Optimized Procedure for Determining the Adsorption of Phosphonates onto Granular Ferric Hydroxide using a Miniaturized Phosphorus Determination Method

Example of isotherms gained with the proposed procedure
Figure 4 shows an example of the results gained when applying the protocol in the case of the investigation of the adsorption of NTMP by GFH at different pH values. NTMP was selected because, with three phosphonate groups, it is the most representative phosphonate for the broad spectrum of possible phosphonates of which the number of phosphonate groups vary between one (PBTC) and five (DTPMP). Furthermore, the molar mass of NTMP (299.05 g/mol) is also located in the middle range of phosphonates (HEDP: 206.03 g/mol, DTPMP: 573.20 g/mol). In Figure 4, the adsorption isotherms, i.e., the loading of the phosphonate above the residual phosphonate concentration, are depicted at different buffers and pH values after a contact time of 1 h. Longer contact times could lead to undesirable abrasion of the material due to too long contact between the particles. For each isotherm, a solution with 1 mg/L NTMP-P and, depending on the desired pH range, buffer in the concentration of 0.01 M was prepared and adjusted to an initial pH value by means of HCl or NaOH. This was 4.0 (AcOH), 6.0 (MES), 8.0 (EPPS), 10.0 (CAPS) and 12.0 (NaOH). Depending on the GFH concentration, as a result of the 1 h contact time, the pH value in the solution changed by a maximum of 2.0: 4.0-6.0 (AcOH), 6.0-7.3 (MES), 8.0-8.2 (EPPS), 9.4-10.0 (CAPS), 10.9-12.0 (NaOH). The PZC of GFH is approx. 8.6, so it is consequential that the pH value in the case of a set pH value > 8.6 decreased due to contact with GFH and increased at a pH value < 8.6. The further away this adjusted pH value was from 8.6, the stronger the pH change was.

Figure 4: Loading of NTMP (initial concentration of 1 mg/L NTMP-P) onto granular ferric hydroxide dosed at concentrations of 0.7-14 g/L after 1 h contact time at room temperature. The following buffers at concentrations of 0.01 mol/L were used at the mentioned pH values in the graph: AcOH (pH 4.0-6.0), MES (pH 6.0-7.3), EPPS (pH 8.0-8.2), CAPS (pH 9.4-10.0) and NaOH (pH 10.9-12.0). The curves plotted are Freundlich isotherms. Please click here to view a larger version of this figure.

All isotherms in Figure 4 were modeled using the Freundlich equation (R² values from left to right with increasing pH: 0.875, 0.905, 0.890, 0.986, 0.952; corresponding n values: 2.488, 3.067, 4.440, 2.824, 1.942; corresponding K_F values: 0.619, 0.384, 0.260, 0.245, 0.141). At pH values of 4-6, a loading of up to 0.55 mg NTMP-P/g was achieved, which corresponds to 1.8 mg NTMP/g. The higher the pH value, the lower the level of adsorption. Iron hydroxides have a large number of Fe-OH groups on their surface, which may be protonated or deprotonated depending on the pH value. With the depth of the pH value, the surface is predominantly protonated, i.e., positively charged, which means that the multidentate phosphonates, which are negatively charged over the almost entire pH range, are attracted. A higher pH value shifts the charge of the iron hydroxide surface in the negative direction, which in turn leads to increased electrostatic repulsion⁷. Interestingly, even at pH 12, which corresponds to an OH^– concentration of 0.01 M, adsorption occurred. Therefore, for successful desorption, NaOH solutions with a much higher concentration must be used.

In comparison to the results of other researchers, the maximum loading of up to 0.55 mg NTMP-P/g of GFH in this work seems to be rather low. Boels et al.¹⁴ found a maximum loading of 71 mg NTMP/g of GFH, which corresponds to 21.7 mg of NTMP-P/g GFH in their experiments with a synthetic reverse osmosis concentrate with 30 mg/L NTMP (9.3 mg/L NTMP-P) at pH 7.85. They used powdered GFH and stirred the synthetic solution, which contained HCO₃^– that also acts as a buffer, for 24 h. Therefore, their results cannot be directly compared to the findings of this work, as they used a much higher initial concentration and powdered GFH, which is likely to lead to a higher surface area and, therefore, results in a better adsorption performance. Additionally, the contact time was significantly longer as in this work. Nowack and Stone⁷ conducted experiments with a 40 µM NTMP solution (3.72 mg of NTMP-P/L) in a 0.42 g/L goethite slurry at a pH of 7.2. The solution was stirred for 2 h leading to a maximum loading of approx. 30 µM NTMP/g goethite (2.79 mg of NTMP-P/g). 1 mM MOPS was used as a buffer. Again, the results cannot be compared directly to the results of this work due to the higher initial phosphonate concentration. In addition, the slurry, which consisted of goethite flocks, had a high surface area. However, the shapes of the isotherms from Boels et al.¹⁴ and Nowack and Stone⁷ agree with the ones of this work, and all of them could be fitted well by the Freundlich model.

Influence of buffer on phosphonate adsorption and required buffer concentration
Previous experiments to determine the adsorption kinetics had shown that also with the use of buffers, an equilibrium pH value is reached within a very short period of time. That pH can deviate significantly from the pH value that was previously set in the phosphonate-containing solution (adjusted pH). This equilibrium pH tends to the PZC of the filter material, which was 8.6 for the granular ferric hydroxide discussed here (according to own investigations). Therefore, it can be assumed that the pH value after the contact time (final pH) is decisive for the extent to which the adsorption of the phosphonate occurs.

Figure 5: Left: loading of NTMP (initial concentration of 1 mg/L NTMP-P) onto 2.5 g/L granular ferric hydroxide as a function of the pH value at different buffer concentrations after a contact time of 1 h. Right: Comparison of the pH value after 1 h contact time with the pH value set in the stock solution before contact with the granular ferric hydroxide at different concentrations of the buffers AcOH, MES, MOPS, EPPS, CAPSO and CAPS. Please click here to view a larger version of this figure.

In the right-hand diagram in Figure 5, the pH values that were set in the NTMP-containing solution at different buffer concentrations are compared with the final pH values after the 1 h contact between 1 mg/L NTMP-P and 2.5 g/L GFH. It becomes evident that a specific correlation between the pH value previously set in the solution and the final pH value was only attainable and thus a relatively reliable pH adjustment was possible only when buffers in concentrations of 10 mM were used. This is reflected in the correlation function determined by means of polynomial regression and reproduced in the diagram on the right. The fact that in the case of buffer concentrations below 10 mM pH values of 2-4 had to be preset in order to obtain final pH values of 6-7 shows that the prediction of the final pH value, which is decisive for adsorption, and thus the safe execution of the adsorption tests for such buffer concentrations were challenging.

In the left-hand diagram in Figure 5, the extent of adsorption of 1 mg/L NTMP-P at 2.5 g/L GFH is depicted as a function of the final pH value for different buffer concentrations. Assuming a linear dependence of the loading on the pH value in the pH range 4-12 according to equation y = ax + b, the values calculated by linear regression for all buffer concentrations investigated were very similar (10 mM: a=−0.0673, b=1.0914, R²=0.9837; 6.6 mM: a=−0.0689, b=1.1047, R²=0.9512; 3.3 mM: a=−0.0672, b=-0.0672, R²=0.9570; 0 mM: a=−0.0708, b=1.157, R²=0.8933). The coefficient of determination, which was the highest for 10 mM buffer, showed very clearly that with this buffer concentration not only the final pH value was easier to adjust, but also the most reliable results with regard to adsorption were achieved. Only the course without buffer indicates possible deviations of the adsorption extent between pH 5 and 7. However, in order to achieve these final pH values without buffering, very low pH values had to be set in the stock solution, some of which were only slightly above 2. Due to the very strong difference between adjusted pH and final pH, it is, therefore, possible that the final pH value was not decisive for the extent of the adsorption in the case of no buffer. It can thus be assumed that the use of Good buffers mentioned in Table 1 has no significant influence on the adsorption of phosphonates onto GFH, i.e., there is no competition for adsorption sites between the phosphonate and the buffer. Such selectivity is only prevalent because the adsorption of NTMP onto GFH is mainly due to the formation of mono- and bidentate complexes¹⁵. Good buffers, on the other hand, have little tendency to form metal complexes^17,19, which is why NTMP is preferably bound by GFH. In the case of adsorbents with a less polar surface, such as activated carbon, it can be assumed that Good buffers also occupy free adsorption sites and thus influence the adsorption of the phosphonate. The use of these buffers to study the adsorption of phosphonates on activated carbon is therefore not recommended.

Calibration of ISO_mini method and compliance with ISO
Figure 6 shows the calibration lines using the internal quality standard (IQS: 1 mg/L KH₂PO₄-P in 0.9 mM H₂SO₄) according to ISO 6878 as well as the modified ISO_mini method for total P and o-PO₄^3--P determination. Based on a linear regression, the calibration function equivalent to ISO 6878 was y = 0.0033 + 0.2833x (R²=0.99978). The linear regression applied to the miniaturized variant for phosphate determination resulted in the calibration function y = 0.0058 + 0.2864x (R²=0.99999). With y = 0.0020 + 0.2890x (R²=0.99985) the calibration function for total P determination according to the ISO_mini method was very similar and very precise as well. All variants had a very high coefficient of determination, which means that the ISO_mini method does not compromise accuracy by the reduction of the sample volume to one-fifth. The conversion equation determined by means of the calibration functions for determining the P concentration in the analysis sample from the measured spectral absorbances is given in the protocol in step 4.15. Experience has shown that the absorbance of the blind sample can usually be neglected since at 880 nm the signal emitted by the photometer can jump very strongly in the very small measuring range. Thus, a measured value of 0.287 at 4 mL sample volume (ISO_mini) corresponded to a phosphorus concentration of 1 mg/L P.

Figure 6: Calibration lines for the determination of total P and ortho-phosphate-P according to ISO 6878 and ISO_mini. An IQS (1 mg/L KH₂PO₄-P in 0.9 mM H₂SO₄) was used in accordance with point 1.9 of the protocol. For the ISO method, the IQS was used in aliquots of 4, 8, 12, 16 and 20 mL and for the modified ISO_mini method in aliquots of 0.8, 1.6, 2.4, 3.2 and 4.0 mL. Please click here to view a larger version of this figure.

Plausibility and buffer-dependent dosage quantities of ISO_mini method
As already mentioned, a reliable pH adjustment in the adsorption test is only possible with a buffer concentration of 0.01 M. However, such a buffer concentration requires a higher K₂S₂O₈ dosage than specified in ISO 6878 for most buffers. In addition, the ISO stipulates that the pH value must be set to 3-10 using a pH probe after digestion. Since such a pH adjustment cannot be carried out in a small screw cap vial, the matching NaOH dosage quantity for different buffer solutions had to be determined. Figure 7 shows the absorbance of different buffer-containing solutions with 1 mg/L NTMP-P when these were digested with different K₂S₂O₈ quantities according to ISO_mini and treated with varying amounts of NaOH after digestion. Accordingly, each matrix was based on the following procedure: 4 mL of a solution was mixed with 0.2 mL 0.9 M H₂SO₄, provided with different K₂S₂O₈ quantities and filled up with H₂O to the same total volume of 9 mL. This was now digested in accordance with the protocol (1 h at 148-150 °C). After cooling, different NaOH quantities were added and filled up to a total volume of 9.4 mL with H₂O. Subsequently, 0.2 mL of ascorbic acid solution and 0.4 mL of molybdate II solution were added. The determination of the absorbance (880 nm) was carried out 4 h after the addition of these color reagents. This time was chosen to ensure that the specific absorbance was stable. A solution with 1 mg/L NTMP-P and 1 M NaOH was also investigated. However, instead of the K₂S₂O₈ and NaOH amounts, the H₂SO₄ amounts were varied to ensure that the pH was low enough for digestion. The targeted absorbance value was 0.287 (see calibration line in Figure 6). Thus, in Figure 7 those values are shown in light green that deviated from this target value by a maximum of 5%. One value in each matrix is highlighted with a dark green color. This marks the K₂S₂O₈ and NaOH dosage quantities recommended for the regular ISO_mini method for this type of buffer solution.

Figure 7: Spectral absorbance (×1000) of different phosphonate- and buffer-containing solutions with different K₂S₂O₈ and NaOH dosage quantities at a wavelength of 880 nm in 1 cm cuvettes. Procedure: 4 mL solution (as shown in the figure and adjusted to the pK_a value of the buffer adapted from the thermodynamic pK_a values of Goldberg et al.²⁰ to a concentration of 0.01 M and 25 °C³¹) was placed in a 10 mL screw cap vial, mixed with 0.2 mL of 0.9 M H₂SO₄ and with different amounts of K₂S₂O₈ (as shown in the figure). Water was then added to obtain a total volume of 9 mL for all samples before digestion. Now the vials were heated in the thermostat at 148-150 °C for 1 h (digestion). After cooling to room temperature, different amounts of NaOH (as shown in the figure) were added and with the addition of water, it was ensured that a total volume of 9.4 mL was present in all vials. 4 h after addition of 0.2 mL of ascorbic acid solution and 0.4 mL of molybdate II solution, the absorbance at 880 nm was determined. In the case of solution l (1 mg/L NTMP-P in 1 M NaOH), the amount of H₂SO₄ was varied instead of K₂S₂O₈. Here, the dosed amount of NaOH in all samples corresponded to 0.4 mL of 1.5 M NaOH, i.e., 0.60 mmol of NaOH. Light green: maximum 5% deviation from target value: 287. Dark green: the recommended setting for this buffer- and phosphonate-containing solution. Dashed line: COD, straight line: ThOD. Please click here to view a larger version of this figure.

Although reductive conditions must prevail in the color formation process and excessive K₂S₂O₈ may interfere with this, the results for solutions a and b (Figure 7), for which no (IQS) or only a very small quantity of K₂S₂O₈ (only NTMP without buffer) is required, show that higher quantities of K₂S₂O₈ than required do not automatically lead to an abrupt reduction of the absorbance. It should also be mentioned here that other phosphonates in solutions analogous to solution b with 1 mg/L PBTC-P (absorbance: 0.3005), 1 mg/L HEDP-P (0.3035), 1 mg/L EDTMP-P (0.2952) or 1 mg/L DTPMP-P (0.2936) were digested entirely using the ISO_mini method according to the protocol with 0.04 g K₂S₂O₈ and 0.6 mmol NaOH. Thus, this method can also be used for phosphonates other than NTMP.

Table 1 shows the theoretical oxygen demand (ThOD) for the oxidation of each buffer and the chemical oxygen demand (COD) measured in a 0.01 M buffer solution by Hach LCK 514 cuvette rapid tests. It is known that potassium dichromate, the oxidant used for the COD determination, does not oxidize organically bound nitrogen³². For Good buffers, the measured COD was always between the theoretical amount for the oxidation of C and H and the oxidation of C, H and S. Only for buffers with a C-OH group (HEPES, EPPS, CAPSO) the measured value did correspond to the theoretical value for oxidation of C, H and S. In buffers that do not contain a C-OH group (MES, MOPS, CAPS), the sulfo group is obviously not degraded completely to sulfate.

For the solutions 7c に 7j, it can be seen very clearly that K₂S₂O₈ quantities significantly below the amount of oxidizing agent required according to the COD of the buffer, independently of the NaOH amount, did not contribute to the achievement of the target value. At 10 mM, the buffer in these solutions had a concentration of approx. 1000 times higher than that of NTMP. If the buffer is not digested, it cannot be guaranteed that the phosphonate can be completely oxidized. Only K₂S₂O₈ quantities beyond the COD contributed to the reliable attainment of the target value. Thus, it was not necessary for all buffers to apply the theoretical oxidant requirement for the complete oxidation of the buffer (ThOD) because the nitrogen and obviously also for some buffers, the sulfo groups were not completely decomposed. Any oxidizing agent beyond the COD did not react with the buffer, and, therefore, there was sufficient excess of K₂S₂O₈ to oxidize the phosphonate. NTMP also contains nitrogen. Although this may not be completely oxidized to nitrate, all phosphonate groups are obviously oxidized to phosphate. Otherwise, one would not find the absorbance that is present for 1 mg/L P. Abundant excess of K₂S₂O₈ did certainly also contribute to the complete oxidation of the phosphonate, but after the digestion some K₂S₂O₈ was still present and could react with ascorbic acid, which is necessary for the reduction of the blue molybdate-phosphate complex. The result was an absorbance lower than the target value.

In each row, the absorbance increased with the amount of NaOH starting from a certain amount of NaOH. Thus, it also occurred that below the amount of oxidizing agent required according to the COD of the buffer, the measured absorbance value could be in accordance with the target value, although NTMP was obviously not completely digested (see solutions 7c, 7f, and 7h). In this case, the increase in absorbance was due to self-reduction of the molybdate ion due to a too low [H⁺]:[Mo] ratio²⁶, and any correspondence is therefore only random. Accordingly, with higher K₂S₂O₈ quantities, more NaOH could be used after digestion, as K₂S₂O₈ reduces the pH value.

In most solutions, the absorbance was also in accordance with the target value even if no NaOH dosing was applied. Occasionally, however, deviations from this value occurred, which may be because the absence of NaOH resulted in the fact that the optimum [H⁺]:[Mo] ratio was not maintained and thus the color complex became unstable. Therefore, irrespective of the analysis solution, a dosage of 0.6 mmol NaOH is recommended, as, thereby, the color complexes proved to be the most stable. Regeneration solutions often have a concentration of 1 M NaOH. One such case is covered by matrix l. Here, it was shown that only a very narrow spectrum of H₂SO₄ dosage is permissible, proving that the use of a pH probe to adjust the pH after digestion may be a safer procedure here.

All dark green absorbance values in Figure 7 (n=12), converted into the total P concentration according to the calibration line in Figure 6, give an average value of 1.013 mg/L. The standard deviation is 0.014 mg/L. The typical deviation from the target value (1.000 mg/L) is therefore only 0.11-2.67% ((1.013–0.014–1.000) / 1.000 × 100% = 0.11%; (1.013 + 0.014–1.000) / 1.000 × 100% = 2.67%). This shows a high accuracy of the ISO_mini method.