Accelerating 1H NMR Detection of Aqueous Ammonia.

Direct electrolytic N2 reduction to ammonia (NH3) is a renewable alternative to the Haber-Bosch process. The activity and selectivity of electrocatalysts are evaluated by measuring the amount of NH3 in the electrolyte. Quantitative 1H nuclear magnetic resonance (qNMR) detection reduces the bench time to analyze samples of NH3 (present in the assay as NH4 +) compared to conventional spectrophotometric methods. However, many groups do not have access to an NMR spectrometer with sufficiently high sensitivity. We report that by adding 1 mM paramagnetic Gd3+ ions to the NMR sample, the required analysis time can be reduced by an order of magnitude such that fast NH4 + detection becomes accessible with a standard NMR spectrometer. Accurate, internally calibrated quantification is possible over a wide pH range.

Introduction

Ammonia (NH3) is one of the largest chemical commodities responsible for about 1.5% of global energy use and associated CO2 emissions from the Haber–Bosch process. Its primary use is as a feedstock for nitrogen-based fertilizers. Electrochemical, fossil-fuel-free methods to produce ammonia are gaining significant interest for the reduction of CO2 emissions, as well as enable ammonia as a carbon-free energy carrier and storage material.[1−3] Direct electrolytic N2 reduction is a renewable alternative to synthesize NH3. To suppress the undesirable hydrogen evolution reaction, a selective electrocatalyst is needed.[4]

The activity and selectivity of electrocatalysts are quantified by measuring the accumulated NH3 in the electrolyte with appropriate detection methods. Recently, a number of papers were published concerning the difficulty of obtaining reproducible results in nitrogen reduction research.[5,6] Such difficulty is related to experimental procedures and significant amounts of NH3 from dust, ambient air, 15N2, and desorption from cell surfaces. In addition, NOx contamination or nitrogen-containing catalyst precursors can be reduced to NH3 during electrolysis, which can be falsely attributed to N2 reduction.[7,8] Reliable testing and analysis procedures including control experiments with an isotope labeled 15N2 are necessary to avoid false positives.[5] Conventional spectrophotometric NH3 detection methods such as the indophenol blue method cannot distinguish between isotopologues of NH3 and require considerable bench time.[9,10]1H NMR spectroscopy is a fast, accessible, and isotopically selective alternative to spectrophotometric methods for NH3 detection, but as we will show below, the sensitivity on a standard 400 MHz spectrometer is insufficient.[11] Here, we present a powerful liquid-state NMR method with sufficient sensitivity on relatively easily accessible NMR spectrometers, i.e., requiring limited field strength and normal sensitivity probes.

Several alternatives to spectrophotometric NH3 detection methods have been proposed recently. Ion chromatography can be used for ammonia detection but isotopologues cannot be distinguished and an overlap of NH4+ with other cations poses a threat to the accuracy of the method.[12] Yu et al. proposed ultrahigh-performance liquid chromatography–mass spectroscopy (UPLC-MS) to measure derivatized solutions of NH3.[13] The method is very sensitive and capable of distinguishing isotopologues of NH3 but requires careful control over the pH. Quantitative 1H NMR has been widely adopted to quantify 15NH4+ from control experiments with 15N2. The acidified form of ammonia, ammonium (14NH4+), and its isotopologue 15NH4+ have an unmistakable fingerprint in the 1H NMR spectrum.[11]

Quantitative 1H NMR is based on the proportional relationship between the signal integral I and the number of protons N responsible for that particular signalwhere KS is a proportionality factor that depends on the physicochemical properties of the sample. To achieve accurate quantification, changes in KS have to be accounted for by a suitable quantification method.[14] Pulse length-based concentration determination (PULCON) uses the principle of reciprocity to correlate the absolute intensity of two spectra measured in different solutions.[15] Nielander et al. successfully applied the PULCON method to NH4+ quantification.[11]

Since fluctuations of KS affect all resonances in the spectrum equally, the ratio of two peaks is independent of KS and can therefore be used for quantification. Typically, an internal standard of known concentration is added as a reference. The concentration of NH4+ can be quantified by either relative or absolute quantification. For relative quantification, a calibration curve is generated by measuring standard solutions of NH4+ and an internal standard.[14] The prerequisites for accurate ammonia quantification with relative quantification were recently described..[16] Absolute quantification allows the calculation of the NH4+ concentration directly from the integral of the peaks of NH4+ and the internal standard without requiring a calibration curve according towhere I, N, and C are the integral area, number of nuclei, and concentration of NH4+ and standard, respectively. Absolute quantification requires that the total time spent to acquire one scan, the interscan delay tscan, is at least 5 times the longest longitudinal relaxation time T1 in the sample.[14] Despite the advantages of absolute quantification (calibration-free and robust) so far, no absolute quantification method has been proposed for ammonia detection. Hodgetts et al. reported that a d1, T1, and proton exchange-induced loss of coherence affects the NH4+ peak, rendering absolute quantification not suitable for NH4+ detection.[16] By adding a suitable paramagnetic salt, accurate absolute quantification becomes possible because the T1 values of both the internal standard and NH4+ are reduced so that there is insufficient time for the proton exchange to induce a loss of coherence as we will see below.

The lower limit of quantification (LOQ) of a detection method is the lowest concentration of NH3 that can be measured within an acceptable time and with an acceptable accuracy. The LOQ depends on the sensitivity, which is calculated from the signal-to-noise ratio (SNR) at a certain interscan delay tscan

To reduce the minimum LOQ by a factor of 2, the analysis time has to be quadrupled.[17] Nielander et al. could detect 1 μM NH4+ in ethanol within 1 h (tscan = 2 s) with a 900 MHz NMR and cryo-probe.[11] The sensitivity difference between a 900 MHz NMR and a standard laboratory NMR (400 MHz) is substantial. The type of probe and the field strength difference lead to 1 order of magnitude lower sensitivity, which leads to 2 orders of magnitude longer analysis time to achieve the same LOQ on a 400 MHz NMR without cryo-probe.[18] To compensate for the lower sensitivity, longer experiments (typically several hours) are necessary to accumulate enough NH4+ in the electrolyte to reach the detection limit, which is unfavorable and which in addition increases the risk of false negatives due to deactivation and of false positives due to contamination. Higher 1H NMR sensitivity is needed to enable laboratories with more standard NMR spectrometers to quantify NH3 efficiently.

The type of the pulse sequence influences the NH4+ sensitivity strongly.[11,16] The signal from the hydrogen atoms of the solvent has to be suppressed to avoid baseline distortions and low receiver gain. Nielander et al. showed that pulse sequences that utilize pulsed field gradients in combination with selective excitation pulses are very effective at suppressing water without removing the NH4+ signal.[11] These pulse sequences use pulsed field gradients to dephase the water resonance and selective pulses to ensure that during acquisition water is completely out of phase while NH4+ is in phase.[19]

The T1 of a molecule influences the sensitivity, because for a given interscan delay tscan, T1 determines the percentage of spins that can relax back to equilibrium in between scans. A smaller percentage relaxation leads to less acquired signal per scan according towhere M and M0 are the magnetization in the z-axis following tscan and at full relaxation, respectively. The interscan delay tscan is composed of the recycle delay d1 and the acquisition time. It is noteworthy that for some pulse sequences, the percentage relaxation only depends on the recycle delay, not on the acquisition time, as will be discussed in more detail below. Reducing the interscan delay, for example, by fast sampling is a well-known strategy to improve the 1H NMR sensitivity.[20] Another strategy to lower the interscan delay is to shorten the T1 of the analyte, which has the advantage that the same percentage relaxation can be achieved at a lower interscan delay.[14]T1 is determined by the fluctuating magnetic interactions due to nearby magnetic moment fluctuations and due to positional changes of surrounding nuclei and moments. Interactions with unpaired electrons of paramagnetic substances are 1000 times larger than typical interactions between nuclear magnetic moments. Therefore, a small amount of a paramagnetic substance is sufficient to lower T1 drastically.[21] This concept is applied in contrast agents for medical magnetic resonance imaging (MRI). The so-called paramagnetic relaxation enhancement (PRE) is also a common strategy to overcome sensitivity barriers for small organic molecules and proteins, because as T1 decreases, more scans can be acquired in the same amount of time.[18,22,23]

Results and Discussion

The Gd3+ ion is widely used for PRE in medical MRI due to its large magnetic moment from seven unpaired electrons.[24] We investigated the influence of paramagnetic Gd3+ ions on the sensitivity for aqueous ammonia detection to enable 1H NMR as a routine analysis tool for NH4+ quantification, with the use of an internal standard (absolute quantification). In agreement with Nielander et al., we found that pulse sequences that are suppressing the water resonance by dephasing it during acquisition are well suited for NH4+ detection.[11] With excitation sculpting (ES), the water resonance is suppressed effectively and a flat baseline is obtained around the NH4+ triplet. However, at 40 μM NH4+, the SNR is only 13.6 for a 12.8 min measurement on a 400 MHz NMR with room-temperature probe (see Figure a). With this sensitivity, it takes 4 h until the accumulated ammonia in the electrolyte produced by a catalyst with intermediate activity becomes quantifiable by NMR (calculation in the Supporting Information, SI). Therefore, we sought to improve the sensitivity by adding 1 mM paramagnetic Gd3+ to the NMR tube. Maleic acid (MA) was added as an internal standard to quantify the amount of NH4+ with absolute quantification. The singlet of maleic acid at ca. 6.21 ppm is sufficiently separated from the NH4+ triplet at ca. 6.9 ppm. The T1 values of both NH4+ and MA decrease drastically after the addition of Gd3+. T1 decreases from 2.16 to 0.14 s and from 2.05 to 0.13 s for NH4+ and MA, respectively. This 15.4-fold reduction of the T1 of NH4+ enables a reduction of the interscan delay by the same factor, which, according to eq , leads to a potential 3.9-fold sensitivity increase (). The linewidth of NH4+ increases only slightly with the addition of Gd3+ from 3.6 to 4.2 Hz.

Figure 1

1H NMR sensitivity gain from 0 mM Gd3+ (black, green, and gray) to 1 mM Gd3+ (red, blue, orange) in the NMR tube measured with different acquisition parameters and pulse sequences. (a) 40 μM NH4+ measured with identical total acquisition time (12.8 min) and a recycle delay d1 of 10 and 0.75 s for 0 mM Gd3+ and 1 mM Gd3+, respectively. Pulse sequence: Excitation sculpting. (b) 40 μM NH4+ measured with identical total acquisition time (10.7 min) and recycle delay (0.5 s). Pulse sequence: Excitation sculpting. (c) 40 μM NH4+ measured with the same acquisition parameters as (b) but with the double pulsed field gradient spin echo (DPFGSE) pulse sequence. All SNR values are averages from a triplet measurement. (d) Effect of 1 mM Gd3+ on T1 of NH4+ and maleic acid. Field strength: 400 MHz.

To show that the measured sensitivity gain matches the sensitivity gain predicted from T1 measurements, we measured a sample of 40 μM NH4+ with and without 1 mM Gd3+ using different acquisition parameters (Figure a–c). In Figure a, the total analysis time is identical in both measurements and d1 is set to 5T1 so that NH4+ has the same percentage relaxation in both cases. With 1 mM Gd3+, the sensitivity is significantly (factor 2.4) higher but not as much as expected from the T1 decrease (factor 3.9). We will later show that a sensitivity gain close to the predicted value can be measured directly by removing an additional 90° pulse that is by default included in the ES pulse sequence. With the default version of ES, the sensitivity gain is lower than expected from the T1 decrease because the additional 90° pulse removes the contribution of the acquisition time to the percentage relaxation. Consequently, the acquisition time only adds time to the total analysis time without improving signal strength and the percentage relaxation depends only on d1. Since the acquisition time makes up a larger fraction of the interscan delay at low interscan delays, the decrease of sensitivity is more pronounced with 1 mM Gd3+ where the acquisition time makes up 0.75 s of the total 1.5 s interscan delay. Without Gd3+, only 2 s out of 12 s interscan delay is the acquisition time that leads to a smaller sensitivity loss. In other words, the interscan delay could be factors 1.2 and 2 smaller for 0 mM Gd3+ and 1 mM Gd3+, respectively. Therefore, the sensitivity gain with 1 mM Gd3+ would increase by a factor 1.3 () from 2.4 to 3.1 if the sensitivity loss would have been equal in both cases. Taking into account ≈15% sensitivity loss due to line broadening, the sensitivity gain is 3.6, which is close to the predicted value.

The experiment shown in Figure a is not sufficient to prove a sensitivity gain because it only shows that with a higher recycle delay, less scans can be acquired in the same amount of time. Less scans will always lead to lower SNR. To prove a sensitivity increase, it is necessary to show that a larger recycle delay is necessary with 0 mM Gd3+ but not with 1 mM Gd3+. This is shown in Figure b, where both 0 mM Gd3+ and 1 mM Gd3+ were measured with low recycle delay (0.5 s) and identical total acquisition time. The sensitivity without Gd3+ is 3.9 times lower, indicating that a large fraction of the signal is lost due to low percentage relaxation. The percentage relaxation at a recycle delay of 0.5 s is 20.7 and 97.2% for T1 of 2.16 and 0.14 s, respectively. Therefore, 4.7 times more signal can be expected with 1 mM Gd3+ in the same amount of time. We assume that 15% of that signal increase is lost due to line broadening with Gd3+, which results in sensitivity improvement by a factor of 4.08. This value agrees well with the experimentally observed value of 3.9.

To study if adding Gd3+ also improves the sensitivity with other pulse sequences, we measured the sensitivity gain with the double pulsed field gradient spin echo (DPFGSE) pulse sequence (Figure c) using the same acquisition parameters as in Figure b. The sensitivity gain with DPFGSE (2.1) is lower than with ES (3.9). The reason for this is that with DPFGSE, the percentage relaxation has to be calculated using the full interscan delay including acquisition time, not just the d1 as for ES. Using the same methodology as in (b), we calculate the percentage relaxation with and without Gd3+ and arrive at an expected sensitivity gain of 1.8, which agrees well with the experimentally measured value. The sensitivity gain is lower than in (b) because with 1 mM Gd3+, the chosen tscan of 1.25 s is almost 9 times longer than the T1 of NH4+, which means that tscan is much longer than the necessary 5T1, and as a result, sensitivity is lost. The previous examples demonstrate that after addition of 1 mM Gd3+, a significant sensitivity gain is observed with different acquisition parameters and pulse sequences, and this sensitivity gain agrees well with the expected values predicted from T1 measurements.

We measured the accuracy of NH4+ quantification with 1 mM Gd3+ by calculating the NH4+ concentration from the intensities of MA and NH4+ using eq and comparing it to the gravimetrically measured concentration (Figure a,b). The method has very good linearity (R2 = 0.999) and an acceptable relative error (≤10%) in the NH4+ concentration range of 30–388 μM with the ES pulse sequence. The relative error is randomly distributed around the abscissa, which suggests that it is caused by integration errors. Higher accuracy (relative error ≤ 5.3%) was obtained with an isotope labeled 15NH4+ (see Figure S2). 15NH4+ can be quantified with higher accuracy because it appears in the NMR spectrum as a doublet, which has inherently higher SNR than the 14NH4+ triplet.

Figure 2

(a) Linearity and (b) accuracy of NH4+ quantification with 1 mM Gd3+. Error bars around each point represent the standard deviation for each triplicate measurement. Pulse sequence: Excitation sculpting, recycle delay: 0.5 s, acquisition time: 0.75 s, total analysis time: 10.7 min, field strength: 400 MHz.

The pH of the catholyte, which is used for detection, can vary over time due to acidic or alkaline species produced in the electrochemical reaction or because of migration of ions induced by the electric field.[25,26] N2 reduction experiments are especially prone to pH changes because the electrolyte volume is minimized to maximize the signal for ammonia detection. Both UPLC-MS and the indophenol method are sensitive to pH changes because the pH influences the reaction that is carried out prior to analysis.[9,13] Therefore, additional dilution steps can be necessary to measure accurately with these methods.

To investigate if the accuracy of our 1H NMR method depends on the pH, we acidified a sample of 388 μM NH3 with different concentrations of H2SO4 (Figure a). Based on the previous finding that the T1 values of NH4+ and MA are very close to each other, we chose a recycle delay of 0.5 s (3T1) for this experiment. For acid concentrations above 37 mM H2SO4, the relative error continuously increases when more acid is added. To investigate if the growing relative error might be caused by changing T1 values, we measured the T1 values of NH4+ and MA at 370 mM H2SO4 (Figure b). The gap between the T1 values of NH4+ and MA is slightly larger at 370 mM H2SO4 than at 37 mM H2SO4 which might explain the larger error. After increasing the recycle delay from 0.5 to 1.5 s to compensate for the increased T1 gap, the relative error decreases to <2% between 37 and 222 mM H2SO4. This suggests that the detection method is accurate over a wide pH range if a higher d1 is chosen to compensate for T1 changes.

Figure 3

Influence of H2SO4 concentration in the NMR tube on the accuracy of NH4+ quantification with d1 of 0.5 s (red) and 1.5 s (black). Error bars around each point represent the standard deviation for each triplicate measurement. (a) T1 of NH4+ and maleic acid at two different H2SO4 concentrations. (b) Acquisition parameters: at = 0.75 s, nt = 512, Pulse sequence: Excitation sculpting, field strength: 400 MHz.

In Figure , even with a high recycle delay of 1.5 s, an unusually high error remains at the highest and lowest acid concentrations. The error at the lowest acid concentrations is in agreement with the results by Hodgetts et al. and is caused by deprotonation of MA below 20 mM H2SO4.[16] Spectra acquired at the highest acid concentration had phasing issues, which had to be corrected by postprocessing the spectrum using the autophasing algorithm in the software package MestReNova. We suspect that the phasing issues are caused by tuning and matching, which become more difficult at high salt concentrations.[14] To achieve maximum accuracy, the acid concentration should not exceed 222 mM.

As discussed previously, with the default settings of the ES pulse sequence, the acquisition time does not contribute to the percentage relaxation so that sensitivity is lost. To determine the maximum sensitivity for NH4+ detection with 1 mM Gd3+, we deactivated the additional 90° pulse at the beginning of the pulse sequence so that both acquisition time and recycle delay contribute to the percentage relaxation. This leads to a significant increase in sensitivity (see Figure ). The sensitivity can be further increased by reducing the interscan delay from 5T1 to 3T1 which is feasible in this case because the T1 values of NH4+ and MA are very close to each other. The SNR of a 40 μM NH4+ sample measured for 14.6 min (interscan delay 3T1) is 47.4. This corresponds to a 1.4-fold sensitivity increase compared with the activated 90° pulse. The relative error is similar to an interscan delay of 5T1 and 3T1 (<6%), indicating that the interscan delay can be reduced without sacrificing accuracy. As discussed above, at high acid concentrations, a higher recycle than 3T1 might be necessary to compensate for T1 changes.

Figure 4

Effect on sensitivity of removing the additional 90° pulse from excitation sculpting pulse sequence and reducing the interscan delay from 0.72 s (5T1) to 0.43 s (3T1) (green). Comparison with literature sensitivity in water. A “standardized sensitivity” was calculated to compare sensitivities measured on different spectrometers (see main text). Error bars around each point represent the standard deviation for each triplicate measurement. NH4+: 40 μM, Gd3+: 1 mM, field strength: 400 MHz.

We remeasured the sensitivity gain after addition of 1 mM Gd3+ to obtain a direct measurement of the sensitivity gain without the interference of the additional 90° pulse. Sensitivity increases of 3.9- and 3.6-fold are measured with 1 mM Gd3+ for interscan delays of 5T1 and 3T1, respectively. These values are consistent with the predicted sensitivity gain from the T1 decrease (3.9). Taking into account the corrected sensitivity gain that we calculated from Figure a (3.1), we estimate that the sensitivity can be increased by a factor of 3.5 ± 0.4 with 1 mM Gd3+, which corresponds to an order of magnitude less analysis time or several hours less ammonia accumulation to reach the detection limit. This sensitivity improvement makes fast 1H NMR NH4+ quantification accessible with a standard NMR spectrometer and reduces the cost of essential control experiments with expensive (≈500 euros/L) 15N2.

It is difficult to compare the sensitivities of two different NMR detection methods if these methods were applied using different spectrometers. The sensitivity can vary an order of magnitude because of different field strength, probe hardware, NMR tubes, postprocessing methods, etc.[18] We attempt to compare our sensitivity with the sensitivity measured by Hodgetts et al. by calculating a standardized sensitivity that takes into account the influence of field strength and type of probe (cryo- or room-temperature probe) on sensitivity (Figure ). The calculation of the standardized sensitivity can be found in the SI. As expected, with 1 mM Gd3+, the standardized sensitivity is significantly higher than the value reported by Hodgetts et al. without Gd3+.

Conclusions

In summary, the 1H NMR analysis time required to quantify NH4+ in aqueous samples can be reduced by an order of magnitude by adding 1 mM paramagnetic Gd3+. This improvement makes 1H NMR NH4+ quantification more accessible and reduces the cost of control experiments with 15N2, which enables faster, more reliable N2 reduction research. A large reduction of the T1 of NH4+ and MA without significant line broadening causes the sensitivity increase. The method has very good linearity (R2 = 0.999) and is accurate over a wide pH range if the interscan delay is increased to compensate for small T1 changes.

Materials and Methods

Materials

14NH4Cl (99.995%), 15NH4Cl (≥98 atom %, 15N ≥ 99 % CP), maleic acid (≥99%), and H2SO4 (≥97.5%) were obtained from Sigma-Aldrich. Gadolinium(III) nitrate hexahydrate (99.9%) was obtained from Fisher Scientific. DMSO-d6 (99.9% D, 0.03% V/V Tetramethylsilan) was obtained from Cambridge Isotope Laboratories. Ultrapure water was produced with a Milli-Q Advantage A10 water purification system (resistivity: 18.2 Ω at 25°C).

Sample Preparation

Ammonia standard solutions (40–500 μM) were prepared fresh daily by adding a suitable amount of NH4Cl to ultrapure water and performing serial dilutions to the required standard concentrations. In a typical experiment, 525 μL of NH4+ standard solution was mixed with 50 μL of 0.5 M H2SO4, 50 μL of DMSO-d6, 25 μL of 12.5 mM maleic acid, and 25 μL of 27 mM Gd3+ solution inside a 1.5 mL Eppendorf tube. This solution (600 μL) was transferred into a 5 mm thin-wall NMR tube (Wilmad). All NH4+ concentrations are reported as concentration in the NMR tube unless otherwise noted. The NMR tube was closed with Norell Sample Vault NMR tube caps (Sigma-Aldrich). The tube was cleaned with ultrapure water and ethanol using an NMR tube cleaner. After cleaning, the NMR tube was dried at 60°C for 1 h and stored in a dust-free environment.

1H NMR Data Acquisition and Processing

1H NMR spectra were acquired on a 400 MHz pulsed Fourier transform NMR spectrometer equipped with an autosampler. An autotunable, temperature-regulated Agilent OneNMR room-temperature probe was used for all measurements. The temperature was set to 25 °C, and the receiver gain was optimized automatically. To avoid baseline distortions and low receiver gain, the water resonance has to be suppressed by a suitable pulse sequence. Good water suppression was obtained with pulse sequences that use pulsed field gradients to dephase the water magnetization and selective pulses to flip the NH4+ magnetization back into phase during acquisition. Two pulse sequences that were preinstalled in the software of our NMR system (vNMRj) were used in this work: Excitation Sculpting (vNMRj: “waterES”) and double pulsed field gradient spin echo (vNMRj: “selexcit”). The waterES pulse sequence has the following structure:

waterES: G1-P90-G1-d1-P90-G2-S180-P180-G2-G3-S180-P180-G3-aq

where G1–G3 are the z-gradients of different strengths, P90 and P180 are hard pulses, and S180 is a selective 180° pulse. During the acquisition time, only the water resonance is out of phase, whereas the rest of the spectrum is in phase, leading to the desired suppression of the water resonance. The block “G1-P90-G1” dephases residual magnetization prior to the next scan and can be deactivated to increase sensitivity, as described in the main text. The z-gradient G1 had a duration of 1.6 ms and a strength of 1.07 G cm–1. The z-gradients G2 and G3 had a duration of 1 ms and a strength of 1.7 G cm–1. The 180° selective pulses had the shape “Wsupp” with a width of 2.5 ms and a power of 13 dB. The selexcit pulse sequence has the following structure:

selexcit: P90-G1-S180-G1-G2-S180-G2-aq

where G1 and G2 are the z-gradients of different strengths, P90 and P180 are hard pulses, and S180 is a selective 180° pulse. During the acquisition time, only the region defined by the selective 180° pulse is in phase, whereas the rest of the spectrum is out of phase. The z-gradients G1 and G2 had strengths of 0.85 and 1.28 G cm–1, respectively, and a duration of 1 ms. The selective 180° pulse was defined as a “q3” pulse shape with a width of 5 ms and a power of 0 dB. The position and width of the selective pulse in the frequency domain were set to 6.63 ppm and 540 Hz, respectively, so that the pulse is positioned between the resonances of NH4+ and maleic acid. The pulse shapes q3 and “Wsupp” that were used to create the shaped pulses in waterES and selexcit are standard pulse shapes available in the software package vNMRj. Equivalent pulse shapes should be available in other software packages.

The data were processed in the software package MestReNova (version: 12.0.1-20560) using the automated tools provided in this software. Unless otherwise noted, an apodization of 4 Hz was applied followed by phasing and baseline correction. The peaks of NH4+ (t, ≈6.9 ppm, 4H) and MA (s, ≈6.21 ppm, 2H) were integrated using the line fitting tool. Using the line fitting tool instead of directly integrating the peaks leads to an approximately 2-fold decrease of the relative error. The three integrals of the NH4+ peaks were added together to calculate the total NH4+ integral. From the ratio of the integral of NH4+ and MA, the concentration of NH4+ was calculated with absolute quantification according to eq . The linewidth of NH4+ is calculated by averaging the full width at half-maximum (FWHM) of the three NH4+ peaks. The signal-to-noise ratio (SNR) was calculated using the “SNR calculation” tool in MestReNova with the noise region defined from 11 to 13 ppm. The SNR values were calculated by averaging three measurements of the average SNR of the three peaks of the NH4+ triplet. The relative error was calculated according towhere ccalcd and cgrav are the concentrations of NH4+ calculated from absolute quantification and from the weight and purity of the NH4Cl that was added to prepare the standards, respectively.

The T1 values of NH4+ and MA were measured using the ES pulse sequence with default setting. Spectra were acquired at six different recycle delays, and the function y(x) = a*(1 – exp(−bx)) was fitted to the integrated peak intensities of NH4+ and MA as a function of d1 using the software OriginPro 2015. Subsequently, the parameter b from the fitting function was inversed to calculate T1. An example of the T1 determination using this method can be found in the SI.