Phosphorus Sequestration and Recovery with Brushite: Validating Chemical Equilibrium and Process Models with Case Studies

Abstract
The increasing prevalence of tighter effluent phosphorus limits, enhanced biological phosphorus removal, and anaerobic digestion pose a unique challenge of phosphorus release and nuisance struvite formation in solids handling processes. Phosphorus sequestration through the addition of calcium represents an alternative approach to phosphorus management in solids processes. Phosphorus equilibrium chemistry is complex, depending on the pH and the concentration of other constituents, and a variety of calcium phosphate precipitates can be formed. To provide Water Resource Recovery Facilities (WRRFs) with an improved understanding of the main calcium phosphate precipitates for recovery, this study applied a chemical equilibrium model using Visual MINTEQ to predict the main precipitates from anaerobic digester (AD) effluent. Designed systematic parameters were used as inputs into the model for existing operational conditions for three WRRFs and for three samples with different molar ratios of Ca/P. Visual MINTEQ was compared with BioWin 6.2 and with X-ray powder diffraction (XRD) analysis for phase identification and quantification. XRD analysis corresponded well with BioWin model outputs in terms of precipitates formation. Our results indicate that BioWin can be reliably used to determine the precipitates in AD samples for calcium phosphate.

Novelty and Relevance
This study applies a chemical equilibrium model for brushite precipitation in AD effluent and compares the results with BioWin and XRD analysis. Based on literature review, this is the first paper that investigates the applicability of Visual MINTEQ for calcium phosphate precipitation in wastewater and a side-by-side comparison of Visual MINTEQ with a wastewater simulator such as BioWin. To date, technology provider inputs have been used for estimating the calcium dose to sidestreams for brushite recovery. This study attempts at verifying results from BioWin and Visual MINTEQ as additional tools to estimate Ca dosing and brushite recovery using full scale data. Introduction Visual MINTEQ, a chemical equilibrium model, has been used to study struvite precipitation using bench scale testing and/or synthetic wastewater (Buchanan et al., 1994; Ali et al., 2003; Wu and Zhou, 2012; Jia et al. 2017; Celen et al. 2007; Miles and Ellis, 2001). Researchers point out that the order of precipitation is determined by kinetic factors and the importance of precipitation kinetics was stressed in many studies (Koutsoukos et al., 2001; Zawacki et al., 1989; Romero-Güiza et al. 2015; da Rocha et al., 2018). Biological wastewater simulators such as BioWin formulates precipitation processes with kinetics equations. In this study, an attempt is made to compare the predictions of BioWin and Visual MINTEQ with XRD analysis using full-scale data.

Methodology and Materials
Samples and XRD Analysis Four AD samples for City of Longmont and two Post Aerobic Digester (PAD) samples from City of Boulder WRRF and Metro Water Recovery NTP were obtained for analytical measurements and XRD analysis. The chemical characteristics of the sludge samples are shown in Table 1. Sample 2 was used for bench scale testing at Ca/P molar ratios of 1.5, 2.5, and 3.5 with addition of 20% CaCl2.

Visual MINTEQ Set up
Visual MINTEQ v3.1 (KTH Vetenskap Och Konst, Stockholm, Sweden) was used to predict the precipitates formed in samples using a constant pH approach at 35°C. Details of model set-up and revisions are described in full paper. Precipitates employed in the model include struvite, bobierrite, newberyite, monetite, brushite, monetite, MgCO3, calcite, vivianite, strengite, and ferrous sulfide.

Experimental Set up
Bench scale experiments were performed using a sample of AD (City of Longmont WRRF) at room temperature. CaCl2 was added, samples were stirred at 200r/min using a magnetic stirrer, and pH was measured after 45 minutes. Photos 1 and 2 show the samples for Conditions 2 and 5 after mixing has stopped. There was no clear visual sign for precipitants formed. Final pH for sample conditions 3, 4, and 5 were 7.2, 7.05, and 6.9, respectively. Samples were sent for XRD analysis and Ortho-P measurements.

BioWin Modeling
A whole plant BioWin model was set up in BioWin 6.2 for Condition 1 for City of Boulder WRRF (Figure 1) and Conditions 2-5 for City of Longmont WRRF (Figure 2 and 3).

Results and Discussion
Testing the Validity of Model Results In order to validate the results and confirm the presence of minerals predicted by the two models, XRD analysis was used. The quantitative crystalline phase analysis for conditions 1 to 6 is shown in Table 2. The observations made based on XRD analysis are presented in Table 3. Figure 4 shows the XRD pattern for condition 1. XRD results for conditions 2 to 5 are presented in Figures 5 to 8, respectively. Figure 9 shows the XRD pattern for condition 6.

Visual MINTEQ Results
Table 4 shows the model output for major precipitates with positive saturation index which indicates supersaturation as well as the solids in finite state (in equilibrium). The observations made based on Visual MINTEQ are presented in Table 5. The Ortho-P reduction, as predicted by Visual MINTEQ, is compared with actual data in Table 6 for Conditions 3 to 5. Visual MINTEQ underpredicted the Ortho-P removal. Visual MINTEQ estimated that the Ortho-P reduction was linked to precipitation of calcium phosphate. The % of total phosphate in dissolved form decreased from 56% for condition 2 to 36% with increasing Ca:P ratios from 1.5 to 3.5. While there are no studies available for calcium phosphate precipitation in AD, studies on brushite precipitation in CalPrexTM reactor shows high removal efficiency of 80-95% at Ca/P molar ratios of 1.2-1.4 based on pilot studies (Li et al., 2018, Cichy et al, 2019). Based on literature values, the Ortho-P reduction values predicted by Visual MINTEQ seem to be low.

BioWin Model Results
Condition 1 The BioWin model (Figure 1) was calibrated for influent fractionation, effluent quality, and solids production. A model builder unit in BioWin with modified default kinetics was used to simulate the PAD process. The model predicts ammonia, pH, and alkalinity well, overpredicts ortho-P concentration and underpredicts metals (Table 7). Ca concentration in PAD can be further adjusted by adjusting influent Ca concentration based on actual data. Figure 10 shows the precipitates and concentrations in AD and PAD. Brushite is formed in PAD, along with small concentrations of ferrous sulfide and vivianite. This is consistent with the findings of XRD analysis, indicating presence of vivianite and amorphous material which most likely includes Amorphous Calcium Phosphate. Conditions 2 to 5 Dynamic modeling for the existing conditions at City of Longmont WRRF was conducted (Figure 2). The model was calibrated for influent fractionation, effluent quality, centrate quality, digesters, and solids production. Influent Ca and Mg was measured in three samples. The model predicts digester Ortho-P, Ca, and Mg very well (Figure 11). For Conditions 3 to 5, CaCl2 was added for Ca:P molar ratios of 1.5, 2.5, and 3.5, which corresponds with Ca dosing of 180, 300, and 420 Ib/d, respectively. The results of BioWin modeling for Conditions 2 to 5 are presented in Table 8.

Comparison of Models
Comparison of two models with XRD analysis was conducted for conditions 2 to 5 (Table 9). Mineral species predicted by BioWin and Visual MINTEQ are generally in agreement. For condition 2, both models predicted both struvite and calcium phosphate which is consistent with XRD analysis. For Conditions 3 and 4, both models show increased quantities for calcium phosphate species. While BioWin predicts some small quantities of struvite formed in Conditions 3 and 4, Visual MINTEQ does not predict any struvite formation. XRD predicts some amorphous content which is likely calcium phosphate and no struvite. For Condition 5, both models predict increased quantities of calcium phosphate species with no struvite formation, which is consistent with XRD analysis.

Summary and Conclusions
The results of this study indicate that while a chemical equilibrium can be helpful in predicting the precipitation of brushite and other phosphate-bearing minerals, BioWin was able to predict precipitates formation and Ortho-P reduction reasonably well. Verification of models with XRD analysis was not possible for the PAD samples, since XRD analysis showed that the majority of the sample was in amorphous form. Verification results with XRD demonstrated that BioWin can be reliably used to determine the precipitates in AD samples for calcium phosphate. The main advantage of BioWin is considerations for kinetics for precipitation and Ca and Mg content of biomass.