Article pubs.acs.org/ac
Application of Microwave-Induced Combustion and Isotope Dilution Strategies for Quantification of Sulfur in Coals via Sector-Field Inductively Coupled Plasma Mass Spectrometry Steven J. Christopher*,‡ and Thomas W. Vetter† ‡
NIST Material Measurement Laboratory, Chemical Sciences Division, Environmental Chemical Sciences Group, NIST Charleston Laboratory, 331 Fort Johnson Road, Charleston, South Carolina 29412, United States † NIST Material Measurement Laboratory, Chemical Sciences Division, Inorganic Measurement Science Group, MS 8391 Gaithersburg, Maryland 20899, United States S Supporting Information *
ABSTRACT: In recent years, microwave-induced combustion (MIC) has proved to be a robust sample preparation technique for difficult-to-digest samples containing high carbon content, especially for determination of halogens and sulfur. National Institute of Standards and Technology (NIST) has applied the MIC methodology in combination with isotope dilution analysis for sulfur determinations, representing the first-reported combination of this robust sample preparation methodology and high-accuracy quantification approach. Medium-resolution mode sector-field inductively coupled plasma mass spectrometry was invoked to avoid spectral interferences on the sulfur isotopes. The sample preparation and instrumental analysis scheme was used for the value assignment of total sulfur in Standard Reference Material (SRM) 2682c Subbituminous Coal (nominal mass fraction 0.5% sulfur). A description of the analytical procedures required is provided, along with metrological results, including an estimation of the overall method uncertainty (4500 at midmass (using medium-resolution slits), which was adequate to mitigate P−H+, S−H+, and O2+ interferences on the envelope of 32−34S isotopes.31 All isotope ratio data for the unknowns, controls, and procedural blanks were collected in the pulse counting detector mode. Table 2 lists the instrumental MS settings for the study, for both the IDICPMS measurements and corresponding isotopic abundance measurements for each of the coals tested. It was not possible to resolve the low abundance 36S+ isotope from the 36Ar+ interference. An estimated 36S isotope abundance of 0.015%32 was used for the purposes of determining S molar masses for the SRM materials tested. Spike Calibration and ID Analysis. Standard Reference Material 3154 Sulfur (S) Standard Solution (Lot No. 892205, Certified S mass fraction 10.30 ± 0.03 mg/g) was used to calibrate the 34S enriched spike solution via reverse ID. Diluted SRM 3154 (≈5075 mg/kg) and ≈2900 mg/kg 34S enriched spike (99.67% ± 0.2% enrichment, Medical Isotopes Inc., Pelham NH, USA) solutions were prepared in dilute acid (1% mass fraction HNO3 in water). The measured enrichment for the 34S after dissolution and background correction was 99.86%, within the specified enrichment. Four gravimetric blends of the natural S (≈90 mg) and enriched 34S (≈90 mg) spike solutions were produced and measured (target 32S/34S ratios near 1.6, corresponding to a nominal error magnification factor of 1). Spike calibration samples were run concurrently with the batch of analytical samples, as is typical for reverse IDMS measurements. A gravimetrically prepared nominal 75fold dilution of the spike solution was used to spike the procedural blanks (n = 6). All IDMS isotope ratio data were corrected for instrumental mass bias by determining mass bias correction “K” factors using theoretical and experimental 32 34 S/ S ratios of SRM 3154. The absolute 32S/34S ratio used for the T/E correction (32S/34S = 22.5667) is derived from published NIST data.33 The IDMS measurement function below was applied, with definitions for each variable provided in Supplemental Table 1:
of 50% (mass fraction) ammonium nitrate in water was pipetted onto the filter paper disc around the base of each pellet, taking care to not touch the pipet tip to the disc or pellet. Similar protocols were used to prepare procedural blanks (adding spike and ignition solution to the filter paper disc; however, the spike solution was diluted by a nominal factor of 75 for procedural blanks). Each prepared combustion platform was carefully lowered into a microwave vessel (containing a preloaded reflux solution at the bottom) using a quartz hook, so that the coal pellet was suspended above (≈1.5 cm) and separated from the microwave reflux solution (Figure 1). The reflux solution consisted of a mixture of 5 mL of 70% mass fraction HNO3 in water + 4 mL of 30% mass fraction H2O2 in water. The vessels were sealed inside the microwave oven rotor and pressurized to 2 × 106 Pa with O2 through the vent valve located in the vessel lid. This pressurization involved opening a vent screw to fill each vessel with O2 and then closing the vent screw to seal in the combustion gas. The combustion apparatus was used with the Multiwave 3000 microwave digestion system manufactured by Anton Paar (Richmond, VA, USA), and the method was created using the software-selectable “combustion” mode of operation. The microwave method was programmed as follows: (1) 1400 W for 1.5 min to initiate combustion and (2) 60 min hold at 1400 W. After combustion, each cooled microwave digest was diluted with water to near the top of the microwave vessel (80 mL mark), to fully wet the vessel walls and cover the entire quartz combustion platform. The samples were vortexed for 60 s and allowed to settle prior to preparation for instrumental analysis. Instrumental Method. Portions of the microwave digests were subsampled and diluted further with water so that the signal intensity of S would fall within the dynamic range of the ICPMS detector. The analytical samples were measured in medium-resolution mode on a Thermo Fisher Scientific (Bremen, Germany) Element XR ICPMS system. The ICPMS was tuned for maximum sensitivity and low oxides using a 1 ng/g Li, In, U solution. The resolution (M/ΔM) was a x × cz × cx =
my mx
×
mz m y′
{
K y1 × R y1 − Kb × R b Kb × R b − K x1 × R x1
}×{
Kb′ × R b′ − K z1 × R z1 K y1 × R y1 − Kb′ × R b′
dx
where i is summed from 1 to 4, to account for the four natural S isotopes. A similar version of the measurement function has been derived;34 however, the NIST measurement function also accounts for test portion dry mass correction (moisture content) and adds several variables (as constants) to incorporate sources of uncertainty related to replication, detector dead time correction, and instrument background correction. This version of the measurement function is not
}×
∑i K xi × R xi ∑i K zi × R zi
− az × cblk × C Rep × Cdt × C bgd
convenient for how IDMS is actually carried out for calculation of mass fractions, but it is presented in this manner to later consider all the potential sources of uncertainty one might encounter measuring sulfur by ID-MIC-SF-ICPMS. A Kragten spreadsheet approach35,36 and the IDMS measurement function was used to estimate the expanded uncertainty for SRM 2682c in determining S by ID-MIC-SF-ICPMS. In a GUM-complaint manner,37 the Kragten spreadsheet numerically approximates C
DOI: 10.1021/acs.analchem.5b03981 Anal. Chem. XXXX, XXX, XXX−XXX
Analytical Chemistry
Article
the results obtained when using propagation of uncertainty to calculate a combined standard uncertainty from the input of the individual components of the measurement function and their associated standard uncertainties. Dry Mass Correction Factors. Separate coal test portions (not destined for combustion) were corrected for mass loss on drying to establish individual mass correction factors (ratio of dry mass to wet mass) for each SRM bottle sampled. Approximate 0.75 g test portions were dried in a gravity convection oven for 2 h (set point = 105 °C). The dry mass correction factors achieved using this method are similar to those achieved by other standard methods.38 Dried test portions were stored in glass weighing containers in a vacuum desiccator containing calcium sulfate to facilitate sample cooling in a dry atmosphere prior to weighing. Five within-bottle samples (≈0.5 g) from a single bottle of SRM 2682c were dried using the same timing and oven set points, in order to establish a moisture measurement uncertainty for SRM 2682c and for controls SRM 2682b and SRM 2682a. Similarly, triplicate (≈0.75 g) test portions from a single unit of SRM 2693 were also measured for the same purpose. It should be pointed out that the mass correction factors are significant for the coals tested in this study, and the uncertainty due to mass correction is likely constrained by the adoption of a specific drying procedure that generates reproducible samples.
Figure 2. Postcombustion picture of high pyritic sulfur (1 wt %) SRM 2685b samples digested in either acidic (HNO3 + H2O2 in water) or basic (50 mmol/L NH4CO3 in water) reflux media. Iron carbonate and iron hydroxide precipitation appears as red-colored particles near the bottom of the combustion tubes labeled base.
(SRM 2685b, Table 1) showed that, if the spike was initially placed in the reflux solution, the time needed to achieve sample−spike equilibrium was extended, with the working hypothesis being that the 34S spike in solution partitioned to residual ash particles generated, before slowly leaching back into solution. This process is depicted in Figure 3 as a change in the S mass fraction of SRM 2685b (directly correlated to a change in the measured 32S/34S ratio) over time, obtained by measuring discrete sampling draws from combusted test portion digests, where the spike was placed either in the reflux solution (n = 3 replicates) or on the combustion platform (n = 3 replicates) prior to combustion. The six combustion digests contained visible residual ash particles, which is consistent with the high residual ash content of the SRM 2685b material. The volume removed for each sample draw/time point was approximately 0.25% of the respective original solution digest volume (0.2 mL/draw from 80 mL total digest volume); the measured S mass fraction should have remained relatively constant upon achievement of sample−spike equilibrium in the biphasic (ash-containing) solution digests. This relatively constant value was only true for the combusted spike samples. The S mass fraction was markedly higher than expected for the Day 1 samples (Figure 3) where the spike was initially present in the reflux solution, followed by a change in S mass fraction toward the expected values over the course of several days. This initial high S mass fraction bias was not observed in subsequent work (Figure 3) when the spike was placed on the combustion platform, as indicated by the flat response of S mass fraction data versus time. The terminal S mass fraction values measured for SRM 2685b (Figure 3, Day 7 results) indicated an approximate 2% bias relative to the certified value, regardless of spike placement. The SRM 2685b certified value24 is based on Carius tube sample dissolution and the isotope dilution thermal ionization mass spectrometry (ID-TIMS) value assignment in 2007. A limited amount of quantitative in-house (NIST) control data collected in 2009, based on the same ID-TIMS certification methodology, indicated a S mass fraction value of 4.58% ± 0.07% in SRM 2685b (≈3% bias). The ID-MIC-SF-ICPMS data fall in between the certified values and the 2009 control data. Qualitative data collected in 2012 from comparative laser ablation SF-ICPMS of pressed pellets of SRM 2685b and the companion candidate SRM 2685c material also suggested that the S mass fraction of SRM 2685b could be shifting to a lower S mass fraction value. Nonetheless, a low bias in the ID-MIC-SF-
■
RESULTS AND DISCUSSION Method Development Details for Application of IDMIC-SF-ICPMS. Initial combustion temperatures exceeding 1400 °C and pressures of 5.6 × 106 Pa have been reported for coal samples ranging from 0.1 to 0.5 g using a specially instrumented microwave.39 Initial combustion temperatures and pressures could not be measured using the stock microwave system. Maximum pressures of 8 × 106 Pa and temperatures exceeding 205 °C were achieved during the reflux that occurred through the hold period. Temperatures of >180 °C were reached after approximately 11 min, and terminal temperatures of >205 °C were reached after approximately 30 min. Chemical conversion of sulfur species occurs during the combustion and reflux processes where intermediate forms of sulfur in the gas or liquid phase are oxidized to sulfate in solution. Water, hydrogen peroxide, acidic, and basic reflux media have all been discussed in the literature as appropriate reflux media for sulfur capture.13,14,40 Initial attempts to combust coals using peracetic acid and base piranha (ammonium hydroxide/hydrogen peroxide) reflux solutions and also 50 mmol/L ammonium carbonate and ammonium hydroxide reflux solutions resulted in formation of precipitates in the microwave combustion tubes (likely iron hydroxides and iron carbonates). A picture comparing coal combustion digests in either acidic or basic reflux media is provided in Figure 2 for SRM 2685b, a bituminous coal with pyrite (FeS2) content ≈1 wt %. A mixture of HNO3/H2O2 for reflux capture of sulfur was required to avoid formation of precipitates (Figure 2, compare vessels labeled acid with vessels labeled base). The use of HNO3/H2O2 reflux media is suitable for downstream ICPMS analyses and serves to drive the oxidation of multiple chemical forms of sulfur in coals to sulfate. Sulfate is the chemical form of S that matches the speciation of the enriched S spike used in this work (34SO42−), which helps ensure isotopic equilibration of sample and spike, a prerequisite for ID analysis to work. Initial Experiment. Early method development trials using a particularly refractory coal of high sulfur and ash content D
DOI: 10.1021/acs.analchem.5b03981 Anal. Chem. XXXX, XXX, XXX−XXX
Analytical Chemistry
Article
Figure 3. Comparison of sulfur mass fraction (%) data for replicate SRM 2685b test portions subjected to ID-MIC-SF-ICPMS using the scheme and reagents in Figure 1, with either 34S spike added to the reflux capture solution (triangles) or 34S spike collocated and combusted with the test portion pellets (circles). The certified S mean and expanded uncertainty for SRM 2685b are shown as solid and dashed black lines, respectively. The measured S mean (based on all days) and expanded uncertainty for SRM 2685b combusted spike samples are shown as solid and dashed blue lines, respectively.
Figure 4. Sulfur signal intensity (integrated counts per second, ICPS) and isotope ratio results obtained for SRM 2685b and SRM 2692c from the temporal isotope equilibration experiment.
hypothesis that the S mass fraction data in Figure 3 could be biased high (initially) because of a delayed onset of sample− spike equilibrium. The 32S/34S IDMS ratios were prepared to fall near 1 in this experiment but were biased high immediately after combustion (Figure 4, day 0 results) and settled near the expected ratios after a day of equilibration (Figure 4, day 1 results). Intensity values for both 32S and 34S signals increased >1 order of magnitude between the day of combustion and 1 day following, suggesting that both sample- and spike-derived sulfur temporarily partitioned to ash particles after combustion. The results presented in Figure 4 strengthen the hypothesis of delayed sample−spike equilibrium when the 34S spike is placed
IDMS method could not be ruled out for this specific material, given the paucity of data. A low sulfur content could be due to incomplete combustion, possibly due to the high pyrite content in SRM 2685b, or incomplete sample−spike equilibration, possibly due to preferential loss of sample sulfur in the vapor phase, or an instrumental measurement problem. Temporal Isotope Equilibration Experiment. Sulfur IDMS intensity and isotope ratio data were generated from a second, independent temporal sampling experiment combusting duplicate test portions of SRM 2685b, and one test portion of SRM 2692c (lower FeS2 and ash content), each containing the 34S spike in the reflux solution, to further test the E
DOI: 10.1021/acs.analchem.5b03981 Anal. Chem. XXXX, XXX, XXX−XXX
Analytical Chemistry
Article
spike may be less uniformly deposited and take longer to absorb or dry. Spiking into the reflux solution residing directly in the microwave vessel is less restrictive in terms of physical volume limitations. This last ID spiking approach might still find application with easier-to-combust samples that do not produce residual ash or for other elements besides S. However, this approach should be considered carefully, based on the sample matrix and chemistry of the target analyte. ID-MIC-SF-ICPMS Data for SRM 2682c and Controls. The most recent application of the ID-MIC-SF-ICPMS methodology using “on platform” spiking is presented next, considering data for the series of 2682 SRMs and SRM 2693 (Table 1). All signals for unknowns, controls, procedural blanks, and isotopic abundance samples were corrected for instrument S background, detector dead time, and mass discrimination/bias to establish the S isotope ratios used in ID quantification or isotopic abundance calculations. A 32S/34S Theoretical/Experimental (T/E) isotope ratio was used to calculate a mass bias “K” factor correction of 1.09 (≈4.5% per u) for samples of SRM 3154 run in medium-resolution mode. No assumption of a mass bias model was considered. This factor was used to adjust the measured 32S/34S isotope ratios for the blend and back-blend ID samples. A spike calibration solution was measured at the beginning of the analysis (just after SRM 3154) and measured several times over the course of a 9 h instrumental analysis to monitor isotope ratio repeatability. The relative standard deviation (RSD) due to isotope ratio repeatability was approximately 0.6% for n = 5 repeat measurements of 32S/34S and was used in the estimate of the isotope ratio measurement uncertainty for IDMS blend and back-blend samples. No systematic drift in the isotope ratio was observed, suggesting that the K factors applied were stable for the experiment. Isotope abundances were measured for each coal SRM tested and the 34S spike after dissolution by monitoring the count rates for 32,33,34S and correcting for corresponding S background count rates. Count rates for 36S were estimated from corrected 32S count rates, assuming IUPAC abundances32 of 95.04% for 32S and 0.015% for 36S. Table 3 lists the measured S abundances and corresponding molar masses calculated for the materials. Blank, Mass Fraction, and Moisture Measurements. The absolute S procedural blanks ranged from 52 to 140 ng for the analysis of SRM 2682c, with a mean blank of 89 ±30 ng (1 SD) for n = 6 blanks. Blank corrections were minor (
