
Ni10.410.4 ± ± ± 2.5

[20] Previous studies have revealed that subway dusts have elevated Fe and Mn contents, and elevated trace metals like Cr and Ni [Sitzmann et al., 1999; Chillrud et al., 2004; Salma et al., 2007; Kang et al., 2008]. The significant positive relationships between Fe and Mn (r = 0.98, p < 0.01), Ni (r = 0.95, p < 0.01) and Cr (r = 0.83, p < 0.01), point to steel containing Mn, Ni and Cr as a major source. However, the correlations between Fe, and Cr, Ni and Mn are much poorer for aboveground subway dusts, which implies a mixture of railway friction materials with dusts from other sources of these elements. The significant Ca pollution in each type of sampled dust can be an indication of the widespread use of cement as a construction material in urban environments.

4. Discussion

[21] Urban dusts include material from natural sources, such as soil, as well as anthropogenic inputs, such as industrial emissions. Dusts from subway platforms have extremely high magnetic susceptibility and SIRM, but have low χfd%, χARM/SIRM and χARM/χ values, which suggest a dominant contribution from coarse PSD/MD grains. SEM analysis reveals that iron sheets occur widely in our samples, which is supported by XRD analysis and strong enrichment of elemental Fe in the subway dusts, especially in underground dusts (Table 2). We regard the iron sheets as the product of rail abrasion. There are claims that iron in subway particles mainly occur as magnetite (Fe3O4) [Karlsson et al., 2005]. Our study, however, reveals that metallic iron is also an important component in the dust, together with iron oxides such as hematite.

[22] The observed spherical magnetic particulates (e.g., Figure 7d) are similar to the magnetic particles in fly ash that are derived from high-temperature fossil fuel combustion [Thompson and Oldfield, 1986; Maher and Thompson, 1999; Lu and Bai, 2008]. They are more abundant in samples from aboveground stations (Figures 4, 6, and 7), especially those from the northern portion of Line 3 in the Baoshan District in northern Shanghai, which indicates that they reflect ambient air input, which is influenced by heavy industries such as steel works and power stations within Shanghai [Shu et al., 2001]. The total suspended particles collected from the heavy industrial area of Shanghai has χ values of 23–245 × 10−7 m3 kg−1 [Shu et al., 2001], and urban topsoils in this area have χ values of 13–196 × 10−7 m3 kg−1 [Hu et al., 2007]. These values are comparable to those for the dust collected from the residential areas, which has χ values of 9–146 × 10−7 m3 kg−1 (Figure 3a).

[23] The studied dust samples contain significant soil-derived components, therefore signs of a natural ferrimagnetic mineral should be present. However, this natural ferrimagnetic signature is masked by the anthropogenic magnetic particles. For example, the χ value of background soil in the Baoshan District of Shanghai has been found to be 3 × 10−7 m3 kg−1 [Hu et al., 2007].

[24] Magnetically, the subway dust mainly contains particulates from the two dominant anthropogenic sources described above. Dusts from the underground stations have much higher χ and SIRM values than those aboveground (Table 1). Bulk dust particle size compositions have no significant relationship with magnetic parameters, implying the variations of magnetic properties cannot be explained by differences in particle size. Due to the relatively confined nature of underground stations, magnetic particles in the underground platforms mostly come mostly from wheel-rail friction. Consequently, ferromagnetic minerals (i.e., iron) with stronger magnetic susceptibility are more abundant. By contrast, aboveground stations receive more dust from the ambient environment, which is magnetically weaker and lowers the magnetic susceptibility of aboveground subway dusts. In addition, iron fragments generated aboveground are likely to be less closely confined and will be more widely dispersed than those generated underground. The alteration of iron fragments due to contact with moisture in the air is also probably much greater for aboveground dusts.

[25] Geochemical analysis supports such an inference. Ti is a conservative element and is unlikely to reflect contamination. Despite the difference in geochemical composition between the dusts from different sites, the ratio of Al to Ti is broadly similar, with a mean value of 13.1 (Figure 3d), which suggests that the subway dust and residential dust have a common natural component. The dusts collected from residential areas are far away from the subway stations and are more likely to be representative of ambient air pollution, which is characterized by the highest Al, Ti, Ca and Na values, but the lowest Fe and Mn concentrations and χ values (Tables 1 and 2). The underground subway dusts have lower Al, Ti, Ca and Na but higher Fe and Mn concentrations relative to dusts collected aboveground, which indicates that they have the lowest proportion of natural crustal material due to limited exchange with ambient aerosols (Table 2). By contrast, aboveground stations receive dust from open air sources which dilutes the effect of any particles derived from rail abrasion and result in lower χ values and intermediate Al, Ti, Ca, Na, Fe and Mn contents (Tables 1 and 2). Furthermore, underground dust has higher SIRM to Fe ratios compared to aboveground and residential dust (Figure 3c), which confirms the stronger ferromagnetic contributions in the former.

[26] There is also large variability in χ and SIRM values within each type of subway platform dust (Figure 3a), which reflects station-specific conditions for dust transport, mixing and deposition. For example, the underground stations on Line 1 are installed with screen doors and they have lower mean χ values (730 × 10−7 m3 kg−1) compared to their counterparts on Line 2 without screen doors (1190 × 10−7 m3 kg−1). This suggests that installation of screen doors reduces the transport of iron fragments onto the platforms, and is therefore environmentally beneficial.

5. Conclusions

[27] Combined magnetic measurements, XRD analysis, SEM/TEM examination and geochemical analysis were used to characterize the magnetic mineralogy and elemental composition of subway dusts in Shanghai. The results indicate that iron scrap and spherical magnetic particulates are the dominant magnetic minerals in subway dusts. We suggest that the iron scrap is the product of wheel-rail mechanical abrasion, while the spherules are derived from fossil fuel combustion products emitted into the ambient atmosphere. Relative to aboveground subway dust, underground subway dust has stronger χ and SIRM values, higher proportion of ferromagnetic iron scrap and higher concentrations of Fe and Mn, but lower Al and Ti concentrations. The relatively confined space of underground platforms limits exchange with ambient airborne particles. By contrast, aboveground subway dust includes ferrimagnetic particles from the ambient air in addition to rail-derived particles, and is therefore magnetically weaker with higher concentrations of Al and Ti from natural sources. Subway dusts have elevated Cr, Cu and Ni contents derived both from rail friction and from other industrial sources. Our results demonstrate that magnetic measurements are well suited to subway dust pollution monitoring studies.


[28] This work was financially supported by the Science and Technology Commission of Shanghai (036505004) and Program for New Century Excellent Talents in University (NCET-08-0195). We thank Frank Oldfield, Ramon Egli, Leonardo Sagnotti, an anonymous reviewer, and Andrew Roberts (Associate Editor) for their constructive reviews and comments.