Start-up and performance of a downflow fluidised bed reactor for biological treatment of yellow wastewater and nutrient recovery

Source separation of domestic effluents in grey water (from sinks and showers or baths), brown water (toilet flush waters) and yellow water (from urinals) has been proposed to allow a better wastewater treatment, to recover resources and reduce pollution (Malila et al. 2019).

Grey water treatment and reutilisation is easier if no excessive organic matter, coliform pathogens, or N and P are present. Grey water is generally classified in two categories based on the types of sources: low pollutant load (sink and showers) and high pollutant load (kitchen and laundry). The main characteristics of grey water are that the organic load and content of nutrients are low; therefore, treatments could involve filtration coupled with chemical disinfection (Jefferson et al. 2004). Grey water with low pollutant load can be recycled and used for domestic cleaning, flushing toilets, washing vehicles, etc. (Elhegazy & Eid 2020).

In the total sewage, the urine stream comprises less than 1%, but it provides about 79% nitrogen, 47% phosphorus and 71% potassium (Friedler et al. 2013). Therefore, urine contains the main nutrients for crops. If adequate recovery procedures were implemented, these nutrients could be used as a sustainable fertiliser (Randall & Naidoo 2018). For instance, considering the average daily urine production per person of 1.5 L (Rauch et al. 2003) and depending on diet, and time of the day, among other factors, the amount of nitrogen in urine can reach a concentration up to 9 g N L⁻¹ (Winker et al. 2009).

It has been suggested that 1.5 L of urine could be applied to one square metre of arable land to fertilise it. Consequently, the annual urine production per person (550 L) would be sufficient to fertilise 330 m² of land per year (Jönsson et al. 2004). However, to recover such huge amounts of N, a decentralised urine storage facility close to the generation site (households, schools, etc.) would be necessary to prevent nutrient losses and nuisances (Udert et al. 2006).

The solubility product of struvite indicates that the minimum concentration for precipitation is [Mg²⁺]=[NH₄⁺]=[PO₄³⁻]=(10^−12.6)^1/3=6.31×10⁻⁵ M. The yield of struvite could increase if addition of PO₄³⁻ and Mg²⁺ is done at the rate of NH₄⁺ production during the biological urea hydrolysis.

The magnesium/ammonium/phosphate ratios (Mg²⁺/NH₄⁺/PO₄³⁻) reported in fresh urine and anaerobically treated swine wastewaters are 1/79/5 and 1/74.9/1.8, respectively (Udert et al. 2003; Kim et al. 2017). However, Kim et al. (2017) reported that the optimum molar ratio of Mg²⁺/NH₄⁺/PO₄³⁻ to improve the recovery of phosphate as struvite from artificial swine wastewater is 1.2/1.0/1.0, in a pH range from 8 to 10.

It is well known that water quality of natural reservoirs has badly deteriorated because of disposal of raw or treated wastewater, leading to eutrophication processes. Therefore, it is important to research for wastewater treatment strategies that look not just for a reduction of pollutant loads, but also for nutrient recovery. This new generation of nutrients could be introduced to new agriculture strategies that also look for reduction of pollution.

The downflow fluidised bed reactor (DFFBR), also known as an inverse fluidised bed, has been used for the treatment of high and low strength wastewaters (Castilla et al. 2000; Mallikarjuna & Dash 2020). The biomass is immobilised on a support of small particles of low-density polyethylene which floats, is dragged down by the downflow and allows for an easy separation of solids denser than water and their recovery at the conical reactor bottom.

The aim of this work was to evaluate the treatment of source separated yellow water (undiluted human urine) using biological hydrolysis to favour the recovery of nutrients as struvite in a DFFBR, inoculated with anaerobic granular sludge, operated in batch mode. To reach this goal three inoculums were initially evaluated for urea hydrolysis.

The recovery of nutrients from human urine might be one of the solutions for the sustainable management of effluents and sanitation. Also, the recovery of nutrients from yellow waters would step up the circular economy of urine. Furthermore, this is the first report on the biological treatment of yellow water using a DFFBR.

Human urine was obtained from 31 volunteer donors (females (17) and males (14) between 13 and 50 years old). Urine was collected in PET plastic bottles. Composites of fresh urine samples (FU) were stored at 4 °C for less than 24 h before being used. The average concentrations of the main components (Table 1) are similar to other reports (Randall et al. 2016; Christiaens et al. 2019).

Three inoculums were evaluated: inoculum FU5 consisted of composite samples of fresh urine obtained from different donors. The composite sample was stored at room temperature for five days, afterwards it was used as inoculum. Inoculum FS was anaerobic granular sludge from a reactor that treated food industry wastewater. Finally, inoculum MGS was anaerobic granular sludge from an upflow anaerobic sludge blanket (UASB) reactor treating municipal wastewater.

The number of ureolytic organisms were estimated by the most probable number (MPN) method using the Stuart's urea broth pH 6.8±2 (Stuart et al. 1945). Tubes were incubated at 25 °C, for six days.

The support material used was low-density polyethylene particles (LDPE, 17070 Pemex), with an average diameter of 0.4 mm and apparent density (ρ) 0.374 g·cm⁻³.

To promote the immobilisation of biomass on support material, initially, 140 g of LDPE were conditioned in a 1 L Erlenmeyer flask containing 500 mL of distilled water, casein peptone (1 g·L⁻¹) and yeast extract (0.5 g·L⁻¹). Particles were kept stirring for 48 h. Subsequently, the particles were transferred to a 2 L glass bottle containing 1.8 L of fresh urine composite sample (FU) and 200 mL of MGS inoculum (volatile solids = 27 g VSS·L⁻¹, specific methanogenic activity = 0.49 g COD-CH₄·g⁻¹VSS·d⁻¹) previously washed with a 0.89% NaCl solution and macerated. The closed flask was left under stirring for 24 h. Finally, the contents of the flask were transferred to the DFFBR, and FU was added until the working volume of 2.3 L was completed. The reactor was kept in recirculation for three days and then batch operation began.

Biomass immobilisation monitoring was evaluated by observational evaluation using a Scanning Electron Microscopy (SEM) technique at 10 and 30 days of operation and the content of immobilised volatile solids (IVS) and specific ureolytic activity (SUA) after 36 days of operation.

Batch tests were conducted to measure the SUA for three inoculums (g Urea-N g⁻¹ VSS·d⁻¹) and immobilised biomass (g Urea-N g⁻¹ IVS·d⁻¹). Table 2 shows the experimental conditions. Fresh urine (FU) was filtered through 0.2 μm membranes for sterilisation. Serum bottles were sealed with rubber stoppers capped with aluminium clamp seals and were incubated on an orbital shaker at 100 rpm. The NH₄⁺-N production and urea consumption were quantified, and the initial slope divided by the amount of VS or by the attached biomass (immobilised volatile solids, IVS). Experiments were run in triplicate. The evaluation of SUA for immobilised biomass was done during the performance of the reactor.

The 2.3 L DFFBR consisted of an acrylic tube (108 cm high, 5 cm internal diameter) connected to a gas-liquid-solid separator to retain the LDPE particles and to fix the water level (Figure 1). The reactor was operated anaerobically at 30±4 °C in batch mode (24 h reaction time).

With a peristaltic pump (Masterflex L/S, 1–600 rpm) the bed was fluidised with a recycle flow rate of 497 L·d⁻¹ equivalent to a downflow superficial velocity (V_d) of 10.5 m·h⁻¹, which expanded the bed of 140 g of LPDE to 25 to 30% of the column. After each batch, the full reaction medium (yellow water) was replaced in the reactor, the liquid-solid-gas separator, and all the silicone tubing (Figure 1). The urine composition varied with each batch (Table 1).

The parameters quantified during the performance of the DFFBR were: pH, urea, NH₄⁺-N, COD, PO₄³⁻ and SO₄²⁻ at the beginning and at the end of each batch.

K_m is the half-saturation constant (mol urea·L⁻¹), V is the specific ureolysis rate (mol urea·g⁻¹ IVS·d⁻¹), V_max is the maximum specific rate of ureolysis, and U is urea concentration (mol urea·L⁻¹).

The ammonium nitrogen and struvite production and COD removal efficiency (η) were measured during urea hydrolysis in a batch operated DFFBR.

To improve the recovery of struvite during the hydrolysis of urea in the DFFBR, Mg²⁺ and PO₄³⁺ were added at a molar ratio of 1.6/1 (250 mL of MgCl₂·H₂O 1.6 M and 250 mL of Na₂HPO₄·12H₂O 1.0 M) with a peristaltic pump during 30 min (through the reactor outlet line, Figure 1(5)), when the pH increased to 9. The reactor continued operating up to 9 h. In this experiment Mg²⁺ and PO₄³⁺ were added in two consecutive batches.

The pH, ammonium, phosphate, and urea were evaluated. The precipitates were removed from the water level and bottom cone of the reactor (Figure 1(4) and 1(5)) and were dried at room temperature for quantification.

The COD was assayed following Standard Method (APHA 2005), the ammonia was measured with a high performance ion selective electrode for ammonia after converting ammonium in to ammonia by adding an ionic strength adjuster that increases the pH above 11. It was then stoichiometrically to converted to NH₄⁺-N for reporting purposes, and urea was measured with the Ehrlich reagent.

Sulphate and phosphate ions were quantified using Hanna colorimetric assay kits and an HI 83200 colorimeter. All samples were filtered through 0.45 μm nitrocellulose membranes.

The biomass formation in the DFFBR was quantified by the volatile suspended solids (VSS) gravimetric method, Standard Method (APHA 2005); this was done after biofilm separation from support material by sonication with successive deionised water washings in an ultrasonic bath.

The morphology of dried precipitates was observed on a Scanning Electron Microscope (SEM JEOL, JSM-5900 LV). Identification of functional groups in the precipitates was done by Fourier transform infrared spectroscopy (PerkinElmer Spectrum 100 FTIR spectrometer) in a wavenumber range from 500 to 4,000 cm⁻¹.

Furthermore, approximately 0.05 g of dried precipitate was dissolved in 25 mL of a 4% hydrochloric acid solution. The NH₄⁺-N was measured in the acid solution with the ammonia ion selective electrode as NH₃-N. Also, elemental Mg²⁺ was analysed by atomic absorption spectrometry (Thermo Elemental SOLAAR).

To identify potential alternative sources of ureolytic microorganisms for urea transformation from human urine, different inoculums were screened at different temperatures. Figures S1, S2 and S3, in the supplementary material, show the influence of temperature on the hydrolysis activity of each inoculum, from following the NH₄⁺-N production. Overall, inoculums evaluated at 25 °C had the lowest ammonium production at 24 h and the final pH for all inoculums was less than 8. However, the experiments conducted at 55 °C showed higher ammonium production than inoculums evaluated at 35 °C (Table 3). The pH at 35 °C was less variable, 8 to 8.9, while at 55 °C the variability was higher, 7.2, 8.0 and 9.0.

The presence of ureolytic microorganisms in FU5 presented the highest MPN·100 mL⁻¹ (1.1×10⁵), followed by MGS (1.5×10⁴), while the FS inoculum presented the lowest MPN·100 mL⁻¹ (2.3×10³). Nevertheless, the ability of the anaerobic granular sludge (MGS) to hydrolyse urea from human urine at different temperatures showed a positive influence on urea hydrolysis at 55 °C (Figure 2).

The specific ureolytic activity of the MGS on human urine was measured at different temperatures and the activation energy (E_a) calculated at the end of these experiments indicated that the pH values were 7.7, 8.8 and 9.0 at 25 °C, 35 °C and 55 °C, respectively. The abiotic hydrolysis was run as a blank, with sterile FU at 35 °C, with no significant pH change 5.9 to 6.1, nor NH₄⁺-N production after 1 d.

The results show that an inoculum was needed; the MGS had better adaptation to urine, the SUA rates increased with temperature and presented the best E_a and SUA. The MGS inoculum was readily adapted to perform urea hydrolysis from human urine, probably because urine is a permanent component in domestic wastewaters, so this inoculum was used in the DFFBR.

In this study, the E_a was 13 kJ·mol⁻¹ and Ao=120·d⁻¹ (Figure S4), close to the values reported for soil ureases, 10 to 24 kJ·mol⁻¹ at 8 to 35 °C (Singh et al. 1992; Juan et al. 2010), and smaller than E_a for jack bean urease, 33 to 31 kJ·mol⁻¹ at 10 to 40 °C, in phosphate and citrate buffer (Huang & Chen 1991).

During start-up, the pH increase was slow; it required 3 days to increase from 5.6 to 9.3 (data not shown). The biomass concentration was 0.72 g IVS·L⁻¹ at 36 days. During start-up and biomass immobilisation, period I (Figure 3), initial pH was 6.2±0.4 and increased to 9.3±0.1, while urea started from 15.8±5.4 g·L⁻¹ (0.263 M), and urea was 93.4% hydrolysed, contributing to an increase in the ammonium concentration from 0.7±0.2 g NH₄⁺-N L⁻¹ to 5.3±1.2 g NH₄⁺-N L⁻¹ (ΔN=0.32 M), which indicates an ammonium recovery of 65%. With an initial COD of 4.1±8 g L⁻¹ a removal of 36% was recorded together with the production of a small amount of precipitate. These events indicated the occurrence of alkaline ureolysis by the immobilised biomass at basic conditions.

The biomass immobilisation was assessed through SEM observations. The LDPE particles before inoculation had an irregular surface with cavities (Figure 4(a)), apparently allowing the biofilm formation. After 10 days, some bacillus and coccus were observed colonising the surface (not shown). At 30 d there was an important biomass immobilisation with different bacterial morphology: bacillus, diplobacilli, coccus and diplococci (Figure 4(b)).

After 36 days of inoculation the specific ureolytic activity was 159±4.7 g Urea-N g⁻¹·IVS·d⁻¹ showing a considerable increase compared with the suspended fresh MGS inoculum at the same temperature (Table 2). Figure 5 shows that urea was almost completely hydrolysed in 0.1 d (91%). This proves that selected ureolytic microorganisms colonised the support material. The NH₄⁺-N production was accompanied by a pH increase from 6.5 to 9.5, while the control, with no inoculum, did not show significant pH and urea concentration changes (not shown).

It was observed that the inoculation and start-up strategy favoured the ureolytic biomass immobilisation and consequently urea hydrolysis was achieved in less than 1 d. In this condition, urea hydrolysis was faster than reported by Neethling (2015), who evaluated two biofilters. In one, uninoculated Kaldnes rings were used as support material and in the second, biofilm carriers from an aerobic membrane biofilm biofilter reactor were used as inoculum. In both biofilters, urine and acetate wastewater were used during the start-up, acetate as carbon source (13 g·L⁻¹) to promote biomass growth. In the former, only undiluted urine was treated, and there was slight indication of ureolytic activity after 24 days. In the latter, the biofilter was fed with urine and acetate wastewater, and ureolytic activity was observed after four days of inoculation.

The treatment of yellow water at ambient temperature, between batches 20 to 108 was evaluated. The FU chemical characteristics were highly variable along the treatment (Tables 1 and 3, Figure 3). The major component was urea 14.3±5.7 g·L⁻¹ with a COD of 9.2±2.8 g·L⁻¹. The high variability of data reported in this work for human urine could be explained by several factors, for instance the age, gender, eating habits, and the level of physical activity of the donors. Our donors included young bachelor students, mature university lectures and a diverse group of professionals. Udert et al. (2003) also mentioned geographical region and environmental factors as a source of variability for the physicochemical characteristics of human urine.

The urea volumetric load (B_vurea) varied between 4.3 g·L⁻¹·d⁻¹ and 28 g·L⁻¹·d⁻¹. Despite the fluctuations of the physicochemical characteristics of yellow water, the performance of the reactor was stable. Overall, the hydrolysis efficiency was high 93.4%. This is the result of ureolysis; urea was transformed into NH₄⁺-N over time, this was confirmed by the rise of NH₄⁺-N concentration at the end of the reaction time (1 d) (Table 3). Besides, the specific ammonium production rate (r_NH4⁺-_N) in the reactor was quantified by the ratio of the NH₄⁺-N production rate to the immobilised volatile solids, minimum of 4.2 g NH₄⁺-N·g⁻¹ IVS·d⁻¹ and maximum of 35 g NH₄⁺-N·g⁻¹ IVS·d⁻¹.

Additionally, a good linear correlation between feed urea concentration and the ureolysis rate was found, r=0.972 (95% confidence interval, 0.954–0.982), which means that a high urea initial concentration would be related to a high ureolysis rate in the DFFBR. These results show that the reactor operating condition was adequate to transform either low or high volumetric loads of urea without the addition of an external carbon source.

Thus, the biological treatment of yellow wastewater in a DFFBR is a viable option. Also, in terms of urea volumetric loading rate, the reactor was more robust in comparison with other biological reactor configurations previously proposed for the treatment of nitrogen rich effluents (Garrido et al. 2001; Eiroa et al. 2004). For instance, Garrido et al. (2001) evaluated ureolysis under anoxic conditions in a multifed upflow filter for wastewater with high concentration of formaldehyde 1.5 g·L⁻¹, volumetric loading rate of urea 0.99 g·L⁻¹·d⁻¹ and hydraulic retention time (HRT) of 1 d. They reported hydrolysis percentages between 65 and 85%. Their volumetric loading rate was considerably lower than the ones reported in this work; furthermore, these authors used external carbon sources.

On the other hand, it is important to note that the mean COD in the influent during the immobilisation period was approximately half of the COD recorded for the performance period (Table 4). The COD removal efficiencies (η) oscillated within the wide range of 5 to 62.8%, while the removal rate (r_COD) ranged between 0.5 and 8.5 g COD·L⁻¹·d⁻¹, which can be attributed to the microbial activity. Also, the biological degradation of organic compounds is possible under anaerobic conditions. The organic compounds could perform as electron acceptors (Udert et al. 2006), and this can be done by microorganisms that can grow in urine (Zhang et al. 2013).

From a wastewater treatment point of view, there are few references that report COD removal from yellow water. For the anaerobic ureolysis of urine in a sequencing batch reactor (SBR), 17±9% of the initial COD was converted to volatile fatty acids with HRT of 0.45 d (Christiaens et al. 2019). In a bioelectrochemical system, degradation of COD in undiluted urine (7 g COD·L⁻¹) reached 75 and 76% in long treatment hydraulic times; 40 and 60 days, respectively (Barbosa et al. 2017). Gao et al. (2018) reported 95% of COD removal from diluted urine by a factor of 10 (1.6 g COD·L⁻¹) in a bioelectrochemical system. In this study the mean removal of COD was 31±16% in undiluted urine at RT of 1 d; thus we consider that removal of COD is an advantage to the process.

The high variability of initial COD concentration (Table 1) was related to the high variability of organic volumetric rate (B_v), which ranged from 2.5 to 14.2 g COD·L⁻¹·d⁻¹. There were two periods that the COD increased; these were observed between batches 31 to 48 and 84 to 99; the COD average was 8.2 g L⁻¹ and 11.4 g L⁻¹, respectively, and in consequence the B_v also increased. These successive increments on COD could have had a negative effect on the urea hydrolysis efficiency. The lowest ureolysis efficiencies were detected between batches 31 to 48 and 84 to 99, with 67 and 70%, respectively. Thus, it is possible that the microbial consortium was temporarily affected by the B_v increment. Ureolytic microorganisms had to readapt to new conditions, though this adaptation was done readily.

During the urine hydrolysis, precipitates in the DFFBR column were observed together with phosphate (26%) and sulphate (24%) mean removal efficiencies (Table 4). These results are in accordance with other similar studies on urine hydrolysis; the pH and ammonia concentration increased during urea hydrolysis, and these are cofactors for phosphate precipitation (Udert et al. 2003; Maurer et al. 2006). These precipitates were analysed by FTIR and SEM.

The downflow velocity in the DFFBR was 10 m h⁻¹, which was lower than other works. Shih et al. (2017) worked with a fluidised-bed reactor; these authors investigated the crystallisation of struvite using synthetic wastewater. Additionally, they used small struvite particles (53–297 μm) to promote growth of crystals, the flow velocity was 19 m h⁻¹, and to increase pH they added NaOH. Under these conditions 90% of phosphorus was recovered. Sathiasivan et al. (2021) were able to recover struvite from synthetic urine using a DFFBR with a downflow of 51.12 m h⁻¹ with low density (836 kg m⁻³) polyethylene particles (5 mm). To promote the formation of struvite the pH was raised to 9.5 with NaOH. In our work the fluidisation velocity favoured the biomass immobilisation and at the same time the mixture within the reactor allowed the precipitation of solids. Furthermore, the configuration of the DFFBR helped to collect the solids in the conical bottom of the reactor and in the liquid-solid-gas separator and water lever control (Figure 1).

There have been important concerns about emissions of ammonia to the environment through agriculture activities (Sigurdarson et al. 2018). One strategy is to reduce the ammonia emissions maintaining the equilibrium pH between NH₄⁺ and NH₃ in aqueous solutions. In this work the average pH reached was 9.0, which is close to the pK_a for these chemical species. Furthermore, the DFFBR was kept tightly closed and a gas scrubber was implemented (Figure 1); the idea was to prevent any loss of NH_3(g); however during the performance evaluation there was no evidence of this gas. Also, there was no presence of the characteristic odour of hydrolysed urine during the biological hydrolysis process. But, during the recovery of solids, the ammonia odour was perceived.

During the performance of the DFFBR, individual kinetics of urea hydrolysis were evaluated by the estimation of kinetic parameters (Figure 6(a) and 6(b)). Temperature and urea concentration in human urine oscillated between 24.4 °C and 30 °C and from 12 g L⁻¹ to 27 g L⁻¹, respectively. This kinetic study was conducted between batch 57 and 107.

Through the evaluation of the biological hydrolysis, the formation of fine white particles in suspension was observed between 0.08 d and 0.196 d reaction time (Figure 6(a)); these particles increased and subsequently sedimented in the conical part of the reactor and in the phase separator, and the colour of these precipitates was brownish-white. Simultaneously, the pH changed from 8.5 to 9.1, so this stage can be considered as the processes of nucleation and growth for struvite crystals and subsequent precipitation. During this period, the maximum production of ammonium and a change of pH and hydrolysis of urea were observed. Later the changes in the concentration of ammonium and pH were minimal. After 0.26 d of reaction the presence of precipitates in the reactor column was not observed and the colour of the urine had changed from yellow to amber yellow.

The ureolysis rate constant (k_u) was calculated for each urea concentration in urine, ranging from 12 g L⁻¹ to 27 g L⁻¹ (Table 5). The average half-life (t_1/2) for urea hydrolysis was 0.25 d±0.2 d. Also, it was found that the highest removal of phosphate occurred at 0.3 d when the highest production of NH₄⁺-N and the pH reached 9.0. These results suggest that the reactor can be operated at t_r between 0.2 d and 0.3 d to remove nutrients efficiently from undiluted urine.

The kinetic constants were calculated with Equation (6), and initial urea concentrations were 12, 13, 23 and 27 g L⁻¹. The K_m value obtained in this work (152 mmol L⁻¹) is within the range from 2.2 to 550 mmol L⁻¹ reported for microbial urease (Mobley & Hausinger 1989; Connolly et al. 2015). This result suggested the presence of urease producing organisms in the DFFBR with a good affinity for high urea concentration. In addition, urease activity is known to vary from species to species (Mobley & Hausinger 1989).

The mean specific ureolytic activity (SUA) was 121 g Urea-N g IVS ⁻¹ d⁻¹; it was higher than reports for auto fermented source-separated urine in a SBR with a mixture of alkaline fermentation sludge and anaerobic digestion sludge at 28 °C (Christiaens et al. 2019 ).

The formation of precipitates was observed throughout the DFFBR evaluation, which had a colloidal consistency in the reactor. After precipitation and separation, it had a gelatinous consistency, and when dried it was a fine white powder, sometimes with a yellowish colour. These precipitates could likely contain nutrients recovered from urine hydrolysis, as reported elsewhere (Morales et al. 2013). Precipitates recovered from the biological hydrolysis of urea present in urine are struvite, a crystal form of ammonium phosphate and magnesium. Besides, ureolytic microorganisms perform urea hydrolysis and favour biomineralisation of struvite and other compounds such as calcium carbonate (Arias et al. 2017). The recovered precipitates were observed in the microscope (Figure 7(a)) and by scanning electron microscopy (Figure 7(b)). The crystal morphologies observed were irregular, elongated and trapezoidal shapes of different sizes; some were rectangular with peak-shaped ends, and these characteristics had been described as typical for struvite (Münch & Barr 2001).

The FTIR technique was used to analyse the precipitates. The spectrum of the solids recovered during urea hydrolysis of human urine and the vibrational band assignments according to literature are shown in Figure 8 and Table 6.

The absorption peaks due to interactions between water and water and between ammonium and water are present at 2,920 to 3,198 cm⁻¹, 1,628 to 1,653 cm⁻¹, 748 to 877 cm⁻¹. The absorptions at 1,431 cm⁻¹ and 1,432 cm⁻¹ are related to antisymmetric bonds characteristic of the NH₄⁺ ions. The vibrational mode tetrahedral PO₄³⁻ ions occurred between 561 to 1,047 cm⁻¹. The absorption regions in this work are in accordance with values reported for pure struvite (Liu et al. 2013; Krishnamoorthy et al. 2020). Therefore, this evidence confirms that the fine white powder recovered at the end of each batch was struvite. Also, it has been reported that struvite is the dominant solid phase in the pH range from 7 to 11, and struvite crystals would be formed around pH 7.5 to 10.5 (Kim et al. 2017). In this investigation the final pH was within these values (Table 4).

The biological recovery of solids ranged from 0.3 to 1.72 g L⁻¹. This apparently low recovery is due to the low concentration of Mg²⁺ in urine, therefore, limiting the struvite formation. However, these values are close to the theoretical yield (1.0 g L⁻¹) considering the mean value of Mg²⁺ concentrations reported previously for urine by Udert et al. (2006) and struvite yield (1.9 g L⁻¹) estimated by (Tilley et al. 2009).

The biological recovery of struvite from urine is fast and represents an alternative for nutrient recovery such as phosphate and nitrogen. Besides, struvite is a slow-release fertiliser, the agricultural application of struvite does not represent a hazard to the environment, nor does it contibute to global warming; struvite contains 5.7% N, 12.6% P and 9.9% Mg (Maurer et al. 2006). In contrast, urea is a fast release fertiliser that contains 46.6% N which is lost to the environment through different ways (Liang et al. 2007); the biggest loss is through volatilization (26.5–29.4%), which contributes to greenhouse emissions.

To increase the yield of struvite recovery, Mg²⁺ and PO₄³⁻ were added to the DFFBR in a molar ratio of 1.6/1.0, respectively, at the end of the biological hydrolysis. The biological urine hydrolysis was achieved at t_r of 0.26 d. At this time, the efficiency of hydrolysis was 85% and removal of PO₄³⁻ and Mg²⁺ was 23.7% and 99.3%, respectively. The NH₄⁺-N concentration increased from 0.96 to 9.1 g L⁻¹, and the pH changed from 7.2 to 9.3. After the biological hydrolysis the quantification of Mg²⁺, PO₄³⁻ and NH₄⁺ was 2.3×10⁻⁵, 0.010 and 0.651 mol L⁻¹, respectively. At 0.26 d of reaction Mg²⁺ and PO₄³⁻ were added (1.6/1.0) over a 30 minute period. Therefore, considering the remaining species involved in the struvite formation and the added salts, the new ratio of Mg²⁺/PO₄³⁻/NH₄⁺ was 1.4/1.0/6.6.

The chemical reaction was almost instantaneous; the calculated recovery of PO₄³⁻, NH₄⁺ and Mg²⁺ at the end of the 30 min. was 98, 33 and 90%, respectively. At the end of the total reaction time (0.38 d), 82.3 g of precipitates were recovered. During the chemical reaction pH decreased from 9.28 to 8.7; these values of pH were in the range of struvite formation. Additionally, the chemical composition of these precipitates was 8.3% and 5.7% of Mg²⁺ and N, respectively.

Struvite quantification was based on mass balance of phosphorus in the aqueous phase; phosphorus was the limiting reactant. Quantification of phosphorus was done after the biological reaction (0.26 d) and at the end of the chemical reaction (0.28 d). The theoretical composition of the precipitates obtained from this experiment were close to that of pure struvite, indicating that at least 84% was struvite by dry weight. However, the presence of other cations such as Ca²⁺, Na⁺ and K⁺ can interfere in the crystal formation of struvite; by replacing the NH₄⁺, K-struvite can be formed (Di Costanzo et al. 2021).

As we said before, the hydrolysed urine was still rich in nitrogen content; it could be used further in other nutrient recovery processes, such as those reported by Zeng et al. (2018). These authors evaluated phosphorus recovery from sludge that is produced in wastewater treatment plants. This sludge is rich in phosphorus. Hydrolysed urine could be used to increase pH in nutrient recovery process and contribute with a nitrogen source.

The DFFBR configuration and operational conditions (30 °C ±4 °C, 10 m h⁻¹ downflow velocity) promoted the adhesion of ureolytic biomass on the LDPE support to a high concentration (0.72 g IVS L⁻¹), which allowed a high reaction rate and an effective separation of struvite, which was dragged down while the bed of immobilised biomass remained in the fluidised bed.

The biomass capable of urea hydrolysis under anaerobic condition was tolerant to high ammonia concentration and elevated pH. The DFFBR proposed in this study operated at environmental temperature demonstrated robustness during the treatment of human urine by biological urea hydrolysis. High ureolysis efficiencies (η_urea=93.4%), and good COD removal efficiencies (η_COD=31.2%) were obtained from batch treatment of yellow wastewater at 1 d.

The results also showed that the variability in the urine components (urea, COD, pH and salts) did not significantly affect the urea hydrolysis and performance of the bioreactor. The ureolysis kinetics showed that the reactor can be operated at low t_r to recover nutrients (1.72 g) from source separated undiluted urine. Kinetic parameters from Michaelis-Menten model were: 0.152 mol L⁻¹ (K_m) and 8 mol g ⁻¹IVS d⁻¹ (V_max) and mean specific ureolytic activity was 121 g Urea-N g⁻¹ IVS d⁻¹.

We consider that the DFFBR appears to be a good system for the simultaneous ureolysis process and recovery of nutrients as struvite. Furthermore, the chemical addition of an external source of Mg²⁺ and PO₄³⁻, in a molar ratio of 1.6/1, to already hydrolysed urine increased the yield of struvite formation (82.3 g of precipitates of which more than 84% was struvite).

Finally, we concluded that ureolysis in a DFFBR was viable for undiluted yellow-water treatment and simultaneous recovery of nutrients. However, the effluent of the reactor, hydrolysed urine, was still a high strength matrix that cannot be disposed of in water bodies, as ammonia is still present. Nonetheless, hydrolysed urine can be used for other green treatments such as microalgae cultivation or as a source of nutrients, if used for irrigation. Also, further exploration of the addition of Mg²⁺ and PO₄³⁻ using different sources to increase nitrogen recovery is needed. Some of these options are under research already. In other words, the treatment of undiluted urine by biological hydrolysis in a DFFBR is one step in the recovery of nutrients that ultimately could save energy, water and even be part of a loop in the circular economy of urine.

This project was granted financial support from The Science and Technology Institute of the Federal District, a Mexican Council of Science, Technology and Innovation (PICSO12-106). Also, to CONACYT, a fellowship was granted to Belem Espinosa Chávez, the Universidad Autónoma Metropolitana-Iztapalapa and Universidad del Mar campus Puerto Ángel. We thank Dr. Omar Viñas Bravo and Vice Chancellor Hector López Arjona from Universidad del Papaloapan for their support in the use of the FTIR spectrometer, Dra. Susana García Ortega for her help in reviewing the manuscript and Anna Cortesio and Josseph M. Carreiro for the English revision.

The authors declare no conflict of interest.

All relevant data are included in the paper or its Supplementary Information.