Optimized low-level liquid scintillation spectroscopy of ³⁵S for atmospheric and biogeochemical chemistry applications

Abstract

Anthropogenic activities, dominated by emissions of sulfur dioxide (SO₂), have perturbed the global sulfur (S) cycle. Uncertainties in timescales of S transport and chemistry in the atmosphere lead to uncertainties in the predicted impact of S emissions. Measurements of cosmogenic ³⁵S may potentially be used to resolve existing uncertainties in the photochemical and chemical transformation of S in the environment. The lack of a simple, effective, and highly sensitive technique to measure ³⁵S activity in samples with low activities may explain the scarcity of published measurements. We present a set of new sample handling and measurement procedures optimized for the measurement of ³⁵S in natural samples with activities as low as 0.20 dpm above background (2σ, integration time = 2 hr). We also report simultaneous measurements of aerosol () and gas phase () collected at inland and coastal locations; the range of observed activities corresponds to SO₂ residence lifetimes of 0.2 ± 0.04 (coastal) - 22.3 d ± 0.04 (inland). These optimized techniques offer the potential for resolving atmospheric processes that occur on 6–12-hour timescales as well as resolving transport phenomena such as stratospheric mixing into the troposphere.

According to the Intergovernmental Panel on Climate Change (1) anthropogenic emissions of sulfur dioxide have perturbed the sulfur cycle on local, regional, and global scales. These emissions and their oxidized end products lead to increases in acid rain (H₂SO₄) and aerosol sulfate () concentrations that play a significant role in climate and the global sulfur cycle. Sulfur-containing aerosol particles serve as cloud condensation nuclei affecting cloud formation and the hydrological cycle. Our knowledge of the chemical and photochemical processes that govern the chemical transformations and transport of sulfur compounds in the atmosphere is incomplete due to the complex, multivalent nature of sulfur and uncertainties in the understanding of aerosol chemistry. Sulfur in the atmosphere exists simultaneously as a solid and gas, further complicating matters. The development of new and/or improved analytical techniques to study the sulfur cycle on short timescales (hours to days) is, therefore, of considerable importance. Here we describe significant advances in the detection sensitivity of the cosmogenic isotope ³⁵S that can be used as a tracer of gas and aerosol phase lifetimes and turnover kinetics with high time and aerosol size resolutions.

Stable isotopic measurements of atmospheric species such as nitrate () and sulfate () have recently been used to provide strong constraints on the oxidative processing of their precursors, NO_x and SO_x, in the atmosphere (2–6). The radionuclide ³⁵S (β-decay to ³⁵Cl, t_1/2 = 87.4 d) is continuously produced in the atmosphere by the interaction of cosmic rays with ⁴⁰Ar and provides an additional opportunity for tracing atmospheric processes. Upon production, ³⁵S rapidly oxidizes to ³⁵SO (lifetime < 1 ms) and to (7, 8). In the atmosphere, undergoes wet deposition (removal below and within clouds), dry deposition (gravimetric settling or interactions with surfaces), or may be oxidized to sulfate () and incorporated onto aerosol particles (which eventually undergo dry and wet deposition). The sink reactions occur on timescales of a few hours to a few days, depending on the local atmospheric environment; thus the concentration of and is expected to vary significantly as a function of time, meteorology, humidity, and location. These variables can vary on timescales of hours to days, which highlights the need of the present work of developing high time resolution measurement capabilities.

The chemical properties of and are expected to be nearly identical to SO₂ and , respectively. Measurements of ³⁵S in aerosol sulfate may be used to better resolve aerosol aging and chemistry, lifetimes of aerosol and gas phase SO₂, and provide a better measure of boundary layer dynamics (9, 10). Previous ³⁵S measurements have been used to calculate SO₂ fluxes (9, 11–14) and depositional rates (wet and dry) (9, 12, 15, 16) and trace atmospheric sulfate deposition into lakes, rivers (17), and catchments (18–22). Some of these papers have included measurements of in bulk aerosols (9, 23).

Because the atmospheric aerosol chemistry of sulfur depends on the type, size, and number density of aerosol particles, one of the principal goals of this work is to improve existing measurement techniques so as to obtain particle-size-resolved measurements of ³⁵S in aerosol sulfate without the need to collect samples for long periods of time that may mask short-term variability.

Recently, a technique using low-level liquid scintillation spectroscopy (LSS) was developed to measure ³⁵S in lake water (17). Our review of this technique revealed that the methods used introduced high backgrounds, prohibiting their use for determining the activities of natural samples with low ³⁵S/S ratios. Here we present improved sample handling and analysis techniques employing LSS and our first field results. The improvements include unique sample preparation procedures, identification and correction of previously unreported backgrounds, as well as a method for optimizing the integration of scintillation spectra. Our improved methods allow for the measurement of the abundance of as a function of aerosol size as well as with sample collection times as short as 12 hr at a coastal and an inland location at similar altitudes and latitudes. Using these measurements, we estimated the overall lifetime of SO₂ at these two locations.

The increased sensitivity to ³⁵S concentrations presented in this paper should allow for the study of a broader range of environmental processes such as boundary layer dynamics, evolution of sulfate in aircraft plumes in the lower stratosphere, deposition of sulfate into the hydrological cycle, kinetics of SO₂ oxidation, and aerosol dynamics (including the transport and evolution of sulfur in remote locations such as Antarctica). In summary, the enhanced sensitivity of our method expands the range of biogeochemical processes that can now be explored.

Results and Discussion

Method Testing.

A Wallac Quantalus 1220 Ultra Low-Level Liquid Scintillation Counter was used for all the measurements and optimization tests that we report here. This instrument minimizes cosmic ray backgrounds using passive and active shielding, multiple multichannel analyzers, and anticoincidence counting techniques. A standard (MP Biomedical) was used to optimize the sample preparation techniques. The instrument’s detection efficiency (ε_D) was determined using a calibrated (absolute) ¹⁴C activity standard from Perkin Elmer (dpm = 96,200)* (24). By comparing the instrument reported count rate (counts per minute, cpm) and the expected activity (disintegrations per minute, dpm) for the absolute standard, the instrument’s overall detection efficiency () was determined to be 0.946. This efficiency is an upper limit to the overall efficiency of the detection technique presented here, since natural samples will also incur losses in their measured activity by being lost during preparation. Given the high activity of the ¹⁴C standard, this efficiency determination is unaffected by electronic noise or sample vial backgrounds.

Energy channel optimization.

The detection system bins photon pulse events into one of 1,024 energy channels, with the higher channels corresponding to highest energies. The β-decay spectrum of ³⁵S is well defined with a maximum decay energy of 0.1675 MeV (15). If the energy spectra of different cosmogenic species are sufficiently distinct, partial-to-complete discrimination between decay events of ³⁵S and those of other radio-nuclides present in the sample (e.g. ¹⁴C) can be achieved by selective integration of the energy spectra. Fig. S1 displays the spectra measured for ³⁵S, ¹⁴C, and radioactive isotopes of barium. While each channel contributes noise to the integrated count rate, integrating counts only in the energy range where ³⁵S decay events occur results in improved signal to noise ratios. In collected aerosol and gas samples, organic molecules containing ¹⁴C may contribute to the background activity, and our optimization considers these contributions. Using the spectra shown in Fig. S1, the ratio of the time- and channel-integrated signal [S_m(ch1,ch2)] and background [B_m(ch1,ch2)] ratio can be used to optimize the counting of samples and is defined as

[1]

where ch1 and ch2 are the start and end channels of integration, respectively, and and are the observed count rates in channel i of ³⁵S and interfering backgrounds (including ¹⁴C, radioactive barium isotopes, and/or scintillation gel). This ratio is plotted separately for the backgrounds expected from barium, the scintillation gel, and ¹⁴C as a function of ch2 in Fig. S2. The ratio of reaches a maximum value close to the channel where the ³⁵S decay spectrum reaches its maximum (ch2 ∼ 450). Integration between 1 (ch1) to 445 (ch2) and 455 (ch2) captures 95% and 97% of all decay events under the ³⁵S spectrum shown in Fig. S2 while minimizing background contributions.

Taking measurements of β^- activities using LSS requires using a scintillation gel that fluoresces when excited by emitted β^- particles as well as a vial to hold the sample-gel mixture. The choice of scintillation vial material, preparation techniques, and the amount of scintillation gel volume used can affect the sensitivity of measurements in LSS.

Major gains in sensitivity were attained by minimizing the background activities of scintillation vials and reagents used in the preparation of samples. The most significant background reductions were obtained by using plastic vials (Fisherbrand, 20 mL, ∼0.145 dpm) as opposed to glass vials (Wheaton Glass-20 mL, 0.45 dpm). Similarly, glass fiber filters (Whatman GF/B), like those recommended by Hong and Kim (17) to isolate and count lake water sulfate as BaSO₄, also possessed relatively high activities (∼2.6 dpm per 2 × 2 inch section). These activities likely originate from the decay ⁴⁰K present in borosilicate glass and, as a result, we avoided the use of glass vials and fiber paper in our subsequent tests and natural sample measurements. Microquartz filters were not tested in this study.

The scintillation gel contributes to the overall activity of a sample but is a strictly necessary reagent for measuring samples because of its role in converting the energy loss from a decay event into photons that are detected by the photomultiplier tubes of the Quantulus. The use of optimized integration channels () resulted in a 30% reduction in the background contributed by the scintillation cocktail for a given gel volume (V_G). The activity measurements reported below use the optimized integration channels. Because measurements of empty vials yielded negligible (< 0.01 dpm) activities, the observed background activity in scintillation vials filled with variable amounts of gel must come from the gel itself. This gel-specific background () is a nonlinear function of V_G and can be fit by a second-order polynomial with respect to V_G as

[2]

Using a (aq.) standard (MP Biomedical, ∼4,000 dpm/mL of solution of H₂O) to ensure that , we determined the counting efficiency of the liquid scintillation technique for aqueous samples of volume 0.1 mL and 1 mL using various values of V_G. These counting efficiencies are shown in Fig. 1. Polynomial fits to these data yield the following counting efficiencies as a function of gel volume:

[3]

[4]

Fig. 1.

³⁵S decay event detection efficiency [ε(V_G)] as a function of gel volume (V_G) used for 0.1 mL and 1 mL aqueous samples. Note that the counting efficiency exceeds 90% for smaller amounts of gel compared to aqueous samples with larger volumes.

Optimizing V_G to maximize sensitivity in weak samples.

We determined the optimal scintillation gel volume as follows: We assumed that an aqueous sample of volume V_SA [0.1 or 1 mL of NaSO₄ (aq.)] is to be counted. Because we want to determine the sample’s unknown activity (dpm) in a gel volume V_G, we need to know what amount of gel volume maximizes our signal to noise. The expected signal, S_m, that is reported by the scintillation system during an integration time Δt is given by

[5]

where ε_D is the overall detection efficiency of the Quantulus detection system (∼0.94) and ε_V(V_G) is the gel volume dependence of the detection efficiency (e.g. Eq. 3 or 4) for an aqueous sample, and ε_prep is the sample preparation efficiency. We note that in our tests with laboratory standards, the sample preparation efficiency was 100% (ε_prep = 1). The noise in the Quantalus scintillation spectroscopic system is dominated by fluctuations in the total number of events detected in the same time interval and is given by

[6]

The ratio is plotted as a function of V_G in Fig. 2 for a weak sample with and Δt = 1 and 2 hr. Examination of Fig. 2 reveals that the signal to noise (and hence the minimum detection limit, MDL) can be maximized for 1 mL aqueous samples when V_G = 2.5 mL. Even more important, however, is the observation that smaller aqueous volumes, in this case 0.1 mL, yield substantially improved MDLs. The main reason for this behavior is that the efficiency of detection [ε_V(V_G)] as a function of V_G for 0.1 mL aqueous samples is already close to 90% when V_G = 1. Additional gains in efficiency obtained by using larger gel volumes are marginal and are offset by the higher backgrounds associated with the scintillation gel. Thus, contrary to previous method descriptions, we find that minimizing the gel volume can significantly improve the MDL for the LSS of ³⁵S.

Fig. 2.

Signal/noise () for 0.1 and 1 mL aqueous samples as a function of gel volume. Note that smaller amounts of water lead to drastic improvements in the minimum detection limit.

Background interferences from barium isotopes.

See SI Methods.

Preparation efficiency.

See SI Methods.

Natural Sample Results.

We collected and measured size-segregated (SI Methods) atmospheric aerosol and SO₂ samples at the Scripps Institute of Oceanography (SIO, latitude = 32.867°N, longitude = 117.257°W) and San Fernando Valley (SFV, latitude = 34.212°N, longitude = 118.061°W) during September 2007. These samples were processed before the optimized techniques described here were developed, and their activities were measured as BaSO₄. Size-segregated aerosol sulfate and gaseous SO₂ samples (SI Methods) were also collected during July 2008 and were measured as Na₂SO₄(aq). The uncertainties, determined by counting statistics, for the 2007 samples are larger than those associated with the 2008 samples due to the difficulties in preparing and measuring these samples as BaSO₄, as we have discussed in our measurement techniques. Overall, the background corrected activities we observed ranged from 0 (within the measurement uncertainty) to 22.20 ± 1.22 dpm (see Tables 1 and 2). All samples were recounted once the naturally present ³⁵S had fully decayed and were spiked with 0.1 mL of the ³⁵S standard. This procedure allowed us to correct for sample self-absorption of scintillation events. The corrected count rates are reported in Tables 1 and 2.

Table 1.

³⁵S activity of aerosol found in particles > 1.5 μm in size at Scripps Institution of Oceanography and San Fernando Valley

Sample	dpm per sample		dpm/μmol S	dpm/μmol nss-S	%NSS
SFV 9/3/2007*	4.90 ± 3.83	453 ± 347	0.21 ± 0.16	0.34 ± 0.26	60.2%
SFV 9/7/2007*	-2.06 ± 3.61	- 187 ± 328	-0.05 ± 0.09	-0.12 ± 0.21	42.9%
SFV 9/16/2007*	-0.88 ± 3.11	-80 ± 282	-0.02 ± 0.09	-0.05 ± 0.19	45.9%
SFV 7/21/2008	16.47 ± 1.06	1494 ± 96	0.43 ± 0.03	1.23 ± 0.08	34.9%
SFV 7/25/2008	-0.30 ± 1.33	-27 ± 121	-0.01 ± 0.03	-0.02 ± 0.11	29.4%
SFV 7/28/2008	5.62 ± 1.22	510 ± 111	0.16 ± 0.03	1.01 ± 0.22	15.5%
SVF 8/1/2008	3.84 ± 1.09	349 ± 99	0.09 ± 0.02	0.22 ± 0.06	39.8%
SIO 9/4/2007*	-0.021 ± 0.643	-2 ± 58	0 ± 0.01	-	0
SIO 9/7/2007*	-0.122 ± 0.653	-11 ± 59	0 ± 0.01	-	0
SIO 9/11/2007*	-0.138 ± 0.691	-16 ± 63	0 ± 0.007	-	0
SIO 7/21/2008	2.16 ± 0.84	196 ± 76	0.36 ± 0.14	-	0
SIO 7/25/2008	3.06 ± 0.86	278 ± 78	0.10 ± 0.03	-	0
SIO 7/28/2008	-1.62 ± 0.74	-147 ± 67	-0.06 ± 0.03	-	0
SIO 8/1/2008	2.26 ± 0.77	205 ± 70	0.07 ± 0.02	-	0

Samples measured as BaSO₄(s) are indicated using *. All other samples (2008) were measured as Na₂SO₄(aq). Air volume per sample (V_air) was equal to 2,000 m³ and the total number of ³⁵S(³⁵N) in each sample was determined using the relationship , where is the radioactive-decay half-life of ³⁵S expressed in minutes and dpm is the measured activity of the sample in units of disintegrations per minute at the time of collection. The volumetric concentrations of ³⁵S were calculated as . Non-sea-salt sulfate was determined using anion and cation measurements of aerosol samples and the molar ratio of sulfate to sodium (0.0604) in sea-salt spray.

Table 2.

³⁵S activity of aerosol found in particles < 1.5 μm in size at Scripps Institution of Oceanography and San Fernando Valley

Sample	dpm per sample		dpm/μmol S	dpm/μmol nss-S	%NSS
SFV 9/3/2007*	10.84 ± 2.87	984 ± 260	0.18 ± 0.05	0.28 ± 0.07	64.1%
SFV 9/7/2007*	13.63 ± 2.84	1236 ± 258	0.27 ± 0.06	0.40 ± 0.08	68.9%
SFV 9/16/2007*	13.18 ± 2.59	1196 ± 235	0.18 ± 0.03	0.27 ± 0.05	66.4%
SFV 7/21/2008	3.76 ± 1.06	341 ± 96	0.03 ± 0.01	0.03 ± 0.01	88.3%
SFV 7/25/2008	20.43 ± 1.33	1854 ± 121	0.10 ± 0.01	0.12 ± 0.01	88.7%
SFV 7/28/2008	22.20 ± 1.22	2013 ± 111	0.16 ± 0.01	0.19 ± 0.01	84.7%
SFV 8/1/2008	9.80 ± 1.09	889 ± 99	0.04 ± 0.00	0.04 ± 0.00	92.7%
SIO 9/4/2007*	-1.09 ± 2.53	-99 ± 230	- 0.01 ± 0.03	-0.01 ± 0.03	75.9%
SIO 9/7/2007*	-1.94 ± 2.29	-176 ± 207	-0.04 ± 0.05	-0.06 ± 0.07	67.3%
SIO 9/11/2007*	3.91 ± 2.45	355 ± 223	0.08 ± 0.05	0.13 ± 0.08	61.8%
SIO 7/21/2008	11.46 ± 0.97	1040 ± 88	0.55 ± 0.05	0.85 ± 0.07	64.0%
SIO 7/25/2008	3.26 ± 0.91	295 ± 83	0.03 ± 0.01	0.04 ± 0.01	81.4%
SIO 7/28/2008	0 ± 0.90	0 ± 82	0.00 ± 0.01	0.00 ± 0.02	72.2%
SIO 8/1/2008	6.39 ± 1.10	580 ± 99	0.06 ± 0.01	0.15 ± 0.03	37.0%

Samples measured as BaSO₄(s) are indicated using *. All other samples (2008) were measured as Na₂SO₄(aq). Air volume per sample (V_air) was equal to 2,000 m³ and the total number of ³⁵S(³⁵N) in each sample was determined using the relationship , where is the radioactive-decay half-life of ³⁵S expressed in minutes and dpm is the measured activity of the sample in units of disintegrations per minute. The air volume concentrations of ³⁵S were calculated as . Non-sea-salt sulfate was determined using anion and cation measurements of aerosol samples and the ratio of sulfate to sodium (0.0604) in sea-salt spray.

While detailed modeling of the observed absolute activities (dpm) is beyond the scope of this work, examination of the ³⁵S activity in the gas and aerosol phases points to the potential of the methods developed here to address present uncertainties in our understanding of the global S budget. We found, with the exception of a few days, that the activities contained within aerosol particles < 1.5 μm were generally higher than those within particles of diameter > 1.5 μm. Despite their proximate latitudes and longitudes, the activities in both size fractions were generally higher at SFV than at SIO. The fraction of non-sea-salt sulfate (NSS) was measured and calculated (SI Methods) for each sample using the total concentration of sulfate and sodium and the molar ratio of sulfate to sodium (0.0604) found in seawater. The SIO samples had a much higher fraction of the total sulfate derived from sea salt, with the coarse aerosols dominated by sea-salt sulfate. The lower ³⁵S activity in these samples is consistent with these aerosol particles being fresh sea-salt aerosols without any measurable amounts of secondary sulfate produced by the oxidation of . A few coarse samples at SIO have an activity > 0, yet have Na⁺ abundances that indicate that 100 ± 8% of the sulfate is from sea-salt spray. An estimate of the amount of non-sea-salt sulfate that would be needed to explain this apparent discrepancy was obtained using the molar ratio of ³⁵S/S ratio from SO₂ collected at SIO, and this estimate indicates that the activity falls within the total measurement uncertainty of the ion-chromatographic method used to determine the cation and anion concentrations.

To provide the proper context to relate our and concentration measurements (reported as activities) to atmospheric processes, we follow a treatment similar to that of Ref. 12 in relating the concentration of to the overall lifetime of SO₂ (τ_overall) in an atmospheric box model. For a stable boundary layer and ignoring mixing from outside of this box, the time-dependent concentration of can be expressed as

[7]

where P_CR is the cosmic ray production rate of ³⁵S. The timescales for sinks of are represented by an oxidative lifetime (τ_ox), a dry deposition lifetime (τ_dd), the lifetime with respect to wet deposition (τ_wd), and the radioactive lifetime (τ_decay) of ³⁵S (radioactive mean-life = 126 days). Combining all sink terms proportional to , defining , assuming steady-state conditions within the box (), and solving the resulting expression for τ_overall yields

[8]

The production rate of ³⁵S in the troposphere reported by Ref. 13 of 4.75 × 10^-4 atoms m^-3 s^-1 was used to determine the integrated production in a 10 km high volume. The results are presented in Table 3.

Table 3.

Summary of concentrations and τ_overall estimated using Eq. [9] and

Sample date	dpm per sample		τoverall (days)
SFV
7/21/2008	4.80 ± 0.19	436 ± 17	10.6 ± 0.4
7/25/2008	9.03 ± 0.18	820 ± 16	20.0 ± 0.4
7/28/2008	6.21 ± 0.17	565 ± 16	13.8 ± 0.4
8/1/2008	10.07 ± 0.19	915 ± 18	22.3 ± 0.4
SIO
7/21/2008	0.10 ± 0.19	9 ± 17	0.2 ± 0.4
7/25/2008	1.19 ± 0.17	108 ± 15	2.6 ± 0.4
7/28/2008	2.02 ± 0.19	184 ± 17	4.5 ± 0.4
8/1/2008	1.81 ± 0.14	165 ± 12	4.0 ± 0.3

Air volume per sample (V_air) was equal to 2,000 m³ and the total number of ³⁵S(³⁵N) in each sample was determined using the relationship: , where is the radioactive-decay half-life of ³⁵S expressed in minutes and dpm is the measured activity of the sample in units of disintegrations per minute. The air volume concentrations of ³⁵S were calculated as .

Inspection of Table 3 reveals significant and systematic differences in τ_overall between the inland (SFV) and oceanic (SIO) sites that likely reflect differences in the predominant mechanisms affecting the lifetime of SO₂ at these two locations. Given the absence of precipitation events during the sample collection times, we ignore the contribution of wet deposition (τ_wd → ∞). Surface resistance models indicate that large differences in deposition velocities (v_dd) exist for SO₂ over land (0.2–0.4 cm s^-1) and seawater (∼0.8 cm s^-1) (25). For a mixing height of 1000 m, these v_dd values translate to deposition lifetimes of about 5.78–2.89 and 1.5 d for SFV and SIO, respectively. An additional factor that determines τ_overall at each of these locations is the removal of SO₂ by OH. This lifetime is given by

[9]

where k₁ is the effective rate constant for the reaction (k₁ ∼ 3 × 10^-12 cm³ s^-1) (26), and 〈[OH]〉_t is the time-averaged (over 24 hr) concentration of OH per cm³. Inserting values typical of polluted environments (〈[OH]〉_t ∼ 1 × 10⁶ cm^-3) leads to estimates for τ_ox ∼ 3–4 d, although the oxidation lifetime of SO₂ in the free troposphere may be longer. At this time, we do not know why the SFV and SIO lifetimes are longer than expected based on oxidative and dry deposition considerations, but these differences may be due to boundary layer shifts and/or mixing of air from the free troposphere. Detailed modeling of these factors, in the future, should help in clarifying their importance.

A similar analysis can be made to interpret the concentration data presented in Tables 1 and 2. To begin, we express the overall rate of change of as

[10]

where τ_ox is again the overall lifetime of SO₂, τ_dd(SO₄) and τ_wd(SO₄) are the dry and wet deposition timescales of aerosol sulfate, and τ_decay is defined as before. Given the ability to discriminate size-dependent activities in aerosols, the activity in aerosol sulfate can be subdivided into two size bins and expressed as

[11]

and

[12]

Here, the oxidative [τ_ox(F,C)], dry deposition [τ_dd(F,C)], and wet deposition [τ_wd(F,C)] lifetimes are broken up into two components to correspond to fine (F) and coarse (C) aerosols. Again, if we assume steady-state conditions and combine the sink terms into one overall timescale for fine and coarse aerosols (), we find that

[13]

and

[14]

The higher (lower) activities per unit volume observed in fine (coarse) aerosol particles, therefore, are expected to depend on (1) the concentration of , (2) the ratio of the overall residence time for fine (coarse) aerosol sulfate, and (3) the rate of SO₂ oxidation and incorporation into fine (coarse) aerosol particles. In this discussion, we neglect the possibility that ³⁵S found in fine aerosol particles may end up in coarse aerosol particles due to aerosol size growth dynamics, although these timescales (hours to days) are expected to depend on total particle concentrations and characteristic particle sizes (27), and this process may be important for very polluted environments such as SFV. If coagulation were the dominant mechanism affecting the relative ratio of activity in fine and coarse aerosols, however, we would expect to see the same ratio of ³⁵S to total S in both of these sizes. Given the above discussion, it is interesting to note that the days with nonzero activities in SIO coarse particles in 2008 also corresponded to days with the lowest (shortest τ_overall) concentrations of . These observations are consistent with an increase of the heterogeneous processing of SO₂ onto coarse aerosol particles and not to changes in the deposition velocities of SO₂, which would only lead to a decrease in the concentration. Given the prevalence of sea-salt particles as a medium for the aqueous phase oxidation of SO₂, we suggest that these observations may be consistent with enhanced uptake and aqueous phase oxidation of during these time periods. Future work with larger datasets and detailed modeling of the coupled gas and aerosol phase chemistry and ambient meteorological conditions will be needed to fully interpret these types of results and their variation. We emphasize that without the optimized methods that we have developed, an in-depth examination of short timescale (∼6–12 hours) processes would not be possible.

In contrast to SIO coarse particles (∼0–3 dpm per 2000 m³), the SFV coarse aerosol size fraction had significantly higher (∼0–16 dpm per 2000 m³) concentrations. In light of Eqs. 13 and 14, these differences may be explained in part by the higher concentrations of seen at SFV (∼4–10 dpm per 2000 m³) compared to SIO (∼0–2 dpm per 2000 m³) and/or substantial differences in the aerosol age (lifetime).

We compare the ³⁵S measurements presented here to the pioneering work of Tanaka and Turekian (9), which reported ³⁵S activities in aerosols collected weekly in 1992 at New Haven, Connecticut. Air samples collected during July and August of 1992 were reported to have activities ranging from 1.1 × 10^-3–14.5 × 10^-3 dpm/μmol S for SO₄ and 6 × 10^-4–2.2 × 10^-3 dpm/μmol S for SO₂. We note that these values are significantly lower than our measurements, which are summarized in Table 1. These differences in the total-sulfur normalized values reported by Tanaka and Turekian (9) and the present values may result from the dilution generated by larger regional SO₂ fluxes emitted by coal burning in the eastern United States. These high sulfur coal emissions in essence dilute the ³⁵S activity when normalized to the total mass of S in aerosols. In contrast, California does not operate coal burning electric plants or have any other large sources of SO₂ that would dilute by a comparable factor. These results highlight the utility of the ³⁵S, as first shown by Tanaka and Turekian (12), to understand differences in regional atmospheric chemistry and transport. Reporting ³⁵S activity with respect to sampled air volume minimizes the dilution effect described above, but normalization has other uses, and thus reporting in both units is advisable.

Conclusion

This work presents improved and optimized methods for measuring ³⁵S in natural samples. These methods have reduced the limit of detection down to 0.200 dpm or 36,000 ³⁵S atoms for a one-hour integration and have been shown to be suitable for measurements of in size-segregated aerosols and . We have identified sensitivity-limiting backgrounds in chemical reagents used for sample preparation and these findings should be considered in future ³⁵S measurements. In light of this work, we suggest that reports of excess ³⁵S in aerosols, in comparison to ⁷Be or ³²P, may be erroneously high due to the high backgrounds that may be introduced by barium and suggests that the conclusions of Osaki et al. (23) may need to be revised. We recommend that ³⁵S activity measurements should be made as NaSO₄ rather than BaSO₄ in weakly radiogenic samples.

In the present work we have also simultaneously measured and aerosol in two distinct locations. The ³⁵S activity differences observed between the two locations, as well as the differences observed in the two aerosol size fractions, illustrates the ability of ³⁵S to further elucidate the local and regional cycling of SO₂ and aerosol SO₄. Given the observed ³⁵S concentrations and the minimum detection limits presented here, use of these techniques should allow for high time resolution studies of SO₂ and sulfate. While this work focused specifically on measurements of aerosol and , the techniques can be readily applied to measure ³⁵S in fog, rain, and snow. For example, given the production rates of ³⁵S (1.5 times that in troposphere) and the residence time of SO₄ in the stratosphere (∼2.5 years) (28), we estimate that the concentration of in the stratosphere, , is , or molecules per m^-3, based on the range of concentrations we report here. The enhanced sensitivities and techniques we present here could be used to measure in snow, and enhanced concentrations of may be used as a proxy for stratospheric intrusions in Antarctica and in samples from other remote field locations.

Because of the length of biological cycles, this technique offers a unique way to measure ³⁵S in fast growing plants, phytoplankton, and bacteria and therefore could provide a unique radioactive tracer to understand these processes.

Methods

Low-level liquid scintillation spectroscopy, combined with a high activity (aq.) standard, was used to develop the optimized ³⁵S sample preparation protocols. A multistage high-volume sampler was used for the aerosol and gas-phase S collections. Complete details can be found in Results and Discussion and in SI Text.