From sea salt to seawater: a novel approach for the production of water CRMs﻿

Natural water certified reference materials (CRMs) are mostly available in a liquid form, and they are produced starting from suitable environmental samples. Many precautions are usually needed to avoid biological or physical degradation, including filtration, acidification, and sterilization. In this study, the drawbacks associated with liquid-based seawater CRMs were tackled by developing a salt-based seawater proxy for nutrients that could be reconstituted in water solution just before use. Phosphate, silicate, and nitrate were chosen as target analytes. Sea salt mimicking the composition of seawater was spiked with an aqueous solution of the analytes and homogenized using a high-energy planetary ball mill (uhom < 1.2%). The salt powder CRM SALT-1 (https://doi.org/10.4224/crm.2022.salt-1) demonstrated good short- and long-term stability for nutrients. When the SALT-1 was reconstituted in water at the 4.0% w/w level, the resulting solution had similar properties with respect to typical seawater in terms of major constituents (± 20%), trace metals, density (1.023 g/mL), pH (8.8–9.0), and optical properties relevant to the photometric characterization. Phosphate and silicate were quantified by photometry (molybdenum blue method, batch mode), whereas nitrate was quantified by isotope dilution GC−MS (uchar < 1.2%). In the SALT-1 reconstituted seawater solution at the 4.0% w/w salt level, the nutrient amount concentration was w(phosphate, PO43−) = 1.615 ± 0.030 μmol/L, w(silicate as SiO2) = 8.89 ± 0.31 μmol/L, and w(nitrate, NO3−) = 18.98 ± 0.45 μmol/L at the 95% confidence (k = 2). Overall, the SALT-1 CRM exhibits similar nutrient profile and general analytical characteristics as the MOOS-3 CRM. However, the SALT-1 has much reduced preparation, storage, and distribution cost, likely much better long-term stability, and it could enable the production of lower cost and more accessible seawater reference materials. Graphical abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00216-022-04098-0.

The determination of nitrite was obtained by UV-vis spectroscopy using the Griess method adapted from Ref. [1,2]. Two reagents were required for the derivatization: a. Reagent NO2-R1. A 5.0 mL volume of concentrated (37%) hydrochloric acid was added to 30 mL of water. The solution was gently mix and allowed to cool down. 0.5 g of sulfanilamide was added and dissolved. The solution was further diluted with water to a final mass of 50.0 g. This solution was stored in a polyethylene bottle at 4 °C for one month.
b. Reagent NO2-R2. 0.1 g of N-(1-naphthyl)-ethylenediamine dihydrochloride was dissolved into 100 mL of water. This solution was stored in a polyethylene bottle at 4 °C for one month.
For color development, a 5 mL volume of the sample (or standard) was mixed with 0.1 mL of NO2-R1. After 5 min, 0.1 mL of NO2-R2 solution was added. The absorbance of the solution was read after 15 min (but within 2 h) at 541 nm. The derivatization was performed in plastic bottles. The absorbance of the derivatized sample was read at 541 nm (5 s integration time using 1 nm spectral bandwidth). The absorbance was measured using a quartz cuvette (1 cm).

S1.2 Photometric determination of phosphate
The determination of phosphate was obtained by UV-vis spectroscopy using the molybdenum blue method adapted from Ref. [2][3][4][5][6]. The reagents required for the derivatization were prepared as follow: For color development, a 6 mL volume of the sample (or standard) was mixed with 0.5 mL of PO4-MIX. Then, 0.2 mL of PO4-AA solution was added. The derivatization was performed in plastic bottles. After 30 min, the absorbance of the solution was read at 890 nm (5 s integration time using 1 nm spectral bandwidth). The absorbance was measured using a quartz cuvette (5 cm).
The residual absorbance of the matrix was also subtracted from the signal reading [3]. For this purpose, a 6 mL volume of the sample (or standard) was mixed with 0.5 mL of PO4-Acid. Then, 0.2 mL of PO4-AA solution was added and the absorbance of the solution was read at 890 nm.
For quantitation, five standards and a reagent blank were prepared in low nutrient seawater. Both linear and quadratic calibration curves were obtained. Uncertainty was evaluated using both Gauss [7] and Monte Carlo approach [8] (Paragraphs S1.7-S1.8 and attached excel file).

S1.3 Photometric determination of silicate
The determination of reactive silicate was obtained by UV-vis spectroscopy using the molybdenum blue method adapted from Ref. [2,9]. The reagents required for the derivatization were prepared as follow: For color development, a 3 mL volume of the sample (or standard) was mixed with 0.12 mL of Si-R2. After 10 min, 0.12 mL of Si-R3 and 0.06 mL of Si-AA were added. The solution was mixed after every addition. Between addition of Si-R3 and Si-AA there was no waiting time. The derivatization was performed in plastic bottles. After 60 min, the absorbance of the solution was read at 810 nm (5 s integration time using 1 nm spectral bandwidth). The absorbance was measured using a quartz cuvette (1 cm).
The residual absorbance of the matrix was also subtracted from the signal reading [3]. For this purpose, a 3 mL volume of the sample (or standard) was mixed with 0.12 mL of Si-Acid. After 10 min, 0.12 mL of Si-R3 and 0.06 mL of Si-AA were added. The solution was mixed after every addition. Between addition of Si-R3 and Si-AA there was no waiting time. After 60 min, the absorbance of the solution was read at 810 nm.
For quantitation, five standards and a reagent blank were prepared in low nutrient seawater. Both linear and quadratic calibration curves were obtained. Uncertainty was evaluated using both Gauss [7] and Monte Carlo approach [8] (Paragraphs S1.7 and S1.8 and attached excel file).

S1.4 GC−MS determination of nitrate
The determination of nitrate was performed by headspace GC−MS using triethyloxonium tetrafluoroborate to convert NO3 − into a volatile derivative [10][11][12][13][14][15]: This derivatization is a single-step aqueous derivatization. A 2.0 mL aliquot of seawater sample was transferred in a 10 mL headspace vial along with 0.2 mL of 15 NO3 − solution of internal standard. The amount of internal standard was chosen to match the concentration of nitrate in the sample. The sample preparation was performed gravimetrically. At this point, 0.1 mL of aqueous sulfamic acid (1% w/w) was added to the mix to remove nitrite. This blend was then reacted with 50 µL of Et3OBF4 aqueous solution. Aqueous solutions of Et3OBF4 are unstable and are prepared just before derivatization. For this purpose, 1.0 g of Et3OBF4 (commercial reagent) was dissolved in 1.0 g of cold water (4 °C) and used within 5 min. This blend was then analyzed by headspace GC−MS. The vial was incubated at 60 °C for 2 min with agitation. A 500 µL volume of headspace was sampled (gas-tight syringe held at 70 °C) and injected in the gas chromatograph. The syringe was cleaned by flushing it with nitrogen for 5 min. The inlet liner (internal diameter of 1 mm) was held at 120 °C and the injection was performed in 7:1 split mode. Separation was obtained on a mid-polarity column (DB-624; length: 30 m; stationary phase: 6%-cyanopropyl-phenyl-94%-dimethyl polysiloxane; 0.25 mm inner diameter; 1.40 µm coating) using the following temperature program: 1.5 min (at 50 °C), then 20 °C/min to 140 °C for a total run time of 6 min (constant flow mode: 1.0 mL He/min). The temperature of the transfer line was 220 °C. Ethyl nitrate was detected in negative chemical ionization mode (CH4 reaction gas; CI gas flow: 40%; source and quadrupole temperature: 150 °C). The acquisition was obtained in selected ion monitoring at m/z 46 and 47 (50 ms dwell time for each ion). The GC−MS chromatogram of the ethyl nitrate derivative is shown in Fig. S13. The integration of the signal was obtained using the Agilent MassHunter software (B.06.00, Built 6.0.633.0, Agile integrator, 2012). Nitrate quantitation was obtained using quadruple isotope dilution [13]. Calculation and uncertainty evaluation can be found in Paragraph S1.9 and in the attached excel file.

S1.5 ICP−MS/MS for phosphate and silicate
An Agilent 1200 series HPLC (Agilent Technologies, Mississauga, Ontario) with an ion exclusion column Dionex IonPac ICE-AS1 (4 x 250 mm) was used to separate phosphate and silicate from the major ions in sea water samples. A PEEK tubing (0.13 mm ID x 1.59 mm OD) was used to connect the column of the HPLC to the ICP−MS/MS introduction system. The sample uptake was 25 µL and the eluent used in the experiment was HCl 0.01% and the flow rate was 0.25 mL/min. All measurement were made with an Agilent 8800 (Agilent Technologies, Mississauga, Ontario) operating in hydrogen mode to measured mass 28 (Si) and 31 (P) on an 8 min acquisition window. A Ga 20 ng/g internal standard solution was simultaneously introduced to monitor the stability of the signal. Hydrochloric acid used for the eluent was purified in-house and the deionized water was from a Thermo Scientific Gen-Pure UV xCAD plus system (18.2 MΩ cm at 25 °C). The standards SRM 3150 silicon solution (lot number: 130912) was from the National Institute of Standards and Technology (NIST, Gaithersburg, MD) and the Supelco phosphate standard 38364 (lot number: BCCD8182) was from Sigma Aldrich (Oakville, ON, Canada). These primary standards were used for the preparation of the standard addition calibration curves. Three different samples of SALT-1 were prepared by dissolving approximatively 4.0 g of sample into 100 g of water. Freshly prepared samples were allowed to rest for 24 hours at room temperature. There aliquots from each samples were prepared and spiked respectively to achieve the standard addition calibration curves (one un-spiked sample and two spiked samples to double and triple the signal of the sample). The blank used was DIW and was also spiked to match the signal observed in the SALT-1. S1.6 Ion chromatography conductivity for major components Chloride, bromide, and sulfate were detected by ion chromatography with conductivity detection. A Thermo Scientific ion chromatography system was used. A single pump ICS-5000 + SP-5 was interfaced with an AS-AP autosampler (p/n 074925

S1.8 ICP−MS for trace metals
A seaFAST from Elemental Scientific (ESI, Omaha, NE) was used as an automated preconcentration system to separate analytes of interest in undiluted seawater samples. The column used in the seaFAST was the CF-N-0200 and it was coupled to the Thermo ELEMENT XR (Bremen, Germany) sector field ICP−MS. About 6.0 mL of sample was injected by the seaFAST autosampler. The analytes of interest were concentrated on the column and then eluted to the ICP−MS. The analytes measured were Cd, Co, Cu, Fe, Mn, Ni, Pb, V and Zn. Some analytes like Mo and U did not need pre-concentration and were measured after simple dilution. For Cr measurement, samples were reduced using hydroxylamine hydrochloride solution, then separated and pre-concentrated offline using a column of silica immobilized diphenylcarbazone. Samples were prepared by diluting approximatively 4.0 g of SALT-1 in 100 g of a solution of nitric acid at pH of 1.6. A three points standard addition calibration was used for As, Cd, Co, Cu, Fe, Mn, Pb, Ni, V and Zn. Isotope dilution calibration was used for Cr, Mo and U. Deionized water (DIW) was obtain from a Nanopure ion exchange reverse osmosis system (Barnstead/Thermolyne, Bosta, MA). Nitric acid was purified in-house using a sub-boiling distillation system. The standard stock solutions were prepared from high-purity metals with purity determined by the NRC GD−MS facility. A SRM 3164 uranium stock solution from National Institute of Standards and Technology (NIST, Gaithersburg, MD) was used. Seawater CRMs NASS-7 and CASS-6 from National Research Council (NRC, Ottawa, Canada) were used as quality control. S1.9 Determination of phosphate and silicate by photometry: uncertainty evaluation by propagation (combined standard uncertainty) Both phosphate and silicate were measured by photometry following quantitation against an external calibration curve (5-points) obtained in matrix-matched low nutrient seawater. For the estimation of the combined standard uncertainty on the final results we followed the JCGM 100 guide [7]. For a general model equation = ( , , ⋯ , ) the combined variance on is given by this equation: In the case of external calibration with linear model ( = + • ), the only correlated variables are the intercept and slope of the calibration curve. Their correlation coefficient is estimated as follow [16]: The combined standard uncertainty on the result was calculate using an Excel VBA function, as reported in the attached Excel file. VBA code of the function: x = (y -a0) / a1 GaussU = Sqr((1 / a1) ^ 2 * (uy ^ 2 + ua0 ^ 2) + ((a0 -y) / a1 ^ 2) ^ 2 * ua1 ^ 2 _ -2 * (1 / a1) * ((a0 -y) / a1 ^ 2) * ua0 * ua1 * r) RSD = Sqr((GaussU / x) ^ 2 + (uw / w) ^ 2) GaussLinearUNC = x * RSD End Function S1.10 Determination of phosphate and silicate by photometry: uncertainty evaluation by Monte Carlo simulation Both phosphate and silicate were measured by photometry following quantitation against an external calibration curve (5-points) obtained in matrix-matched low nutrient seawater. The uncertainty on the final result was also estimated using a Monte Carlo method. Uncertainty on the x-axis (mass fraction) was evaluated by uncertainty propagation on the gravimetrically prepared standards. Uncertainty on the y-axis (normalized signals) was estimated from the ANOVA study for homogeneity: the within-group variance was used as an estimate for the uncertainty on y-axis.
The Monte Carlo uncertainty on the result was calculated using an Excel VBA function, as reported in the attached Excel file. VBA code of the function:  Quantitation of nitrate was obtained by exact-matching quadruple isotope dilution as reported previously [13,14]. Value assignment and uncertainty evaluation (combined standard uncertainty by propagation [7]) were obtained by applying the following Excel VBA function:  Figure S4. Absorbance of SALT-1 matrix as a function of pH When the SALT-1 is solubilized in water (4.0% w/w) the pH of the solution is ~8.9 with a minor opalescence which cause a positive reading of absorbance. The absorbance reported in the figures was measure using a 5 cm quartz cuvette. When the pH of the SALT-1 was corrected to lower values (8.2, 7.6, 7.2), the opalescence disappeared and the residual absorbance of the solution was comparable to that of the low nutrient seawater from OSIL. The determination of phosphate, silicate, and nitrate by photometry requires derivatization under acidic conditions [2]. In this experiment, a SALT-1 sample was solubilized in water at the 4.0% w/w level (pH 8.9). Another unit of SALT-1 was baked at 105 °C for 21 days. The reconstituted solution of this sample showed a pH of 10.5 and it was quite turbid. The third (reference) sample was the low nutrient seawater from OSIL. Each of these three samples was acidified under typical conditions for determination of phosphate, silicate, and nitrite (Paragraphs S 1.1 to S 1.3 "residual absorbance of the matrix"). The absorbance reported in the figures was measured using a 5 cm quartz cuvette. In the top graph the three samples were prepared under conditions for phosphate measurement; the middle one for silicate; the bottom one for nitrite. The photometric response of phosphate, silicate and nitrite was studied in the [0.1, 1] absorbance interval. 10 standard solutions were prepared in low nutrient seawater and the response was measured as described in Paragraphs S1.1 to S 1.3. Although the calibration plots are "visually" linear with R 2 > 0.997, the residual graphs shows a nonlinear response.   Figure S10. Rotational matrix effects: SALT-1 vs low nutrient seawater One SALT-1 sample was reconstituted in water at the 4.0% w/w level. One nutrient solution was prepared in low nutrient seawater to closely match SALT-1 concentration for phosphate and silicate. Both materials were analyzed as-is and after increasing levels of phosphate and silicate. The ratio between the slopes of the calibration curve obtained in the SALT-1 medium was not significantly different than the one obtained in low nutrient seawater for both analytes. No relevant (rotational) matrix effects were observed. Therefore, the matrix-matching external calibration was used for quantitation.    Spacing: number of units produced before the sampling for homogeneity study. Bottle ID: before the final bottling, the SALT-1 was portioned in 32 x125 mL Nalgene bottles: the bottle ID column keeps track of the bottle number where the corresponding SALT-1 units was coming from. Δ Time (min): this is the time from when the Bottle ID was opened to when the corresponding SALT-1 unit was produced (monitoring the effect of ambient humidity on the homogeneity). Day: this column keeps track of the day of bottling (the SALT-1 was bottled over four days).