Measurement of number and mass size distributions of light-absorbing iron oxide aerosols in liquid water with a modified single-particle soot photometer

Abstract

Here, we used a modified single-particle soot photometer (SP2) coupled with a concentric pneumatic nebulizer to measure the size-resolved number and mass concentrations of light-absorbing iron oxide aerosols (FeO_x) in liquid water (C_NFeOx and C_MFeOx, respectively). The SP2 could selectively detect individual FeO_x particles in mixed wüstite–fullerene soot laboratory samples and melted Arctic snow samples. The nebulizer efficiency for FeO_x particles was about 50% within the 70–650 nm diameter range, as derived from the ratio of the volume of ammonium sulfate before and after extraction by the nebulizer and the size-resolved transmission efficiency in the nebulizer–SP2 sampling line. Uncertainty from the boundary lines empirically drawn to discriminate the scatterplots of FeO_x and black carbon in the mixed wüstite–fullerene soot suspensions and snow samples was approximately 3.0% and 10%, respectively. Overall uncertainty in total C_NFeOx and C_MFeOx (220–1400 nm) was approximately 19% and 18%, respectively. After storage at 4 °C for 16 months, the FeO_x particle size distributions in melted Arctic snow had remained stable, and C_NFeOx and C_MFeOx had changed by less than 19% and 1.0%, on average, respectively. Most of the FeO_x on dust particles measured by this system was estimated to be in the diameter range smaller than 1000 nm, considering the nebulizer efficiency for dust particles. The high accuracy of the C_NFeOx and C_MFeOx measurements will help to improve our quantitative understanding of the wet deposition of FeO_x and provide more accurate estimates of the effects of FeO_x on snow surface albedo.

EDITOR:

• Hans Moosmüller

1. Introduction

Iron oxide aerosols (FeO_x) absorb solar radiation and heat the atmosphere. Absorption of solar radiation by FeO_x particles deposited onto high- and mid-latitude snow can decrease the albedo of snow surfaces and accelerate snowmelt (Ito, Lin, and Penner 2018; Matsui et al. 2018; Moteki et al. 2017), which may contribute to the acceleration of global warming. However, deposition of FeO_x-containing particles to the ocean where iron is limited can also influence ocean primary productivity and marine ecosystems, increasing carbon dioxide uptake into the ocean (Jickells et al. 2005; Matsui et al. 2018).

FeO_x particles are emitted both from anthropogenic and natural sources. Anthropogenic sources are mainly blast furnaces of iron manufacturing facilities, coal and heavy oil combustion, and the brakes of motor vehicles (Fu et al. 2012; Machemer 2004; Sanderson et al. 2016). Natural sources include emissions of mineral dust particles from desert soils (Nowak et al. 2018). Emitted FeO_x particles would become hydrophilic while aging during transport; therefore, it is likely that FeO_x-containing particles with diameters of less than a few micrometers are removed from the atmosphere mainly by precipitation. To improve our understanding of the wet deposition of FeO_x, accurate measurements of the number and mass size distributions of FeO_x in hydrometeor samples are needed.

Generally, Fe concentrations are measured in water samples by means of bulk chemical analyses using inductively coupled plasma mass spectrometry or energy dispersive X-ray fluorescence (Bhatia et al. 2013; Iwamoto et al. 2009); however, these approaches provide no data on the size distributions of Fe-containing particles in the samples. To address this issue, Yoshida et al. (2016) used a modified single-particle soot photometer, based on a laser-induced incandescence technique (Moteki and Kondo 2010), to selectively detect individual FeO_x and black carbon (BC) particles and measure their size distributions in the atmosphere. This technique has also been used by others to measure the number and mass size distributions of FeO_x particles in the atmosphere (Moteki et al. 2017; Ohata et al. 2018; Yoshida et al. 2018; Yoshida et al. 2020). Lamb et al. (2021) have also measured the size distributions of FeO_x in the atmosphere using their own modified SP2 (Lamb 2019; Schwarz et al. 2010).

Mori et al. (2016) used a differently modified SP2 with a concentric pneumatic nebulizer (Marin-5; CETAC Technologies, Inc., USA) to measure the size-resolved number and mass concentrations of BC in liquid water with an accuracy of approximately 16%. In the present study, for the first time, we used the same nebulizer–SP2 system to measure the size-resolved number and mass concentrations of FeO_x in liquid water (C_NFeOx and C_MFeOx, respectively). Here, we describe our methodologies for measuring the size distributions and associated uncertainties of FeO_x in water, and then apply our approach to examining the stability of FeO_x in water during long-term storage.

2. Methods

To determine the masses of the BC and FeO_x particles in the study samples, we used a nebulizer–SP2 system, as reported previously (Katich, Perring, and Schwarz 2017; Mori et al. 2016; Schnaiter et al. 2019). In brief, water samples were agitated by an ultrasonic bath for 15 min to remove any particles from the walls of the glass container (Mori et al. 2014) and then injected into the nebulizer at a constant flow rate (3.0 or 6.0 μL s⁻¹) by a peristaltic pump (Reglo Digital, Ismatec, Switzerland). Nebulization of the samples generated droplets with a majority size of 2–3 µm, which were then evaporated at 140 °C in the nebulizer’s heated spray chamber. A portion of the BC and/or FeO_x particles was extracted by the nebulizer at a constant flow rate (16 cm³ s⁻¹ at standard temperature and pressure) and introduced into the SP2 for measurement of the masses of the individual particles in the carrier gas. To quantify the number and mass size distributions of FeO_x in the water samples, we determined the size-resolved number and mass concentrations of FeO_x in the carrier gas (Sections 2.1 and 2.2) and the nebulizer extraction efficiency for FeO_x particles (Section 2.3).

2.1. Using the SP2 for the detection of airborne FeO_x

SP2 instruments have been used previously to measure the masses of individual airborne FeO_x (e.g., magnetite) and BC particles (Moteki et al. 2017; Yoshida et al. 2016). Inside an SP2, the airborne sample is passed through a Nd:YAG laser beam (λ = 1064 nm) and particles contained in the sample absorb energy from the laser, heat up to their evaporation temperature (e.g., of 3300 K for FeO_x and 4100 K for BC), and incandesce (Yoshida et al. 2016). In the modified SP2 of Yoshida et al. (2016), the incandescence signal (S_i) is detected using two photomultiplier tubes, one sensitive to the blue band range (λ = 300–550 nm) and one sensitive to the red band range (λ = 580–710 nm), and the ratio of the peak S_i of the blue band to that of the red band, which is hereafter referred to as the color ratio, is then determined (Moteki and Kondo 2010). This color ratio is a key parameter to discriminate between FeO_x and BC particles because its value strongly depends on the blackbody temperature of individual light-absorbing aerosols. After discriminating FeO_x and BC particles by color ratio, the masses of individual FeO_x particles are derived from the experimentally determined relationship between peak S_i of the blue band and the masses of laboratory samples of FeO_x (e.g., magnetite and wüstite) (Yoshida et al. 2016). Note that although the SP2 can detect FeO_x particles such as magnetite and hematite, the detection of other FeO_x particles, such as ferrihydrite, is difficult due to their weaker light-absorption efficiency (Yoshida et al. 2016).

2.2. Discrimination of FeO_x from BC

Commercial SP2 instruments are generally able to measure the individual masses of BC particles within a mass equivalent diameter (D_m) range of 70–600 nm. However, our modified SP2 can measure the individual masses of BC particles within a D_m range of 70–4170 nm, assuming a void-free density of 1.80 g cm⁻³ (Mori et al. 2016; Moteki and Kondo 2010). The extended D_m range of our SP2 is accomplished by replacing a position-sensitive Si-avalanche photodiode with a H6779 photomultiplier tube and by lowering the sensitivity of the photomultiplier tube. Hereafter, our wide-range SP2 is referred to as wSP2.

To confirm that FeO_x particles could be discriminated from BC particles by the wSP2, we measured the relationship between peak S_i of the blue band and color ratio by using laboratory samples of a finely powdered wüstite (Kojundo Chemical Laboratory Co., Ltd., Tokyo, Japan) and fullerene soot (Alfa Aesar, Inc., Ward Hill, MA, USA) (Section 2.2.1), which were used as standard materials for FeO_x and BC, respectively (Yoshida et al. 2016), and by using environmental samples (Section 2.2.2). Yoshida et al. (2016) have previously demonstrated that the color ratios for finely powdered magnetite and wüstite samples are comparable.

2.2.1. Laboratory samples

The wüstite and fullerene soot laboratory samples were dispersed separately into ultrapure water and then aerosolized using an aerosol generator (AG-100, Droplet Measurement Technologies, Longmont, CO, USA). After passing through a diffusion dryer (Model 3062, TSI Inc., Minnesota, USA) and a neutralizer (Model 1090, TSI Inc., Minnesota, USA), the aerosolized particles were mass sorted in an aerosol particle mass analyzer (APM model 3601, Kanomax, Japan) and introduced into the wSP2. The shapes of the scatterplots of the peak S_i of the blue band and color ratio were very different and the color ratios were mainly in the ranges of 0.7–0.9 for wüstite and 1.2–1.8 for fullerene soot (Figures 1a and b). The results indicate that the two laboratory samples could be separated by a boundary line (i.e., solid lines in Figures 1a and b).

Figure 1. Relationship between peak incandescence (Incd.) signals (S_i) and color ratios for laboratory samples of (a) wüstite, (b) fullerene soot, and (c) mixed wüstite–fullerene soot; snowpack, surface snow, and subsurface snow samples from (d) Alaska, (e) Finland, (f) North Siberia, (g) Greenland, and (h) Ny-Ålesund; and (i) precipitation samples from Ny-Ålesund during 2016–2021. Empirically determined boundary lines discriminating FeO_x and BC scatterplots are shown as black lines in each panel. The color code indicates particle counts on a logarithmic scale. The right axis in (a) shows the approximate mass equivalent diameter (D_m) of magnetite particles.

A similar scatterplot was also obtained for mixed wüstite–fullerene soot suspensions (Figure 1c; sample preparation is described in Appendix A). The scatterplots for wüstite and fullerene soot particles could be clearly discriminated; however, there was an area where the scatterplots overlapped (log₁₀(Incd. peak S_i) value range, 3.5–3.9 [i.e., 160 nm < D_m < 220 nm for the FeO_x particles]; color ratio range, 1.0–1.2; Figure B1). Increasing the lower threshold used for the log₁₀(Incd. peak S_i) value from 3.9 (D_m = 220 nm) to 4.25 (D_m = 320 nm) reduced the overlap, decreasing the uncertainties in the C_NFeOx and C_MFeOx values. Although the C_MFeOx (D_m > 320 nm)/C_MFeOx (D_m > 220 nm) ratio was close to unity, the average C_NFeOx (D_m > 320 nm)/C_NFeOx (D_m > 220 nm) ratio was 0.60 ± 0.08 in our precipitation and snowpack samples. Therefore, to obtain statistically reliable dataset and minimize the overall measurement uncertainty, we used the FeO_x data with D_m > 220 nm.

A calibration curve for the relationship between peak S_i and the mass of individual wüstite and magnetite particles was conducted and used to derive FeO_x mass concentrations (Figure 2). The calibration curve for the wüstite was very similar to that for the magnetite, which is consistent with our previous findings (Yoshida et al. 2016). Size-resolved mass concentrations of FeO_x in air were derived from the calibration curve by assuming a magnetite density of 5.17 g cm⁻³.

Figure 2. Regressions between peak incandescence signal (S_i) and the masses of individual particles of magnetite and wüstite. Calibration curves were fitted with the least squares method.

2.2.2. Melted snow and precipitation samples

We also investigated the peak S_i–color ratio relationships of samples of melted snow (surface snow, subsurface snow, or snowpack) collected from Alaska, Finland, North Siberia, Greenland, and Ny-Ålesund, and samples of precipitation collected from Ny-Ålesund (Figures 1d–i). The snow samples were melted at room temperature before storage, whereas the precipitation samples received no additional treatment; all samples were stored in a refrigerator at 4 °C until analysis. Details of the snow and precipitation sampling methods are described in Mori et al. (2019) and Mori et al. (2021), respectively. Data related to the samplings and sample storage periods are summarized in Table 1. For the Arctic snowpack and precipitation samples, the scatterplots for the FeO_x and BC particles could be discriminated sufficiently. However, the absolute values of the color ratio depended to some extent on the average YAG laser voltage during the analysis (Table C1). Therefore, we empirically determined the boundary lines discriminating the scatterplots for the FeO_x and BC particles for each set of samples (Yoshida et al. 2016). The change in the color ratio depending on the YAG laser voltage and the uncertainty associated with empirically determining the boundary lines are discussed in Section 3.

Table 1. Details of the snow samples used in the present study.

Region	Sampling area (latitude, longitude)	Type of snow sample	Sampling year	Number of sites	Number of samples	Sample storage period, months	References
Alaska	62.8–70.2°N, 144–151°W	Snowpack	2015	11	32	1	Mori et al. 2019
Finland	68.0–69.9°N, 27.0–27.5°E	Snowpack	2013	6	6	15	Mori et al. 2019
North Siberia	70.3–70.8°N, 147–149°E	Surface snow	2015	5	5	1	Mori et al. 2019
Greenland	63.1–77.1°N, 30.0–49.7°W	Surface and subsurface snow	2015	7	21	9	Mori et al. 2019
Ny-Ålesund	78.9°N, 11.9°E	Snowpack	2013	1	1	13	Mori et al. 2019
Ny-Ålesund	78.9°N, 11.9°E	Precipitation	2016–2021	1	275	7	Mori et al. 2021

2.3. Derivation of C_NFeOx and C_MFeOx

Some of the aerosol particles generated by the nebulizer are lost inside the chamber or during transmission from the nebulizer to the wSP2 mainly due to gravitational deposition, Brownian diffusion, and inertial deposition caused by flow constriction. Therefore, before deriving the C_NFeOx and C_MFeOx, we first need to estimate the size-resolved transmission efficiency (Tr) for transmission of the sample from the nebulizer to the laser beam in the wSP2, and nebulizer extraction efficiency (ε) for FeO_x particles.

2.3.1. Particle transmission efficiency (Tr)

We calculated the size-resolved Tr for the polystyrene latex (PSL) spheres and magnetite particles in the carrier gas using on the method described by Kulkarni, Baron, and Willeke (2011), assuming a density and a shape factor of the aerosol particles. The shape factor is generally derived from the effective and true density values of the aerosol particle (Moteki and Kondo 2010). We assumed that the shape factor of magnetite is the same as that of BC (Yoshida et al. 2018; Yoshida et al. 2020).

The Tr values were largely size-independent within the D_m range of 50–2000 nm for PSL spheres and the range of 50–650 nm for magnetite, whereas the values decreased at D_m values larger than these ranges mainly due to gravitational deposition (Figure 3a). The Tr values for hematite (density = 5.25 g cm⁻³) were very similar to those for magnetite (data not shown). Our calculation also shows that the larger FeO_x particles were lost mainly during transmission of the carrier gas from the outflow of the nebulizer to the inlet of the wSP2. The uncertainties in the calculation of the Tr values, arising from uncertainties in the conductive tube length and the flow rate of the carrier gas, were within 10% for the D_m range of 50–1000 nm, on average. At larger D_m values, Tr was more sensitive to changes in the density and shape factor of the aerosols. However, it should also be noted that FeO_x-containing dust particles will be lost more efficiently than internally unmixed FeO_x particles (hereafter, “pure FeO_x particles”) for the same FeO_x mass due to larger total masses.

Figure 3. (a) Theoretically calculated transmission efficiency for polystyrene latex (PSL) spheres and magnetite particles, assuming densities (ρ) of 1.05 and 5.17 g cm⁻³ and shape factors of 1.0 and 1.52, respectively. (b) Extraction efficiency (ε) of the nebulizer, as determined experimentally for water suspensions of PSL particles with known number concentration and those of Aqua-Black 162 particles with known mass concentration (data modified from Mori et al. 2016) or calculated for PSL and FeO_x particles. Bars indicate the systematic errors of ε, which are derived from the typical errors for the number concentrations in a carrier gas and liquid water, peristaltic pump flow rate, and nebulizer flow rate.

2.3.2. Nebulizer extraction efficiency (ε) for FeO_x particles

Mori et al. (2016) used ammonium sulfate solutions of known mass concentrations to measure the volume concentration of dried ammonium sulfate in a carrier gas (15–900 nm) using a system that comprised a scanning mobility particle sizer spectrometer coupled with a nebulizer. They showed that approximately half of the total volume concentration of ammonium sulfate particles within a D_m range of 15–900 nm was lost in the nebulizer. Theoretical nebulizer extraction efficiencies (ε) can be calculated by using size-resolved Tr and volume loss rate (VLR) values as follows: (1) $$\mathrm{Calculated}\mathrm{ }\epsilon =Tr\times (1-\mathrm{VLR})\mathrm{ }\left(\mathrm{\%}\right).$$(1)

The VLR depends on the peristaltic pump flow rate. In the present setup, VLR value was 0.5 at a peristaltic pump flow rate of 3.0 μL s⁻¹ and 0.65 at a flow rate of 6.0 μL s⁻¹. Within the 70–2000 nm D_m range, the measured ε values for PSL and Aqua-Black 162 (Tokai Carbon Co., Ltd., Tokyo, Japan) at the peristaltic pump flow rate of 3.0 μL s⁻¹ were both around 50%; at larger D_m (2000–5000 nm), the ε values gradually decreased (Figure 3b). Our calculated ε values for PSL were consistent with the experimentally determined values. Katich, Perring, and Schwarz (2017) and Schnaiter et al. (2019) derived ε values of about 35%–50%, using water suspensions of PSL and fullerene soot particles with known number and mass concentrations.

We used Equation (1)(1) $$\mathrm{Calculated}\mathrm{ }\epsilon =Tr\times (1-\mathrm{VLR})\mathrm{ }\left(\mathrm{\%}\right).$$(1) to estimate the size-resolved ε values for FeO_x (ε_FeOx). Within the 50–650 nm D_m range, ε_FeOx values were stable at about 50%, on average. For larger particles, the ε_FeOx value gradually decreased with increasing diameter (Figure 3b).

2.3.3. Size distribution of FeO_x in water

The mass size distribution of FeO_x in water (dC_MFeOx/dlogD_m) was derived from the peristaltic pump flow rate (V_pump), nebulizer flow rate (F_neb), size-resolved ε_FeOx(D_m), and mass size distribution of FeO_x in the carrier gas (dM_FeOx/dlogD_m), as follows (2) $$\frac{\mathrm{d}{C}_{\mathrm{MFeOx}}}{\mathrm{d}{\log D}_{\mathrm{m}}}=\frac{\mathrm{d}{M}_{\mathrm{FeOx}}}{\mathrm{d}{\log D}_{\mathrm{m}}}\times \frac{F_{\mathrm{neb}}}{V_{\mathrm{pump}}{\epsilon }_{\mathrm{FeOx}}\left(D_{\mathrm{m}}\right)}.$$(2)

Using the number size distribution of FeO_x in the carrier gas (dN_FeOx/dlogD_m), the number size distribution of FeO_x in water (dC_NFeOx/dlogD_m) was similarly derived. The C_NFeOx and C_MFeOx values were calculated by integrating dC_NFeOx/dlogD_m and dC_MFeOx/dlogD_m, respectively, over a D_m range of 220–4000 nm.

3. Uncertainties of C_NFeOx and C_MFeOx

We evaluated the measurement uncertainty of the C_NFeOx and C_MFeOx using laboratory samples (Section 3.1) and Arctic snow samples (Section 3.2).

3.1. Laboratory samples

We prepared separate wüstite and fullerene soot suspensions of known mass concentrations (i.e., pure suspensions), which we then combined to obtain mixed suspensions with volume ratios (R) with respect to the wüstite component of 0.52, 0.71, and 0.84; the pure wüstite suspension had an R of 1 and the pure fullerene soot suspension had an R of 0. The detailed procedure for preparing these suspensions is described in Appendix A.

Using the relationships between peak S_i and color ratio determined earlier (Figures 1a–c), we derived the size-resolved C_NFeOx and C_MFeOx of the five suspensions (Figures 4a and b). The values of C_NFeOx and C_MFeOx of the pure wüstite suspension were 10 µL⁻¹ and 3.1 μg L⁻¹, respectively, whereas those of the fullerene soot suspension were very close to zero. The values of C_NFeOx and C_MFeOx of the three mixed samples were 5.30–8.60 µL⁻¹ and 1.56–2.90 μg L⁻¹, respectively. We also derived the number and mass concentrations of fullerene soot in these samples within the 75–4000 nm D_m range (C_NBC and C_MBC, respectively). The C_NFeOx/C_NBC and C_MFeOx/C_MBC ratios in the mixed samples were 0.02–0.08 and 0.75–3.45, respectively. These ratios were similar to the averages for the Arctic snow samples (0.02 for the C_NFeOx/C_NBC ratio and 0.99 for the C_MFeOx/C_MBC ratio).

Figure 4. (a) Number and (b) mass size distributions of FeO_x in a pure wüstite suspension, three mixed wüstite–fullerene soot suspensions, and a pure fullerene soot (FS) suspension. The volume ratio of the wüstite suspension to the total volume (R) was 1.0, 0.84, 0.71, 0.52, and 0.0. (c) Number and (d) mass size distributions of FeO_x in a pure wüstite suspension and three mixed suspensions, normalized by total FeO_x number and mass concentration, respectively, and their averages.

We calculated the C_NFeOx and C_MFeOx of the mixed samples by multiplying the measured C_NFeOx and C_MFeOx of the pure wüstite suspension by R (= 0, 0.52, 0.71, 0.84, and 1.0). The C_NBC and C_MBC of the mixed samples were also calculated in a similar way by using a multiplication factor of 1–R.

The shapes of the measured FeO_x number and mass size distributions in the mixed samples changed little with R (Figures 4c and d). The measured C_NFeOx was close to the calculated C_NFeOx value (Figure 5a), and the measured and calculated C_MFeOx values agreed to within 3%, on average (Figure 5b). The measured C_NBC and C_MBC values were also very close to those calculated (Figures 5c and d). These results indicate that the FeO_x particles could be clearly discriminated from the BC particles by using our empirically determined boundary lines, and that the measurement uncertainty of C_NFeOx and C_MFeOx associated with the discrimination procedure was within 3%, on average.

Figure 5. Relationships (a) between the measured and calculated C_NFeOx and (b) between the measured and calculated C_MFeOx. (c, d) Same as (a, b) but for fullerene soot.

3.2. Arctic snow samples

We used snow samples collected in the Arctic to derive the uncertainties of C_NFeOx and C_MFeOx arising from the uncertainties in the factors used to calculate dC_NFeOx/dlogD_m and dC_MFeOx/dlogD_m (i.e., dN_FeOx/dlogD_m, dM_FeOx/dlogD_m, F_neb, V_pump, and ε_FeOx).

The derived N_FeOx and M_FeOx in the carrier gas depended on the boundary lines chosen between FeO_x and BC. The empirically determined boundary lines are shown in Figures 1f–i for the snow samples from North Siberia, Greenland, and Ny-Ålesund, and the precipitation samples from Ny-Ålesund. Together with a pit sample from Greenland (Table 2), the average color ratio of the FeO_x particles was approximately 0.7. In the sample analyses, the YAG laser voltage in the SP2 was lower for samples from Alaska (average 1.97 V) and higher for samples from Finland (average 4.0 V) (Table C1). Median color ratios, derived from the Gaussian fitted frequency distributions of FeO_x and BC particles (Figure 6), were markedly influenced by the average voltage of the YAG laser (Figure 7a). In contrast, the geometric standard deviation (GSD) and the difference of the median color ratios determined for the FeO_x and BC particles changed little with voltage (Figures 7b and c). Thus, the peak S_i–color ratio distributions were shifted to smaller values for the Alaska samples and to larger values for the Finland samples. Due to the lack of representativeness, we excluded the boundary lines for the Alaska and Finland snowpack samples from the derivation of the average boundary line.

Figure 6. Frequency distributions of the color ratio for FeO_x and BC in snowpack, surface snow, and subsurface snow samples from (a) Alaska, (b) Finland, (c) North Siberia, (d) Greenland, and (e) Ny-Ålesund; and in (f) precipitation samples from Ny-Ålesund. The average voltage of the YAG laser beam is shown at the top of each panel. Solid lines indicate Gaussian fitted FeO_x and BC number distributions.

Figure 7. (a) Median color ratio and (b) geometric standard deviation (GSD) derived from the Gaussian fitted frequency distributions for FeO_x and BC in snowpack, surface snow, and subsurface snow samples (Figure 6) versus average voltage of the YAG laser beam. (c) Relationship of the difference of the median color ratio between FeO_x and BC versus the average voltage of the YAG laser beam. Solid lines indicate least squares fitted regressions.

Table 2. Details of the snow sampling conducted in Alaska and Greenland.

Region (Site)	Sampling area (latitude, longitude)	Type of snow sample	Sampling year	Number of samples	Figures in the present study	References
Alaska (FA7)	63.6°N, 148.8°W	Snowpack	2015	5	Figures 9a, b and 10	Mori et al. 2019
Greenland (EGRIP)	75.6°N, 36.0°W	Pit	2017	1	—	Goto-Azuma et al. 2021; Komuro et al. 2021
Greenland (EGRIP)	75.6°N, 36.0°W	Surface snow	2015	1	Figure 8b and c	Mori et al. 2019
Greenland (SC-ALB)	69.6°N, 49.4°W	Surface snow	2015	4	Figures 9c, d and 10	Mori et al. 2019

Figure 8a shows the five remaining boundary lines overlaid on a composite peak S_i–color ratio scatter plot for the 21 surface and subsurface snow samples collected from Greenland (Table 1), which were used to derive the number and mass size distributions of FeO_x in the carrier gases for each boundary line. We used a surface snow sample from East Greenland Ice-core Project (EGRIP) camp in Greenland (Table 2) to investigate the differences in the shapes of the size distributions of FeO_x and found that the shapes of the size distributions did not markedly depend on the boundary lines used (Figures 8b and c). The average (±1σ) values of N_FeOx and M_FeOx were 0.20 ± 0.01 (±6.7%) cm⁻³ and 41 ± 1.0 ng m⁻³ (±2.5%), respectively, indicating the small dependence of these values on the choice of boundary line. The N_FeOx and M_FeOx values of the surface and subsurface snow samples from Greenland, derived by using the five boundary lines, agreed within 10% and 7.0%, respectively. Thus, we concluded that the uncertainties in N_FeOx and M_FeOx associated with the determination of the boundary lines were approximately 10% and 7.0%, respectively.

Figure 8. (a) Five boundary lines discriminating between FeO_x and BC particles overlaid on a scatterplot of the peak incandescence signal and color ratio for the 21 snow samples from Greenland. These boundary lines were empirically determined by using the scatterplots obtained from each sample (i.e., surface and subsurface snow samples from Greenland, surface snow samples from North Siberia, snowpack samples from Ny-Ålesund (Svalbard), pit samples from Greenland, and precipitation samples from Ny-Ålesund). (b, c) Number and mass size distributions of FeO_x (extracted from a Greenland snow sample) in the carrier gas, for each boundary line, respectively. D_m, mass equivalent diameter.

The overall uncertainties of the C_NFeOx and C_MFeOx determinations within the D_m range of 220–1400 nm were approximately 19% and 18%, respectively, as derived from the typical errors for dN_FeOx/dlogD_m (±10%), dM_FeOx/dlogD_m (±7.0%), V_pump (±5.0%), F_neb (±5.0%), and ε_FeOx (±15%).

4. Long-term stability of FeO_x concentrations in water

We also investigated the changes in the FeO_x number and mass size distributions during long-term storage in a refrigerator at 4 °C for five snowpack samples from Alaska and four surface snow samples from Greenland (Table 2). The samples were examined immediately after melting and again 16 months later for the Alaska snowpack samples, and 6 years later for the Greenland surface snow samples. At the first analysis, the average (±1σ) values of C_NFeOx and C_MFeOx for all 9 samples were 6.8 ± 7.1 µL⁻¹ and 1.9 ± 1.7 µg L⁻¹, respectively. For simplicity, we represent C_NFeOx and C_MFeOx at t months or t years after the measurement taken immediately after melting (Day 1) as C_NFeOx (t) and C_MFeOx (t), respectively.

The shapes of the number and mass size distributions of FeO_x in snowpack from Alaska changed little during the 16-month storage period (Figures 9a and b). On average, the values of C_NFeOx (Day 1) and C_NFeOx (16 months) agreed to within 4.0%, and those for mass concentration agreed to within 6.0%. For the surface snow from Greenland, the average C_NFeOx (6 years)/C_NFeOx (Day 1) ratio was 1.17 ± 0.24, and that for mass concentration was 0.72 ± 0.30. The changes in the shapes of the number and mass size distributions of FeO_x were greater than those of the other samples (Figures 9c and d), partly due to the low C_NFeOx and C_MFeOx. In the D_m range of 220–600 nm, on average, the values of C_NFeOx (6 years) and C_NFeOx (Day 1) agreed to within 20%, and those for mass concentration agreed to within 10%.

Figure 9. (a) Number size distributions of FeO_x in a snowpack sample from Alaska measured immediately after melting (Day 1) and 16 months later, and (c) in a surface snow sample from Greenland measured at Day 1 and 6 years later. (b, d) Same as (a, c), but for mass size distributions. Bars indicate ±1σ Poisson distributions. D_m, mass equivalent diameter.

The values of C_NFeOx and C_MFeOx (16 months or 6 years) in all samples were well correlated with those immediately after melting (slope of the regression line = 0.81 and r² = 0.95 for C_NFeOx, Figure 10a; slope of the regression line = 0.99 and r² = 0.91 for C_MFeOx, Figure 10b). The average (±1σ) C_NFeOx (16 months or 6 years)/C_NFeOx (Day 1) ratio was 0.99 ± 0.30, and that for mass concentration was 0.86 ± 0.35. On average, for the melted snow samples, C_NFeOx and C_MFeOx agreed to within 19% and 1.0%, respectively, and the shapes of the size distributions of FeO_x changed little during the 16-month storage period. The high reproducibility of the size distributions of FeO_x in water verified that the measurements after storage were reliable.

Figure 10. (a) Correlation between C_NFeOx determined immediately after melting and 16 months and 6 years later. (b) Same as (a) but for mass concentrations. Solid lines indicate least squares fitted regressions.

5. Measurement of FeO_x in dust particles

Thus far, we have estimated the uncertainty of the measurements of pure FeO_x particles which are not internally mixed with dust particles. Next, we calculated the size range of FeO_x on dust particles. We denote D_total as the mass equivalent diameter of a dust particle and D_FeOx as that of FeO_x present on its surface. We calculated the Tr and ε values for dust particles by assuming a density of 2.7 g cm⁻³ (Hansell et al. 2011) and a shape factor of 1.2 (Adachi et al. 2021). The Tr and ε values for the dust particles gradually decreased with diameter (Figure 11). The ε values were somewhat higher than those for the magnetite particles (i.e., pure FeO_x particles) because of the lower density of the dust particles. The maximum D_total that can be substantially detected by the system is approximately 3000 nm.

Figure 11. Same as Figure 3, but for dust particles, assuming a density (ρ) of 2.7 g cm⁻³ and a shape factor of 1.2. For comparison, the theoretically calculated transmission efficiency and extraction efficiency for a magnetite particle are also shown.

We define (3) $$\alpha =\frac{D_{\mathrm{FeOx}}\mathrm{ }}{D_{\mathrm{total}}} <1$$(3) as a measure of the degree of internal mixing of each FeO_x particle. In the Arctic, D_total is often larger than 1000 nm (Adachi et al. 2021; Nagatsuka et al. 2021). Adachi et al. (2021) showed an example of a dust particle with a D_total of ∼2000 nm and a D_FeOx of ∼100 nm, as measured by transmission electron microscopy. In this case, α is about 0.05.

The maximum D_FeOx that can be measured by the system will be about 3000α nm, considering that the maximum detectable D_total is about 3000 nm. For α = 0.25 and 0.5, the upper detection limit of D_FeOx is estimated to be 750 and 1500 nm, respectively. Actually, α will be considerably lower than these values, as suggested by the example of Adachi et al. (2021). Therefore, it is likely that the majority of pure FeO_x contained in dust particles will be in the size range smaller than 1000 nm for reasonable values of α.

6. Summary

Here, we developed a technique for measuring the size distributions of light-absorbing iron oxide (FeO_x) in water by using a combined concentric pneumatic nebulizer and modified wide-range single-particle soot photometer. To estimate the overall accuracy of determinations of the number and mass concentrations of FeO_x in water (C_NFeOx and C_MFeOx, respectively), we investigated the uncertainties of the key factors involved in the measurements: selective detection and calibration of FeO_x; extraction efficiency of the nebulizer for FeO_x particles; overall error estimates; and stability of FeO_x concentrations in water during storage.

We first investigated the relationship between the peak incandescence signals of the blue band and color ratios determined with our wSP2 for finely powdered wüstite, fullerene soot, and mixed wüstite–fullerene soot samples. We demonstrated that the wSP2 separately detects FeO_x and BC particles. For several melted Arctic snow samples, FeO_x and BC particles were discriminated in the resulting scatterplots by using empirically determined boundary lines, taking into account the dependence of color ratio on the YAG laser voltage.

We calculated the nebulizer extraction efficiency (ε) for magnetite particles using an assumed density of 5.17 g cm⁻³ and shape factor of 1.52. For magnetite particles within 50 nm < D_m < 650 nm, the ε values were about 50% and gradually decreased with increasing D_m above this upper limit.

The measurement uncertainties of C_NFeOx and C_MFeOx, associated with the separation of FeO_x from BC, were estimated to be less than 3.0% for mixed wüstite–fullerene soot suspensions. The variabilities of the number and mass concentrations of FeO_x, arising from five empirically determined boundary lines discriminating between FeO_x and BC in the scatterplots obtained for snow samples, were 10% and 7.0%, respectively. The overall measurement uncertainties of C_NFeOx and C_MFeOx within the D_m range of 220–1400 nm were 19% and 18%, respectively.

We evaluated the reproducibility of the number and mass size distributions of FeO_x in water for several melted Arctic snow samples after storage for 16 months or 6 years. On average, C_NFeOx and C_MFeOx were stable to within 19% and 1.0%, respectively, indicating that the number and mass size distributions measured at 16 months after the samples were initially melted were not markedly altered.

We estimated that the maximum diameter of dust particles detectable by our nebulizer–wSP2 system was about 3000 nm. The major portions of FeO_x on dust particles measured by the present system were estimated to be in the size range smaller than 1000 nm with reasonable values for the degree of internal mixing.

We expect the present measurement technique to be useful for improving quantitative understanding of the wet deposition of FeO_x. In future studies, it will be important to compare FeO_x mass concentrations measured by the present approach with total Fe concentrations measured by other methods, such as inductively coupled plasma mass spectrometry.

Nomenclature
Dm	=	mass equivalent diameter [nm]
CMFeOx	=	mass concentration of light-absorbing FeOx in liquid water [µg L−1]
CNFeOx	=	number concentration of FeOx in liquid water [µL−1]
εFeOx	=	nebulizer extraction efficiency of FeOx particles [%]
Fneb	=	nebulizer flow rate at standard temperature and pressure [STP cm3 s−1]
Vpump	=	peristaltic pump flow rate [µL s−1]
Tr	=	transmission efficiency [%]
MFeOx	=	mass concentration of FeOx in a carrier gas [STP ng m−3]
NFeOx	=	number concentration of FeOx in a carrier gas [STP cm−3]

Acknowledgments

The authors thank J. Matsushita, T. Aoki, F. Nakazawa, W. Shigeyama of the National Institute of Polar Research; A. Sato, formerly of the Snow and Ice Research Center; A. Tsushima of Chiba University; S. Omiya of the Civil Engineering Research Institute for Cold Region; S. Matoba, A. Sugimoto, and S. Takano of Hokkaido University; M. Schneebeli and K. Steffen of the WSL Institute for Snow and Avalanche Research; and the staff of the Norwegian Polar Institute for collecting the precipitation samples at Ny-Ålesund and melted snow (i.e., surface, subsurface, snowpack, and pit) samples from Alaska, Finland, North Siberia, Ny-Ålesund, and Greenland. The pit samples were collected under The East Greenland Ice-Core Project (EGRIP) directed and organized by the Center of Ice and Climate at the Niels Bohr Institute and US NSF, Office of Polar Programs. We also thank K. Adachi of Meteorological Research Institute for useful comments on FeO_x on dust particles.

Data availability statement

The data that support the findings of this study are openly available at https://ads.nipr.ac.jp/dataset/A20220428-001.

Additional information

Funding

This work was supported by the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT); Japan Society for the Promotion of Science KAKENHI Grants (JP1604452, JP18H04140, JP19H05699, JP19K20441, JP20H04980, JP20H00638, and JP22H01294); the GRENE Arctic Climate Change Research Project; the Arctic Challenge for Sustainability (ArCS) project (JPMXD1300000000); the Arctic Challenge for Sustainability II (ArCS II) project (JPMXD1420318865); and the Environment Research and Technology Development Funds (JPMEERF20172003 and JPMEERF20202003) of the Environmental Restoration and Conservation Agency of Japan. EGRIP is supported by funding agencies and institutions in Denmark (A. P. Møller Foundation, UCPH), the United States (US NSF, Office of Polar Programs), Germany (AWI), Japan (NIPR and ArCS), Norway (BFS), Switzerland (SNF), France (IPEV, IGE), and China (CAS).

Appendix A

Three suspensions with known mass concentrations (a pure wüstite suspension, a pure fullerene soot suspension, and a mixed wüstite-fullerene soot suspension) were prepared as follows:

Pure wüstite suspension. Wüstite granules (Kojundo Chemical Laboratory Co., Ltd., Tokyo, Japan) were ground with a spatula. Using an electronic balance (MS603, Mettler Toledo, Tokyo, Japan), 1.12 g wüstite was added to 100 mL ultrapure water in a 100-mL glass beaker, and mixed well using a vortex mixer (SRT010AA, Advantec MFS, Inc., Tokyo, Japan). Then, 1 mL methanol was added to the mixture, followed by mixing with the vortex mixer and sonication in an ultrasonic bath for 15 min.

Pure fullerene soot suspension. Using the same electronic balance, 0.020 g finely powdered fullerene soot (Alfa Aesar, Inc., Ward Hill, MA, USA) was weighted and added to 500 mL ultrapure water in a 500-mL glass beaker. After mixing well, 6 mL methanol was added to the mixture.

Mixed wüstite–fullerene soot suspension. Due to the high mass concentrations of the wüstite and fullerene soot suspensions, we diluted the two pure suspensions by a factor of about 3000 and 1000, respectively, and then mixed the dilutions together. The volume ratio of the diluted wüstite suspension volume to the total volume was adjusted to 0.52, 0.71, and 0.84.

Appendix B

Figure B1. Same as Figures 1a–c, but showing values for the S_i range from 3.5 to 6.0.

Appendix C

Table C1. Average YAG laser voltages, median color ratios and geometric standard deviations for the frequency distributions of FeO_x and BC particles, and the differences of the median color ratios determined for the FeO_x and BC particles.

		FeOx particles		BC particles		Difference of the median color ratio determined for the FeOx and BC particles
Region	Average (±1σ) YAG laser voltage (V)	Median color ratio	Geometric standard deviation	Median color ratio	Geometric standard deviation
Alaska	1.97 ± 0.10	0.59	1.08	1.08	1.10	0.49
Finland	4.00 ± 0.06	0.87	1.10	1.39	1.15	0.52
North Siberia	3.40 ± 0.06	0.70	1.11	1.21	1.13	0.51
Greenland	3.54 ± 0.07	0.71	1.00	1.27	1.13	0.56
Ny-Ålesund (snowpack samples)	3.03 ± 0.04	0.68	1.06	1.19	1.11	0.51
Ny-Ålesund (precipitation samples)	2.81 ± 0.07	0.69	1.07	1.22	1.11	0.53