Organic-rich lake sediments were collected using a gravity corer in a pristine region in Northern Ontario, Canada, as part of a regional geochemical study by the Ontario Geological Survey [34] [35] . A subset of lakes that returned high concentrations of total Cr were chosen for subsequent re-sampling for Cr(VI) quantification [35] - [37] . These samples were taken from the top 5 to 6 cm at the sediment-water interface and stored in 180 mL amber glass jars. All samples were kept refrigerated until they were delivered to the analysis lab within 7 days after collection and stored there at 5˚C ± 3˚C. Sample extraction and analysis was carried out within 30 days of sample receipt. Hexavalent chromium in soil standards were acquired from the National Institute of Standards and Technology (SRM 2701, NIST, Gaithersburg, MD, USA), Phenova (QC-CR6-SOIL, Golden, CO, USA) and Fluka (SQC012, Sigma-Aldrich, Oakville, ON, Canada) and contained the certified Cr(VI) concentrations listed in Table 1.

Extractions of Cr(VI) from soil reference materials and lake sediments were carried out for 1 h at temperatures varying between 90˚C to 95˚C using different methods. The respective amounts of sample, reference material

^aConfidence interval with a probability of 95%; ^bAcceptance limit, corresponding to confidence interval with a probability of 99%.

and extraction solvent are listed in Table 2. All samples and enriched standards were accurately weighed with an analytical balance and 0.0001 g precision.

Based on U.S. EPA method 3060A [21] , reference materials and samples were digested in an alkaline procedure using a 0.28 M Na₂CO₃/0.5 M NaOH solution and heated to 90˚C ± 5˚C on a hotplate while stirring constantly. In addition, the approach described by Fabregat-Cabello et al. [24] was followed and a 0.05 M solution of ethylenediaminetetraacetic acid tetrasodium salt (EDTA, pH 10) was used to complex extracted Cr(III) and digest Cr(VI) in solid matrices. For this method, sediment samples or the SQC012 reference material were weighed, an appropriate amount of ⁵³Cr(VI) and ⁵⁰Cr(III) internal standards (Applied Isotope Technologies, Pittsburgh, PA, USA) added and the mixture suspended in the alkaline EDTA solution. The spiked standard solutions were diluted from stock solutions with certified concentrations of 10.4616 mg∙L^{−1 53}Cr(VI) and 10.8999 mg∙L^{−1 50}Cr(III).

Due to the large difference in Cr(VI) concentration between sample and reference material, the concentration of the enriched chromium isotope standards was varied accordingly to maintain a concentration ratio between analyte and standard of 0.1 to 10. Standard solutions were added in a 1/100 to 1/200 (v/v) ratio of standard/extraction volume at concentrations of 0.5 and 10 mg∙L⁻¹ for samples and reference materials, respectively. The obtained mixtures were then subjected to extraction in pre-cleaned 100 mL Teflon tubes using a Multiwave 3000 microwave system (Anton Paar, Graz, Austria) by ramping to 90˚C for 10 min and reaction for 1 h at 89.4˚C ± 0.8˚C (measured with IR sensor). In addition, a ThermoMixer^® C with Thermo Top (Eppendorf, Mississauga, ON, Canada) was used with a temperature setting of 95˚C ± 4˚C (accuracy based on manufacturer’s specifications) applying 5 mL vials (Eppendorf) and 500 rpm rotation to facilitate mixing. After cooling down to room temperature, all extracts were centrifuged at 4000 rpm for 20 min in a 5810 centrifuge (Eppendorf).

Prior to analysis, all extracts were diluted 10 to 5000-fold (v/v) with Milli-Q water (Millipore (Canada), Etobicoke, ON, Canada) and filtered using Mini-UniPrep™ syringeless filter devices having 0.2 μm polypropylene (PP) filter media and PP housings (Whatman, Forham Park, NJ, USA). Analysis was carried out using an ICS 5000⁺ ion chromatograph consisting of autosampler, 2 dual pumps, auxiliary pump (0.5 mL water∙min⁻¹ to flush the ion suppressor), eluent generator equipped with EluGen potassium hydroxide cartridge and ASRS-300 2 mm ion suppressor operated at 45 mA (all Thermo Scientific Dionex, Sunnyvale, CA, USA). Separation was achieved by flushing 200 μL of analyte through guard column (Dionex, AG20, 2 mm i.D. × 50 mm) and analytical separation column (Dionex, AS20, 2 mm i.D. × 250 mm) at 0.3 mL∙min⁻¹ using the following gradient: 0 - 6 min: 30 mM KOH, 8 - 13 min: 60 mM KOH, 16 - 20 min: 30 mM KOH. After passing through the ion suppressor at 0.3 mL∙min⁻¹, the eluate flow was combined with 0.2 mL∙min⁻¹ methanol (Thermo Fisher Scientific) using a Tee connection. The sample analysis was carried out on an API 3200 triple-quadrupole MS (SCIEX, Concord, ON, Canada) operated in negative ion mode using multiple reaction monitoring (MRM). The eluate was ionized by a TurboSpray electrospray source (SCIEX) with the following settings: curtain gas: 50.0, spray voltage: −3500 V, gas temperature: 450˚C, nebulizer gas 50, and turbo gas: 60. The MRM settings are displayed in Table 3. For data acquisition and handling, Analyst software version 1.6.1 and 1.6.2 (SCIEX) and DCMS Link version 2.12 (Dionex) were applied.

A classical linear multipoint calibration (MPC) quantification was carried out for selected samples. A 1000

^aUsed only during MPC quantification; ^bUsed only during SID quantification.

mg∙L⁻¹ CrO₄ standard (Sigma-Aldrich) was diluted to 9 concentrations between 0.010 to 1.0 μg∙L⁻¹ Cr(VI) and analyzed by IC-MS/MS along with alkaline-digested sediment and reference material extracts. Prior to IC-MS/MS analysis, the samples and calibration standards were spiked with an internal standard in a 50:1 (v/v) ratio containing 10 ng∙mL⁻¹ (≥98%, Sigma-Aldrich, Isotec Stable Isotopes) and 100 ng∙mL⁻¹ (98%, Cambridge Isotope Laboratories, Tewksbury, MA). Two quality control (QC) samples containing 0.800 and 0.200 μg∙L⁻¹ Cr(VI), respectively were analyzed with each batch of samples. The QC samples were prepared by diluting a 1000 mg/L CrO₄ standard (Inorganic Ventures, Christiansburg, Virginia, USA) with Milli-Q water. Sample concentrations were calculated relative to the peak areas of the internal standard. was monitored for quality control purposes.

After double spiking with isotope-enriched ⁵⁰Cr(III) and ⁵³Cr(VI), the ratios of ⁵²Cr/⁵³Cr(VI) and ⁵⁰Cr/⁵²Cr(VI) were monitored for [HCrO₄]⁻. In addition, transitions for [Cr^IIIEDTA]⁻ composed of isotopes 50, 52 and 53 were investigated. Each sample was analyzed in 4 replicates on the API 3200 triple-quadrupole system. Cr(VI) concentrations were calculated relative to the internal standard peak area of ⁵³Cr(VI) with a calculation tool supplied by Applied Isotope Technologies. The underlying mathematical equations have been published elsewhere [31] . The internal standard allowed for compensation of Cr(VI) to Cr(III) conversions occurring during extraction, as well as matrix suppression effects in the IC-MS/MS system.

Reference materials were also analyzed by colorimetry based on U.S. EPA method 7196A [27] . Extracts prepared by alkaline Na₂CO₃/NaOH digestion were pH adjusted to pH 7.5 ± 0.5 with nitric acid. Cr(VI) in the mixture was then complexed with 0.5% (w/v) 1,5-diphenylcarbazide (DPC) solution in acetone and the pH adjusted to pH 2 ± 0.5 using sulfuric acid. The absorbance of the solution was measured at a wavelength of 540 nm using a 2100 spectrophotometer (Cole-Parmer, Montréal, QC, Canada). The quantification of Cr(VI) concentrations was carried out based on a calibration curve composed of 5 standards in the range of 0.01 to 0.4 mg∙L⁻¹.

In order to determine the water content of the sediments, 10 g of sediment sample were dried for 16 h at 105˚C ± 5˚C. The mass difference between wet and dry weight was divided by the wet weight and defined as sample moisture content.

A novel quantification method of low-level Cr(VI) using IC-MS/MS was developed. Anion exchange chromatography allowed for the separation of the two chromium species from each other and from most interfering inorganic and organic sample components. Electrospray tandem mass spectrometry (ESI-MS/MS) permits the detection of Cr(VI) as [HCrO₄]⁻ → [CrO₃]⁻ in negative ion mode, but cannot analyze Cr³⁺ directly. By applying ESI-MS/MS, we thus compromised the ability to detect both species with an immense gain in sensitivity and selectivity for Cr(VI). The method detection limit (MDL) was calculated by analyzing a 0.050 μg∙L⁻¹ Cr(VI) standard multiple times and multiplying the resulting standard deviation of the determined Cr(VI) concentration with a one-tailed Student t-value (α = 0.01, n = 18) of 2.567. Using this approach, an MDL of 0.007 μg∙L⁻¹ Cr(VI) was estimated. Considering an extraction volume of 0.05 L, a sample amount of 2.5 g and a dilution factor of 100 to reduce chromatographic interferences with extraction solvent used for alkaline digestion, this value translates to 0.014 μg∙g⁻¹ wet weight Cr(VI) in solid matrices.

In a first attempt to develop a quantification method for Cr(VI) in lake sediments, three different reference materials were extracted in 15 replicates, respectively, using alkaline digestion. Due to the lack of sediment SRMs certified for Cr(VI), soil reference material was used. The extracts were analyzed by IC-MS/MS with a linear multipoint calibration curve (MPC) quantification method and compared with colorimetric measurements after formation of a DPC-Cr(VI) complex. The average Cr(VI) concentrations (n = 15) are listed in Table 1. The colorimetric approach generated values between 73% and 90% of the certified concentration, whereas the IC-MS/MS method resulted in recoveries of 95% and 111% for the Phenova and NIST reference materials, respectively. When looking at the individual concentrations shown in Figure 1, an excellent correlation between colorimetric and IC-MS/MS data was observed. This suggests that the extraction itself caused a large portion of the data variance for each set. Colorimetric data were typically 20% lower than Cr(VI) concentrations determined by IC-MS/MS, with both methods showing slight biases in opposite directions when compared with the certified concentration. However, for NIST 2701, one third of the samples were quantified within the 95% confidence interval of the certified concentration for both methods, respectively. For the Phenova standard, all samples were detected within the 99% confidence acceptance limits. For the colorimetric method, a bias towards

lower Cr(VI) concentrations can occur in the presence of reducing agents present in soils, such as Fe(II), sulfide or organic matter and is amplified by the acidic conditions required for the DPC-Cr(VI) complex formation [23] . A multi-laboratory study for NIST 2701 found even lower average Cr(VI) concentrations of 384.3 ± 73.5 µg∙g⁻¹ and only 70% recovery when analyzed by DPC-UV-VIS [38] .

As shown by Kingston and coworkers [33] , alkaline digestion can cause oxidation of soluble Cr(III) of up to 60%, which could not be accounted for using a traditional MPC approach with IC-MS/MS and resulted in overestimation of the Cr(VI) concentrations for one out of two reference materials in this study. The addition of Mg(II) ions and a phosphate buffer as recommended in U.S. EPA 3060A in the presence of oxidizing conditions was not applied for this data set in order to reduce interferences with the IC separation and MS/MS detection method. Additionally, matrix matching between samples and calibration standards proved to be difficult considering the complexity of each individual solid matrix. The internal standard counteracted matrix suppression/enhancement effects and elevated background signals to some degree, since data analysis without its consideration resulted in even higher Cr(VI) recoveries between 113% and 118%. However, the slightly stronger retention of the standard on the anion exchange column and different ionization efficiency during electrospray ionization made accurate compensation challenging. Both methods were also used to analyze 38 lake sediment samples collected in a pristine area in Northern Ontario. For the colorimetric approach, all sample results were below the method detection limit of 0.2 µg∙g⁻¹ Cr(VI). In addition, the dark-brown color of the extracts interfered with the analysis by UV-VIS. When applying the IC-MS/MS method to the environmental sample extracts, the results for 8 samples were below the MDL of 0.01 µg∙g⁻¹ wet weight Cr(VI).

In order to compensate for any oxidation and reduction processes altering the Cr(VI) concentration during the extraction, a speciated isotope dilution approach [24] [31] - [33] [39] was applied for the quantification of Cr(VI) in reference material and samples. An alkaline EDTA extraction [24] was chosen over the alkaline digestion based on U.S. EPA 3060A [21] for the SID-MS experiments for several reasons: Alkaline EDTA extractioncauses presumably less to no Cr(III) oxidation than U.S. EPA 3060A [24] [39] , reduces precipitation of Cr(III) and allows direct formation of the [Cr^IIIEDTA]⁻ complex [39] . Theoretically, conversion of Cr³⁺ to [Cr^IIIEDTA]⁻ improves not only its retention on the anion exchange column, but also allows indirect detection by ESI-MS/MS of the otherwise inaccessible ion. Since electrospray ionization suffers from severe ion suppression in the presence of ionic eluent compounds, the IC conditions used herein had to be modified from the elution parameters suggested by Fabregat-Cabello et al. [24] : The isocratic elution with 5 mmol∙L⁻¹ EDTA, pH 10 as mobile phase was replaced with a 30 to 60 mM hydroxide gradient. The Dionex system used in this study offers online post-column electrochemical removal of hydroxide ions in the mobile phase and is thus ideally suited for ESI-MS/MS analysis. The extraction time was kept at 1 h to facilitate comparison between samples analyzed by alkaline and EDTA extraction. The presence of 50 mmol∙L⁻¹ EDTA in the sample extracts required at least 1/10 dilution prior to IC-MS/MS analysis in order to retain the chromate ion on the anion exchange column.

EDTA extraction facilitated by microwave and ThermoMixer was applied to 36 and 47 lake sediment samples, respectively, as well as the SQC012 soil reference material. All extracts were analyzed by SID-IC-MS/MS. None of the analyzed lake sediment or SQC012 samples showed any formation of [CrEDTA]⁻ for the ⁵⁰Cr(III) standard or ^natCr(III). It has to be noted that the sensitivity of [CrEDTA]⁻ by ESI-MS/MS is significantly lower than for [HCrO₄]⁻ with a detection limit above 2 µg∙L⁻¹. Although the lack of a [CrEDTA]⁻ signal did not allow the simultaneous quantification of Cr(VI) and Cr(III), accurate Cr(VI) results were obtained for the certified reference material without consideration of any Cr(III) to Cr(VI) species interconversion. As pointed out before [24] , the complexation of Cr(III) with EDTA and the consequential absence of oxidation processes thus enables the practice of a single spike DID-MS procedure for Cr(VI) analysis in solid matrices. In addition, the detection of oxidized ⁵⁰Cr spike as [H⁵⁰CrO₄]⁻ allowed for a control tool to estimate the extent of potential oxidation.

Figure 2 contains typical chromatograms for the three chromium isotopes analyzed as chromate ions for a lake sediment sample and SQC012. The three chromatographic traces reflect the abundance of several species: ^natCr(VI), Cr(VI) spike enriched for ⁵³Cr, as well as potentially oxidized ^natCr(III) and oxidized Cr(III) spike enriched for ⁵⁰Cr. The main contribution for the [H⁵³CrO₄]⁻ and [H⁵²CrO₄]⁻ signals originated from the ⁵³Cr(VI) spike and natural ⁵²Cr(VI), respectively. In contrast, the relative distribution for the [H⁵⁰CrO₄]⁻ trace was heavily dependent on the presence of oxidation reactions. SID-IC-MS/MS analysis on all SQC012 extracts amounted in

Cr(VI) concentrations within the acceptance criteria with deviations between 0.2% to 18% from the certified concentration (see Table 1 and Figure 3(b)). The abundance of natural Cr(VI) was found to be very low in the environmental samples compared to the SQC012 standard. Intensities of ⁵³Cr(VI) spiked into sediment samples were typically significantly lower than in double-spiked method blanks. Furthermore, extraction recoveries of Cr(VI) peak areas in sediment samples spiked with natural Cr(VI) pre- and post-extraction ere low in some cases and ranged from 6% to 90%. These findings suggest that the high organic content of the analyzed sediment samples triggered analyte and spike reduction processes despite the high pH of the extractant.

Even with signal losses, Cr(VI) quantification after EDTA/microwave extraction yielded results above MDL for all the 36 analyzed samples. For 9 out of 47 samples digested with the ThermoMixer, no quantifiable signals for ⁵²Cr(VI) were detected after 60 min extraction. For 4 out of those 9 sediment samples, reducing the extraction time to 15 min allowed to increase the signal above MDL and thus quantify the Cr(VI) concentration in the sample. For SQC012, the extraction duration did not influence the intensity of any of the analyzed [HCrO₄]⁻ isotopes and the calculated Cr(VI) concentration when applying 95˚C for 15, 30, 45 and 60 min in the ThermoMixer (data not shown). For 5 out of 47 samples, no quantifiable signal for ⁵²Cr(VI) and ⁵³Cr(VI) could be detected despite a 4-fold increase in spike concentration and/or shorter extraction times of 15 min in additional test experiments.

In order to compare the efficiency of the two different extraction protocols for microwave and ThermoMixer directly, aliquots of 8 lake sediment samples were extracted in both heating systems and analyzed by SID-IC-MS/MS. The resulting Cr(VI) concentrations are shown in Figure 3. While the concentrations correspond very well for the selected samples, standard deviations were typically larger for samples extracted in the microwave, probably mostly due to lower absolute peak areas. Lower signals could be the result of a higher reduction rate of Cr(VI) in the microwave or result from different extractant-to-sample ratios. As shown in Table 2, the ratio between sediment sample and extraction solvent was increased from 1/20 (w/w) with microwave- assisted extraction to 1/9 (w/w) when using the ThermoMixer in order to accommodate the 10-fold reduced volume capacity. As a consequence, the total concentration of ^natCr(VI) in the extracts was increased by a factor of 2.2. Consequentially, the decreased EDTA volume did not reduce the relative amount of extracted Cr(VI), but potentially improved the precision of the calculated concentrations. It has to be emphasized that this experiment did not allow a direct comparison between the heating systems itself, but rather between the two different extraction protocols that were commonly applied in our laboratory.

For SQC012, the sample to extractant ratio was increased from 1/500 to 1/225 when switching from micro- wave to ThermoMixer extraction with no significant changes in accuracy of Cr(VI) quantification, as shown in Figure 3(b).

In order to estimate how representative a 0.5 g subsample was for the whole sample set, 6 sediment subsamples were taken from the same jar and extracted with EDTA. Analysis with SID-IC-MS/MS yielded a Cr(VI) concentration of 0.068 ± 0.006 µg∙g⁻¹ wet weight (N = 6, n = 24) with a coefficient of variation of 19%. Despite the difficulty of adequate mixing in a highly aqueous solid/liquid system, this experiment illustrated the adequacy of 0.5 g sample sizes for trace-level sediment samples. Single subsampling duplicates analyzed along with sample extracts resulted in average recoveries of 200% ± 58% (N = 14) for the ThermoMixer and 77% and 121% (N = 2) for the microwave extraction.

Figure 4 shows a direct comparison between one sample set digested with U.S. EPA method 3060A and analyzed using an MPC-MS approach and two sample sets processed with EDTA and an adapted SID-MS approach. All digestion/extraction and analysis methods yielded trace-level Cr(VI) concentrations. Samples analyzed without correction for Cr(VI) reduction with SID-MS yielded a lower average Cr(VI) amount in the analysis extracts and a lower maximum concentration. This finding suggests potential reduction processes even in the presence of the strongly alkaline Na₂CO₃/NaOH extractant. SID-MS results compensated for Cr(VI) reduction and thus provided typically higher average and maximum Cr(VI) concentrations. When comparing the wet weight data, the concentration distributions for ThermoMixer and microwave extracts aligned fairly well. However, after correction for the moisture content, the dry weight microwave data showed larger Cr(VI) concentrations than the ThermoMixer data. The moisture values have to be treated with caution since most samples contained over 90% water. Furthermore, the water content was determined in different sample aliquots than the

Cr(VI) analysis. A comparison between wet weight data seems thus more meaningful for sediments samples with high levels of water.