δ 15 N analysis of ammonium in freeze ‐ dried natural groundwater samples by precipitation with sodium tetraphenylborate

To perform the δ 15 N isotopic analysis of ammonium (NH 4+ ) with an elemental analyzer (EA) coupled to an isotope ‐ ratio mass spectrometer, it is necessary to isolate the NH 4+ prior to the analysis. Existing methods are work ‐ intensive and time ‐ consuming. For broader applicability in the environmental sciences, it is desirable to simplify the sample preparation process. The method used here is based on the insolubility of ammonium tetraphenylborate ((C 6 H 5 ) 4 BNH 4 ) in water. Its suitability for the stable isotope measurement of δ 15 N values for NH 4+ has already been proven (Howa et al. Rapid Commun Mass Spectrom . 2014;28:1530 ‐ 1534). However, there are no studies on the usability of the method for the analysis of ammonium ‐ containing water samples. In this study, the method was tested for its applicability to natural groundwater samples. To achieve the necessary high NH 4+ concentrations for complete (C 6 H 5 ) 4 BNH 4 precipitation, the water samples were first freeze ‐ dried and then prepared for the analysis. To precipitate the NH 4+ , sodium tetraphenylborate ((C 6 H 5 ) 4 BNa) was added to the samples. The precipitate was then separated from the water by filtration using a membrane filter and analyzed using an EA interfaced with an isotope ratio mass spectrometer to determine the nitrogen isotope ratio.


| INTRODUCTION
Nitrogen and its compounds are crucial and ubiquitous components of natural ecosystems. However, the excessive supply of nitrogen compounds in natural cycles leads to numerous problems. 1 Therefore, it is essential to reduce the environmental nitrogen load rapidly and effectively. For this purpose, a better understanding of the nitrogen cycle process is necessary. [1][2][3] For the identification and quantification of conversion processes, the measurement of the stable isotope ratios of the nitrogen in dissolved NH 4 + from water samples can be valuable. 4 It is necessary to isolate NH 4 + prior to analysis when performing a δ 15 N analysis of NH 4 + by elemental analyzer/isotope ratio mass spectrometry (EA/IRMS). The most relevant sample preparation methods for isolating NH 4 + from aqueous samples are distillation, mercury precipitation, cation exchange, and the diffusion method (DM). 5 The distillation method was one of the first sample preparation methods for NH 4 + to be developed, but it is very timeconsuming and expensive, and it is unsuitable for large numbers of samples. 6 Mercury precipitation enables NH 4 + isolation from aqueous samples by forming a sparingly soluble ammonia-mercury complex. 7 However, the handling of mercury compounds is hazardous to human health and the environment. Another way to isolate NH 4 + from aqueous samples is via cation exchange, 8 and cation-exchange resins are used for this purpose. The accumulated NH 4 + is then redissolved and converted into a solid form. However, this method is not applicable to samples with high cation contents.
Furthermore, it is limited to increasing the concentration. Also, it is not a direct preparation method, as further treatment of the solution is necessary. The DM is a relatively common method for isolating NH 4 + from aqueous samples. However, it takes a long time for the NH 4 -N to be converted. During this time, further transformation processes could take place, leading to fractionation. 9 The tetraphenylborate (TPB) method is based on the poor solubility of ammonium tetraphenylborate ((C 6 H 5 ) 4 BNH 4 ) in water (9.92 mg/L at 25°C 10 ), and the method is known for its qualitative and quantitative detection of NH 4 + . 11

| Preservation of natural groundwater samples
Waters, especially surface water, sewage and groundwater, are subject to changes due to physical, chemical or biological reactions that can occur between sampling and analysis. 14

| Freeze-drying of water samples
The freeze-drying process was carried out with a freeze dryer (LD plus 1-4; Martin Christ Gefriertrocknungsanlagen, Osterode, Germany). The required sample volume, which contained 1 mg of NH 4 -N, was placed into glass containers, sealed with a sealing film (PARAFILM® M, American National Can, Chicago, IL, USA), frozen, and then freeze-dried.
Prior to freeze-drying, the sealing film was perforated with a needle.

| On-line isotope analysis
All in-house standards and the precipitated (C 6

| Diffusion method for NH 4 + isolation from aqueous samples
To test the applicability of the TPB method for natural water samples, a comparison with the DM was carried out. The DM 18 was initially performed on the in-house standards Lab-32 and Lab-14 to estimate the accuracy of the preparation method. Next, natural groundwater samples were prepared with the DM and the results were compared with these measured by the TPB method.

| TPB method testing with non-freeze-dried water samples
To identify the direct influence of sample preparation on the δ 15 N values, the three different NH 4 + in-house standards were dissolved in 3 mL of deionized water and then treated by the modified TPB method. This series of measurements was referred to as`non-freezedriedonly NH 4 +´. Next, NaNO 3 (Lab-13) was added to the solutions at varying amounts to show that the deviation of the measured isotope ratios from the expected ratio was independent of the NO 3 − concentration. This series of measurements was referred to as`non- Another series of measurements (`nonfreeze-driedplus K +´) was performed to identify the influence of competing K + (added as KCl). Finally, measurements were made with all three ions in the water sample (`non-freeze-driedplus NO 3 − + K +´) . Table 1 lists the details of the four measurement series.

| TPB method testing with freeze-dried water samples
To test the effect of freeze-drying on isotope fractionation, various sample volumes were prepared with a constant amount of NH 4 + and then freeze-dried. For this purpose, the in-house standards were again used, dissolved in deionized water. First, the water samples were freeze-dried without acidification. In a further series of tests, the samples were acidified with concentrated H 2 SO 4 to pH 2 before freeze-drying. The details of the measurement series can be found in Table 2. For comparison, the series of undried samples from the previous experiment was used.

| Evaluation of the isotopic data
Each sample was measured multiple times by EA/IRMS. The mean value (x _ ) and the associated standard deviation (sd) were calculated from the measurements.
As the measured isotope ratios were normalized using the international reference materials, and the actual uncertainty of the isotopic data (u measured ) was determined using a`Kragten spreadsheet´for the estimation of measurement uncertainty arising from normalization 16     international standards used for normalization. This, in turn, included the reported standard deviation of the true isotope ratios of the reference materials, as well as the standard deviation of the measured isotope ratios of the reference materials. For more detailed information, see supplement S1 (supporting information).
For the calculation of the combined uncertainties (u combined ) of the in-house standards prepared by the TPB method (or the DM), the uncertainties of the measured isotope ratios (u measured ) and the uncertainties of the in-house standards (u expected ) that were previously determined by EA/IRMS measurement were calculated by a simple error propagation (see Equation 1).
To calculate the error caused by the sample preparation, the RMSE of the in-house standards was used, which is equal to the total uncertainty (u total ), which is calculated from the sample preparation (u preparation ), the uncertainty of the in-house standards (u expected ), and the measured isotope ratios (u measured ). Equations (2) and (3) were used for this calculation.
For the statistics, the significance level was set at α = 0.05. The Shapiro-Wilk test was used to test the data for a normal distribution. Based on that, a normal distribution could be assumed for the data used. Therefore, an analysis of variance (ANOVA) or a ttest was performed for the significance analysis; several studies have proven that ANOVA is relatively robust against the misallocation of a normal distribution. 19,20 For the correlation analysis, the Pearson test was used. RStudio software was used for the statistical analysis. 21

| International standards and in-house standards
For quality assurance of the isotope data, the measured isotope ratios  Table 4 the long-term δ 15 N values of the used in-house standards are shown.

| TPB method testing with non-freeze-dried water samples
All double or triple isotopic δ 15 N measurements of the samples prepared by the TPB method had a pooled standard deviation of u measured ≤0.2‰.  For better comparability of the results of the different salts, the actual nitrogen isotope ratios will not be stated but rather the difference between the measured (δ 15 N measured ) and the expected (δ 15 N expected ) values. Henceforth, this is referred to as the`deviation´.
First, the`only NH 4 +´s eries was considered and evaluated (see Table 1). The comparison of the three in-house standards showed no significant difference in their mean deviations (ANOVA: mean deviation = 0.65‰, sd = 0.28‰, p = 0.97, see Figure 1). Taking the uncertainties of the in-house standards (u expected ) into account, the error propagation law yields combined uncertainties of u combined = ±0.41‰ for Lab-14, u combined = ±0.39‰ for Lab-32, and u combined = ±0.26‰ for Lab-8 (see Equation 1) in the`only NH 4 +´s eries.
The deviation from the expected isotope ratio was detected to be an offset in the measured isotope ratios of +0.65‰. The reason for this offset could not be identified. One possible explanation might be that the heavy isotope fractions were preferentially precipitated by the TPB. However, this is doubtful because the concentration measurements of the filtrate gave NH 4 + residual amounts of <1-2% based on the initial amount. Thus, at least 98% of the dissolved NH 4 + was precipitated and retained as precipitate on the filter.
Fractionation by poor recovery rates is therefore very unlikely.
Another explanation could be that drying the precipitate at 60°C led to fractionation. It may, therefore, be useful to freeze-dry the filters instead. This question requires further investigation. The most likely explanation is that this offset was a bias from the EA/IRMS system.
Based on previous studies, 22,23 it can be assumed that the calibration of the organic TPB samples with the international standards leads to a bias due to incomplete reduction of USGS 34. In Figure 2  It could be assumed that the nitrogen isotope ratio of NH 4 + was not influenced by that of NO 3 − . Nevertheless, the measured nitrogen isotope ratios of the`plus NO 3 −´s eries were examined more closely.
There was no significant difference between the three in-house standards (ANOVA: p = 0.17). Figure 3 shows the influence of the

| TPB method testing with freeze-dried water samples
All duplicate or triplicate EA/IRMS measurements of the samples had a standard deviation of ≤0.2‰. There were only three exceptions within the samples that were freeze-dried without acidifying (with sd = 0.3‰, sd = 0.8‰, and sd = 1.1‰). As before, for better comparability of the measured isotope ratios between the different NH 4 + in-house standards, the deviation between the measured and the expected isotope ratios is reported rather than the isotope ratios themselves.
In Figure 4, the deviations of the measured nitrogen isotope ratios from their expected isotope ratios for the freeze-dried samples are shown. Freeze-drying without acidification of the water samples cannot be used for volume reduction, because the error increases substantially with increasing sample volume. At between 2 mL and 25 mL of sample volume, the mean deviation was between 0.46‰ ± 0.18‰ and 0.75‰ ± 0.35‰ and thus did not significantly differ from that of the undried samples. However, at a volume of 50 mL the mean deviation was 0.39‰ ± 1.14‰, and at 500 mL the mean deviation was 3.88‰ ± 5.65‰. These findings show that there is increased isotope fractionation with increasing volume. The individually measured isotope ratios in the series 'freeze-dried -

| Uncertainty of the sample preparation method
The DM was initially performed on the in-house standards Lab-32 and

| TPB method testing with natural water samples
The pooled standard deviations of the multiple EA/IRMS measurements were <0.3‰ for the DM and <0.2‰ for the TPB method.
In Figure 5, the δ 15 N values of NH 4 + in the natural groundwater samples are shown. The nitrogen isotope ratios determined by the TPB method were compared with those determined by the DM. The left diagram shows the actual measured nitrogen isotope ratios. In the right diagram, the data analyzed by the TPB method were corrected against the determined offset of +0.65‰. The DM analyses are unchanged. The line in the middle of the Figure marks the exact accordance of the measured isotope ratios. Apparently, there was no significant difference in the isotope data of the natural groundwater samples between the two preparation methods (paired t-test: t(21) = 1.04, p = 0.31 (corrected for offset in the TPB method)).
These results indicate that the TPB method could be used for natural water samples with more complex matrices. The data is provided in supplement S2 (supporting information).

| CONCLUSIONS
Based on the results of the analyzed in-house standards and the error propagation law, the uncertainty added by the sample preparation with TPB was ±0.1‰, with an offset of +0.65‰. To use the TPB method, high NH 4 + concentrations are required. Therefore, the sample preparation was preceded by freeze-drying. All water samples must be acidified to pH 2 before freeze-drying. Freezedrying requires a volume-dependent correction of the measured δ 15 N values.
It is recommended that at least two in-house standards should be provided, in the form of (C 6 H 5 ) 4 BNa. All samples should be calibrated with these standards; so no offset correction is necessary. For water samples that contain more than twice as much K + as NH 4 + , it is recommended to that standards should be used that have an Since fractionation depends on freeze-drying, which may differ in different laboratories, it is recommended that calibration curve should be created with at least two standards for the correction of the measured values, depending on the dried sample volume. In this study, the fractionation was −0.001‰/mL at pH 2, i. e., 0.001‰ must be added to the measured value per mL of freeze-dried water sample. In this case, the correction could be omitted for volumes below 100 mL.