The objective of this study was the preparation of a salt-based material which could be reconstituted in water and used as a quality control test sample for the determination of nutrients in seawater. In the following paragraphs, the preparation and the certification of the SALT-1 CRM are described. Besides the assessments required for certification (i.e., homogeneity, stability, and determination of the property values), attention was given to the characterization of the most relevant properties of the 4.0% w/w SALT-1 aqueous solution, such as density, pH, UV–vis spectra, major components, and trace metals to ensure that the material was an appropriate proxy for natural seawater.
Preparation of the SALT-1 CRM
A combined nutrient spike solution was prepared in water at these levels of mass fraction: w(NO3−) = 1.15%, w(Si) = 0.217%, w(PO43−) = 0.145%. A 1.5-mL aliquot (1.54 g) of this solution was mixed with 600 g of marine salt in a planetary ball mill. Specifically, 1.5 mL nutrient solution and 600 g marine salt were mixed in a 1-L zirconia jar containing milling zirconia balls (⌀5 mm = 940; ⌀10 mm = 154; ⌀20 mm = 3). The material was milled using a 90-min program where the ball mill was operated at 42 Hz in bidirectional mode: 10 min on and 10 min off. Before opening, the jars were left to cool for 40 min. In a milling cycle, two 1-L jars were employed. The final product (~ 1 kg) was sieved into a 4-L PE bottle, and further homogenized for 15 min using an elliptical shaker. For the production of SALT-1, the process was repeated 4 times yielding ~ 4 kg of product. For further homogenization, the material was transferred into a 20-L PE carboy and shaken manually for 5 min; then, it was portioned back in 4 × 4-L plastic bottles and shaken again for 30 min on the elliptical shaker. The final salt mix was then dispensed into 32 × 125-mL plastic bottles (~ 115 g each) which were thermosealed in trilaminate bags. The material was finally aliquoted into 850 × 12-mL PE bottles (Fig. 1). The target mass of SALT-1 in each unit was 4.1 g, and the mass of each unit was recorded. The mass distribution of the SALT-1 lot is reported in Fig. S1.
Hygroscopicity of the SALT-1
The manipulation of this large surface area (potentially hygroscopic) material posed some initial moisture adsorption concerns. Therefore, its atmospheric moisture adsorption was studied. 3.5 g of SALT-1 was spread on a Petri dish and the mass increase due to moisture adsorption was < 1.5% over 1-h time (t = 20 °C, RH = 40–60%, Fig. S2). Since the preparation of each individual SALT-1 unit took ~ 2.1 min (Fig. S3), all manipulations were carried out under normal laboratory conditions and the finished material was kept tightly closed in PE bottles sealed into trilaminated pouches. Under these conditions, the long-term storage was also favorable. Over a period of 3 years, the average mass increase for the SALT-1 units was only 0.28% (n = 49, RSD = 38%).
Reconstitution: optical properties and pH
The SALT-1 was reconstituted in water to a final mass fraction of 4.0% w/w. This composition was chosen to match average seawater matrix with a density of 1.023 g/mL. Although SALT-1 solubilized instantaneously, the reconstituted material showed a slight opalescence barely noticeable by naked eye. While studying a possible explanation for this effect, it was noticed that the pH of the SALT-1 solution (pH 8.8–9.0) was at the high end of the ocean water pH interval . Moreover, the pH of the material was influenced by the storage temperature: when the SALT-1 was baked at 105 °C for 21 days, the pH of its 4.0% w/w solution was 10.5 and it looked quite turbid. Since sodium bicarbonate is a likely component of the artificial sea salt used to prepare the SALT-1, it was hypothesized that the “loss of acidity” caused by heating could be due to thermal degradation of bicarbonates :
Even milder conditions could cause a slight pH increase: a SALT-1 unit left at 40 °C for 21 days generated a solution with a pH of 9.1.
The alkaline excess could be corrected by adding acid to the SALT-1 solution. Figure 2 reports the pH variations by adding 50-μL aliquots of 0.1 M HCl into the reconstituted solution of SALT-1 (i.e., ~ 4.1 g of SALT-1 in 100 mL volume of solution). When 1.5 mL of 0.1 M HCl was used to correct acidity, the pH of the resulting solution was between 7.7 and 8.0 (a pH between 7.2 and 7.3 could be obtained by adding 2.4 mL of 0.1 M HCl).
The slight opalescence of the SALT-1 solution was related to pH and likely depended upon the complex aqueous equilibria that regulates solubility and precipitation of Mg(OH)2 and CaCO3 . When the SALT-1 solution was buffered to pH values lower than 8.2, the opalescence disappears and the residual absorbance of the matrix was comparable to the values obtained in low-nutrient seawater (Fig. S4). Furthermore, under acidic conditions typically required for the determination of phosphate, silicate, and nitrate by photometry , the residual absorption of the matrix was comparable to low-nutrient seawater, even when the SALT-1 was baked at 105 °C (Fig. S5).
Reconstitution: nutrient solubilization
The effect of pH adjustment and time on the nutrient composition of 4.0% w/w SALT-1 solutions was studied. As shown in Fig. 3 (Table S1), the pH-adjusted SALT-1 solutions (pH 7.2 and 8.0) showed no systematic differences when analyzed for phosphate, silicate, and nitrate with the exception of phosphate: after the third day—in the untreated SALT-1 sample at pH 8.9—its response underwent a significant drop. This effect was likely due to a biological contamination in that sample.
As for the silicate response, the higher values obtained at pH 7.2 could be explained by the silicate blank of the 0.1 M HCl used to buffer the medium. For the methods used, pH correction was not altering nutrient response; therefore, all certification experiments were performed on the 4.0% w/w SALT-1 solution prepared in water. When the SALT-1 is used with analytical methods that are sensitive to the initial pH of the sample, 0.1 M HCl could be used to adjust the pH accordingly (Fig. 2).
From Fig. 3 (Table S1), it can be noticed that the dissolution of phosphate and nitrate is instantaneous, whereas the silicate requires 24 h at room temperature (Fig. 4 and Table S2). For all quantitations described in this study, the 4.0% w/w SALT-1 solution was prepared in high-purity water and stored in PE bottles at room temperature. The solution was manually shaken to facilitate dissolution. Phosphate and nitrate were measured within 48 h after dissolution. Silicate was measured 24 h after dissolution (up to 3 days).
Major components: metals and anions
The artificial sea salt was chosen in order to create a material with a matrix composition similar to natural seawater. The reconstituted seawater prepared from SALT-1 (4.0% w/w) was analyzed for major components and compared to other seawater samples. Table 1 shows the results obtained by ICP−OES and ion chromatography with conductivity detection for major metals and anions normally found in seawater. For most components, the composition of the 4.0% w/w SALT-1 solution was within 20% of reference seawater samples. The only exception was boron which was present in the SALT-1 solution at a much lower level with respect to reference samples. However, boron in seawater is a trace element whose concentration can vary between 0.5 and 9.6 mg/L ; therefore, the difference reported in Table 1 can fit into the natural variations for the element.
Trace metal screening by ICP−MS
The 4.0% w/w SALT-1 solution was also screened for common trace metals measured in seawater. The trace metal content of SALT-1 was then compared to the NRC CASS-6 and NASS-7 seawater CRMs. Table 2 reports the ratio between the trace metal content measured in the SALT-1 solution and the two CRMs. For most elements, it was noticed that SALT-1 was close in metal composition with respect to CASS-6 and NASS-7: the content of As, Cd, Co, Cr, Cu, Mo, Ni, V, and Zn in the SALT-1 solution was within one order of magnitude (higher or lower) of the metal content of either CASS-6 or NASS-7.
The content of Fe, Mn, and Pb in the SALT-1 solution was significantly higher than the CASS-6 and NASS-7 by an average factor of 45, 245, and 731, respectively. On the other hand, the U content of the SALT-1 solution was only 1.7% of that of CASS-6 and NASS-7. Since the mass fraction of these trace elements is well below the micrograms per gram level, the variability has a modest impact on the physical–chemical properties of the matrix.
As shown previously, the artificial sea salt used for SALT-1 preparation had similar properties as natural seawater. Since the residual nutrient content was low, the analytes were spiked into the matrix. For this purpose, 600 g of sea salt matrix with 1.5 mL of a concentrated nutrient solution were homogenized using a planetary ball mill (Fig. 1).
After processing, the SALT-1 was further homogenized and split in smaller aliquots. Sub-portioning was aimed to reduce the exposure time of the matrix to the atmospheric humidity during the final bottling process. The SALT-1 was transferred into 32 × 125-mL PE bottles and 14 of them were analyzed for bulk homogeneity. As reported in Fig. S6, the RSD on the 14 measurements was < 1% for nitrate, phosphate, and silicate. As the bulk homogeneity was satisfactory, the material was aliquoted into the final form for further investigations. The recovery of the spiked analytes was quantitative for nitrate (98%) and phosphate (101%). The recovery for silicate was 109%, likely due to the presence of endogenous silicates in the salt matrix.
The homogeneity study was then performed on the SALT-1. Sixteen units were selected from the batch of 850. The selection was performed to cover all the possible inhomogeneity sources arising from bottling (Table S3). The 16 units were analyzed in triplicate and in randomized order to eliminate the effect of any drifts. The response was normalized to the gravimetric composition of the sample blends, and the data were analyzed using a one-way ANOVA approach. The relative homogeneity components are shown in Table 3, and all calculations are reported in the ESM. Each unit of SALT-1 is intended to be used in whole, without subsampling (~ 4.1 g sample into 100 mL of solution).
Stability and redox properties
During the first exploratory experiments, nitrite was the analyte used to assess milling and homogenization process. Although the planetary ball milling (Fig. 1) was suitable to provide a highly homogenous material, significant loss of the added nitrite occurred. As shown previously, other nutrients did not degrade. It was hypothesized that during mixing at elevated temperature, nitrite could be converted to nitrate. Such an effect was studied by mass spectrometry following the conversion of 15N isotopically enriched nitrite. For this purpose, 90 mg of isotopically enriched Na15NO2 was dissolved in 1.5 mL of water and blended with 600 g of artificial sea salt resulting in w(15NO2−) of 100.7 mg/kg. The homogenization was performed using the planetary ball mill with the same program employed for the preparation of the SALT-1. After mixing, 59% of the added 15NO2− was converted into 15NO3−. This result confirmed that the loss of nitrite followed an oxidative pathway. The sample was kept at room temperature and analyzed over time. After 3 days, the nitrite-to-nitrate conversion was 72%. However, no further conversion was observed after the third day when the sample was kept at room temperature (Fig. S7). To complete this series of experiments, the residual content of nitrite was measured on two aliquots of sample previously kept at 40 °C (34 days) and 105 °C (12 days). Despite the significant degradation of nitrite, 7.7 mg/kg and 0.59 mg/kg 15NO2− (out of the 100.7 mg/kg originally spiked) were still found in the respective samples. The conversion of nitrite to nitrate suggests a temperature-modulated redox chemical process on the salt surface. This effect is a limit to the stability of those analytes that can undergo redox degradation.
Short-term stability modeling extremes of transportation conditions was evaluated. Three sets of three SALT-1 units were incubated at -20 °C, 20 °C, and 40 °C, respectively. The reference SALT-1 unit was kept at 4 °C (storage condition). After 10 days, one unit of SALT-1 stored at the three different temperatures was withdrawn and placed at 4 °C along with the reference unit. This operation was then repeated after 21 and 30 days. All nine SALT-1 units along with the reference one were analyzed within repeatability conditions. Each unit was analyzed three times, and the sequence was randomized to avoid effects associated with drifts. The response of the samples was normalized to the response of the reference unit. As shown in ESM, no trends in the nutrient response were observed under typical transportation condition resulting in ushort = 0.00% as shown in Table 3.
The SALT-1 units are stored at 4 °C enclosed in thermally sealed triple laminated bags in order to limit air exchange and potential moisture uptake. Under these conditions, issues related to the long-term stability of the material were not noticed. The material was produced in April 2018 by gravimetric addition of nutrients into the artificial sea salt. The SALT-1 was analyzed for the first time in August 2019, and the certification was performed in May 2021, following one further analysis in October 2021. As reported in Fig. S8, no significant trends in nutrient concentrations were observed; as a consequence, zero uncertainty component for the long-term stability was assigned (ulong, Table 3). Monitoring of stability will continue throughout the lifecycle of the CRM.
SALT-1 property values
Fifteen units were sampled across the SALT-1 lot and used for characterization. The analyses were carried out over 3 weeks: each week, five SALT-1 units were analyzed in duplicate for phosphate, silicate, and nitrate. The replicate analyses were independent from each other: the second analysis was carried out after 24 h from the first one and a fresh set of standards was prepared for quantitation. All mathematical models for the calculation of the analytical results and the uncertainties are reported in the ESM. Briefly, nitrate results were obtained by exact matching quadruple isotope dilution following triethyloxonium derivatization [43, 50, 51], and the corresponding uncertainty was calculated by error propagation (see Paragraph S1.11 ). For the determination of phosphate and silicate, external calibration using standards prepared in a matrix-matched medium (low-nutrient seawater, OSIL) was employed. Some minor nonlinear trends were observed for these two analytes (Fig. S9); therefore, the standard addition calibration was not implemented [53, 54]. As reported in Fig. S10, the external calibration using matrix matching was adequate to account for rotational matrix effects. In fact, the calibration curve prepared into the 4.0% w/w SALT-1 solution was not statistically different with respect to the one obtained in low-nutrient seawater, for both phosphate and silicate. Furthermore, both the 4.0% w/w SALT-1 solution and the low-nutrient seawater medium showed similar kinetics of color development for phosphate, silicate, and nitrite (Fig. S11).
For phosphate and silicate, the calibration was obtained with a 5-level calibration plot where the middle point was exactly matched to the mass fraction of the analytes in the sample. Results were calculated using both linear (y = a0 + a1x) and quadratic (y = a0 + a1x + a2x2) model; since the relative differences between the two approaches was less than 0.3% on average (Fig. S12), the results from the more intuitive linear model were used for value assignment instead of the quadratic which offers some allowance for potential nonlinearities. For phosphate and silicate, the uncertainty on single measurement (i.e., characterization uncertainty) was estimated with both uncertainty propagation  and Monte Carlo approach . As shown in Fig. S12, the uncertainty propagation estimate was more robust; therefore, it was used to assign the uncertainty component due to characterization (uchar, Table 3). In Table 4, all approaches for the evaluation of the characterization uncertainty are compared, including the relative standard deviation from repeated measurements. Although being based on different principles, all three estimates are in a reasonable agreement.
Two calibration standards were used for quantitation, one from NIST and one from MilliporeSigma (traceable to NIST). Within the limit of the experimental error, there were no differences between results from using the two individual sets of standards. Furthermore, within each measurement sequence, one quality control sample (MOOS-3 CRM) and one spike recovery sample were analyzed. As shown in Tables 5 and 6, quantitative recoveries were obtained.
The mass fraction of the nutrients in the SALT-1 powder is as follows: w(phosphate, PO43−) = 3.744 ± 0.049 μg/g, w(silicate as SiO2) = 13.04 ± 0.42 μg/g, and w(nitrate, NO3−) = 28.73 ± 0.57 μg/g. These values were obtained by average of the analytical results. The corresponding uncertainty (k = 2) was obtained by combining the contributions due to homogeneity and characterization (zero uncertainty was assigned to short- and long-term stability). The density of the resulting 4.0% w/w SALT-1 solution was 1.02321 ± 0.00091 g/mL (k = 2, n = 16) as measured with a 25-mL pycnometer.
Phosphate and silicate by HPLC ICP−MS/MS
Value assignment for both phosphate and silicate was obtained using the traditional photometric methods. In order to confirm these values, an independent approach based on HPLC ICP−MS/MS was developed (Paragraph S1.5). On a total of 11 measurements by standard addition on three units of SALT-1, an average phosphate content of 3.64 ± 0.24 μg/g (RSD 6.7%, k = 1) and an average silicate content of 13.54 ± 0.47 μg/g (RSD 3.4%, k = 1) was obtained. These values are in reasonable agreement with the ones obtained by photometry as reported in the previous paragraph.
Reconstitution of the SALT-1
The SALT-1 was designed as a proxy for the MOOS-3 CRM . Despite SALT-1 lacking nitrite, all other nutrients are close to the level found in MOOS-3. When the entire content of a SALT-1 unit was reconstituted in water to a final volume of 100 mL, the amount concentration of nutrients in the resulting solution was as reported in Table 7 . Such values are within a factor of 4 with respect to the nutrient content found in the MOOS-3. Although all the testing for the certification of SALT-1 were performed at the 4.0% w/w level, the reconstitution of SALT-1 to higher (or lower) salinities would allow expanding the scope of the CRM to salinity ranges different than those of typical seawater.