3.2. Determination of Mineral Precipitation Likelihood
Calculation of probable mineral precipitates’ SI’s by chemical equilibrium modeling is performed in a wider pH value range than foreseen for the actual planned experiments (pH from 8 to 9.5, at a 0.25 step increment), to evaluate the index variability.
Table 3 shows the calculated SI values for all the possible P-containing solid phases and examined conditions. Struvite shows positive SI at all pH levels, in accordance with the literature [
21]. Out of all possible calcium phosphate compounds, ACP and HAP also show high positive SI at all pH values. Dicalcium phosphate dihydrate (DCPD) (CaHPO
4·2H
2O) SI is positive in the pH range 8–8.25, negative afterwards, indicating unlikelihood of precipitation at high pH values. Dicalcium phosphate (DCP) (CaHPO
4) shows positive SI up to pH = 9.
Based on these simulated results, the likelihood of P mineral precipitation in the selected test conditions (pH = 8.5 and 9.0) is assumed to be high for HAP, ACP, bobierrite, magnesite, calcite, struvite and DCP (in that order), and unlikely for the remaining compounds. It is important to mention that high calculated SI values do not necessarily mean that higher precipitation will actually occur: this, in fact, is highly dependent on actual operating conditions, and needs confirmation by experimental and analytical methods.
These findings are in line with those of other studies [
27,
44]. Struvite, the mainstream P-recovery mineral form, shows lower precipitation likelihood than other “useful” P forms; struvite analogues K-struvite (MgKPO
4·6H
2O) and Na-struvite (MgNaPO
4·7H
2O) indicate no theoretical possibility of precipitation, due to negative calculated SI values. These considerations will be of importance in the subsequent interpretation of experimental results.
3.3. Synthetic Wastewater P-Recovery Tests
Each experiment was conducted under continuous flow for a duration of four reactor’s HRTs.
Figure 4a, shows cumulative NaOH aliquots (mmols) added to the reactor for pH control in experiments 1–6. In experiments 3–4 (HRT = 60 min) the cumulate line stabilizes approximately after three HRTs, indicating that the system has reached steady state equilibrium as far as reactions (5–7) are concerned. In experiments 1 and 2 (HRT = 30 min), however, NaOH addition by the PID system continues well past 3 HRTs, indicating that 30 min HRT is not sufficient for all the above reactions to complete. Comparing the total amount of NaOH required for process completion after 3 HRTs (
Figure 4b) between experiments 1, 2 (HRT = 30 min) and 3, 4 (HRT = 60 min) a noticeable difference can be seen. This is visible to a lesser degree in experiments 5, 6 (HRT = 120 min). Higher NaOH consumption at lower HRT means the precipitation reactions (5–7) have not reached equilibrium, yet. The difference is negligible in experiments 5, 6 (HRT = 120 min), and more clearly visible at pH = 9.0, indicating that a greater degree of reaction completion is reached at higher HRTs. In general, with the exception of experiment 6 (HRT = 120 min, pH = 9.0), the process is accelerated by higher pH levels (
Figure 4a). Based on these results, it may be concluded that HRT of around 60 min should be sufficient for all precipitation to be completed.
The most important effect of pH is observed on overall P removal yields.
Figure 5 shows solute phosphorus concentration vs. time for all experiments: concentration decreases rapidly at the onset of the reaction, and then reaches a nearly stable value, indicating that P removal is mainly related to the system’s pH, and not much dependent on the elapsed time of the reaction (or reactor’s HRT). This indicates that P is removed rapidly through crystallization reactions, but afterwards it takes time for precipitates to grow in size and settle. This can be seen both in experiments 1–6 (
Figure 5a) and 7–9 (
Figure 5b), the former under addition of stoichiometric chemicals (Mg and ammonia salts) addition, the latter without.
With Mg and ammonia salts addition (for struvite formation), P removal is higher at pH = 9 (
Figure 5a), but this requires higher NaOH dosage to the system, which is expensive. On average, 61.2% and 90.4% P removal were achieved for pH = 8.5 and 9.0, respectively, in experiments under controlled Mg:NH
4:P = 5:5:1 solute molar ratio. Without the chemicals addition, 82.6% removal was achieved using NaOH to control pH at 9.0 (
Figure 5b). Using Ca(OH)
2 as alternative pH buffer at pH = 8.5 and 9.0, without addition of Mg and ammonia salts (with final precipitates other than struvite), removals of 75.3% and 87.2% were obtained, respectively. This may be a significant advantage in full scale applications of the process, since the cost of Ca(OH)
2 is much lower (by about tenfold) than that of NaOH. Oscillations of P concentration that can be seen in some cases during experiments, for example at around 120 min in the experiment with Ca(OH)
2 at pH = 8.5 and RT = 60, could be attributed to the dissolution of some of the phosphate minerals and the consequent P release, in addition to occasional measurement error.
A first-order kinetic model was fitted to obtain the constant rates of the P removal reaction (Equations (9) and (10)) in experiments 1–6:
where C
0 is the initial P concentration, C
eq is the P concentration at equilibrium (steady state) condition, C is the P concentration at time t, and k the reaction’s kinetic constant.
Results of kinetic model fitting show the effect of HRT and pH on P removal rates (
Table 4). These increase with HRT and, for the same HRT, with pH, as resulting from the slope of the plots of ln(C − C
eq) versus time (
Figure 6). This effect is due to the nucleation rate increase with pH increase, causing more surface area to be available for crystal growth [
17]. This analysis confirms that the optimal process conditions with NaOH buffer and chemical addition are HRT = 60 min and pH = 9.0, when the fastest removal is achieved.
3.4. Precipitates Analysis
Precipitates collected from all synthetic solution experiments were subject to full characterization with FTIR, TGA, XRD, ICP and elemental analysis.
FTIR showed similar spectra for all experiments (
Figure 7a, shows those for experiment 4). Contrary to initial assumptions, but in accordance with subsequent chemical modeling, FTIR analysis gave no direct evidence of the presence of struvite. Correlated peaks of phosphate groups and H-bonds [
45] were visible around 1000–1100 cm
−1 and 3400–3500 cm
−1, respectively. The latter peak is related to the presence of water in the precipitates, while observed peaks in the range 1200 to 2000 cm
−1 could not be ascribed with certainty to this mineral. Two peaks at around 1400–1500 cm
−1 could be related either to the H-N-H bond in struvite or to the carbonate group resulting from precipitation of calcite, while the other peak at around 1650 cm
−1 could be associated to ACP. The peak at around 900 cm
−1 is typical of the HPO
42− group but could also be related to carbonates [
46].
This result is not in accordance with previous lab-scale findings, in which direct evidence of struvite precipitation was observed for experiments at Mg:NH
4:P = 5:5:1 molar ratio from real wastewater [
24]. A possible explanation could be the presence of ACP precipitation inhibitors in real wastewater liquor, not contained in the synthetic solution [
47]. In addition, previous experiments [
24] were carried out in batch mode, at much longer HRT (18–24 h), giving struvite ample time to grow and precipitate in significant amounts, contrary to the present continuous flow, low HRT experiments, more closely resembling actual operating conditions.
XRD precipitate analyses (
Figure 7b, relative to experiments 3, 4, 7 and 9) show a similar, specific pattern with a broad peak around 30°, associated to the amorphous phase in ACP [
48]. Specific peaks of struvite are not visible in XRD patterns. Furthermore, TGA (
Figure 7c, shown for experiment 4) of precipitates shows a weight loss at around 150 °C, corresponding, on average, to 18–22% of what could be attributed to reversible removal of adsorbed water molecules. This could also be related, however, to the loss of NH
3 caused by the presence of struvite in precipitates [
49]. Two minor weight losses (2–4% and 3–5%, respectively) are visible at 300–500 °C and 700–800 °C. The latter could be mainly due to presence of calcite (Equation (11)), the former could be related to irreversible water removal, confirming CaHPO
4 precipitation (Equation (12)) [
49,
50].
ICP and elemental analysis reveal the ionic composition of obtained precipitates. Based on their C, N and P content, the amounts of present PO
43−, NH
4+, CO
32− groups can be calculated (
Table 5), considering that these are the only possible ionic forms containing those elements in these conditions. Water content is calculated considering total hydrogen content and subtracting the content of NH
4+. Results show low error levels for total precipitate estimates in most experiments.
Since struvite is the only possible precipitate compound containing NH4+, any amount of this ion found must correspond strictly to this mineral form. Based on stoichiometric ratios observed in struvite (Mg:NH4:PO4 = 1:1:1), equivalent amounts of these moles are needed; furthermore, there are 6 moles of water in a mole of struvite that will be subtracted from total water content. Similar chemical mass balance procedure was followed for the other groups. CO32− content, for example, can be related to both calcite and magnesite; however, magnesite, notwithstanding the high calculated SI value, precipitates significantly only at pH > 9.5. Therefore, CO32− in these conditions could be associated with, and hence is subtracted from, the Ca2+ precipitate.
The case for the PO
43− group, however, is more complicated. The model suggests that different calcium phosphate compounds such as ACP, DCP and HAP, as well as Mg-based compounds such as bobierrite, could precipitate. The detected amount of Ca
2+, after subtracting the fraction associated to calcite, can therefore be related to calcium phosphate compounds. Precipitation of HAP is kinetically unlikely under the chosen operating conditions, despite having positive SI, since ACP and DCP forms precipitate first, and their transformation to the HAP form is very slow [
27]. In fact, in FTIR spectra, the typical sharp peak of OH
− associated to HAP, expected at around 3500 cm
−1, is not visible in any of the precipitates, suggesting the absence of this mineral. FTIR, on the other hand, confirmed the presence of the HPO
42− group, with its associated peak at around 900 cm
−1 quite evident in the observed spectrum (
Figure 7a) confirming the presence of DCP precipitates.
At this point, it is difficult to determine how much of the remaining Ca2+ could be related to DCP, and how much to ACP. An estimation may be attempted by subtracting the remaining Ca2+ related to Ca-P forms from total PO43− content, assuming that all of it is related to ACP, in Ca3(PO4)2 form. This assumption will be further discussed later on.
Finally, the remaining Mg
2+ content, after accounting for struvite formation, must necessarily be related to bobierrite (the only other Mg-P compound with positive SI according to chemical calculations). Taking into consideration bobierrite’s formula Mg
3(PO
4)
2·8H
2O, and stoichiometrically subtracting any Mg
2+ content from total phosphate and water contents, the remaining precipitates must contain only phosphate and water. A residual PO
43− content confirms the presence of the HPO
42− group, and therefore of DCP precipitation. This invalidates the previous assumption considering only ACP formation from Ca
2+. Considering this PO
43− amount to be in DCP form, and consequently adjusting the Ca
2+ balance, all precipitates’ compositions can finally be calculated.
Figure 8 shows final ionic ratios and composition of final precipitates for all experiments under NaOH addition (tests 1–6).
Figure 8a shows that at pH = 8.5 the Ca/P ratio decreases with increasing HRT, while Mg/P and Mg/Ca ratios increase. The same occurs at pH = 9.0, but to a lesser extent (
Figure 8b).
Figure 8c,d shows the composition of final precipitates based on the mass balance calculations for experiments 1–6. At pH = 8.5 (
Figure 8c), increase in struvite and a slight decrease in ACP content occurs with HRT increase. The explanation could be due to a precipitation lag of struvite compared to calcium phosphates, caused by slower reaction kinetics [
23,
45]. Clearly, 30-min HRT is not sufficient for significant struvite precipitation, as previously observed. At pH = 9.0 (
Figure 8d), however, this pattern is not so evident.
The possible reason for this could be that, at the lower pH, ionic concentrations in solution are still relatively high because of lower SI values (
Table 3), leading to shorter times needed to reach equilibrium. Moreover, moving from pH = 8.5 to 9.0, SI values for Ca-P forms increase considerably more than struvite’s, leading to lower precipitation of the latter compared to the former. In all cases, the main fraction of precipitate consists of ACP and DCP. Studies suggest that, in alkaline conditions, HPO
42− could be incorporated within ACP, and would not exist as a separate DCP solid phase [
51]. In addition, the SI value of DCP is much lower than ACP’s, and consequently a considerable amount of Ca-P precipitation could in fact occur as ACP [
44].
Figure 9 shows molar ratios and composition of final precipitates for the experiments conducted without chemical addition (7, 8 and 9), only using NaOH or Ca(OH)
2 as pH buffers, at HRT = 60 min. Results show significantly limited struvite precipitation due to low Mg/Ca ratios in solution in all conditions, in line with other studies [
52,
53]. At pH = 9.0, addition of Ca(OH)
2 leads to higher calcite precipitation, a situation not favorable for the final composition of the precipitates. However, with the same buffer agent at pH = 8.5, ACP formation increases both compared to the previous case and to NaOH-driven conditions, without increasing or decreasing calcite precipitation. In addition, bobierrite precipitation is slightly higher at pH = 8.5 compared to pH = 9.0 when using Ca(OH)
2. It can therefore be concluded that the use of Ca(OH)
2 to control pH at 8.5 yields a more favorable combination of precipitates, in term of useful P recovery, compared to the use of NaOH or Ca(OH)
2 at pH = 9.0.