PULSE process: recovery of phosphorus from dried sewage sludge and removal of metals by solvent extraction

ABSTRACT Phosphorus (P) is an indispensable nutrient for agriculture. Recovery and recycling of phosphorus from waste streams is necessary to ensure a circular P economy and reduce dependence on disproportionately distributed mineral P resources. In this study, a new process called ‘PULSE’ is presented for the recovery of P from sewage sludge, which can handle high metal contents. The process involves drying of sludge prior to acidic leaching to overcome the challenge of solid–liquid separation at low pH and to reduce the overall material flows. Another key point of the process is the removal of metals using reactive extraction to obtain a high-quality product with good plant availability. Laboratory experiments were conducted to evaluate and select the best process options. A chemical equilibrium tool was developed to simulate the unit operations of the process for optimization. Dissolution of P from sludge depends on leaching pH and the fraction of inorganic P in the sludge. The maximum P leaching efficiency for the sludge used in the study was between 65 and 70%. Under the tested conditions, Fe, Cd, Cu, Hg, Pb, and Zn were successfully removed from the sludge leach liquor by reactive extraction. The recovered product has a nutrient mass fraction of about 51% that includes Ca, PO43-, Mg, and K. Pot trials confirmed that the agronomical efficiency of the product is comparable to that of triple superphosphate. GRAPHICAL ABSTRACT


Introduction
Phosphorus (P) is a vital element required to sustain life.About 95% of P mined are used for agricultural applications predominantly as fertilisers [1,2].Mineral P occurs in nature as sedimentary phosphate rocks which are formed by the combination of physical and chemical phenomena [2].The distribution of phosphate rock resources is disproportionate and only a few economically exploitable significant P reserves were discovered in the past century mostly located in Africa and North America [2].Currently in the EU, phosphate rock mining is only carried out in Finland which covers just about 10% of EU's P demand and remaining 90% is imported [3].Although other phosphate rock deposits are found in EU they are currently not exploited due environmental and economical constraints [2,4].Therefore, due to the scarcity of P resources and its importance, phosphate rock was added to the EU critical raw-materials list in 2014 to promote its sustainable use [5].A substantial fraction of the used P ultimately ends up in different waste streams such as wastewater [6].One of the ways to ensure sustainable utilisation of P and to establish a circular P economy is to recover P from waste streams.P in wastewater is sequestered in the sludge produced during the biological wastewater treatment and frequently its removal from the wastewater is supplemented by chemical precipitation with Fe or Al salts to meet the regulatory effluent quality.The resulting sludge is therefore a potential source for P-recovery.However, the direct application of biological sludge to agricultural land is becoming increasingly restricted due to potential presence of various pollutants and the application of chemically precipitated sludge is generally not allowed under EU legislation to avoid contamination of the receiving soil [7,8].This highlights the need to develop appropriate technologies for the recovery of P so that it can be used as a fertiliser.
Several processes for P-recovery from wastewater, sludge, sludge-liquor, and sludge ashes have been developed in recent years to produce secondary P products [8][9][10].Non-metallic chemical precipitation of P as struvite or calcium phosphates directly from wastewater and sludge liquors has also been studied in detail [10].While direct precipitation of P from wastewater and sludge liquor offers some advantages such as lower operation costs and simple technology, it may not be feasible where the influent concentration of heavy metals is higher or if the P concentration is lower.P-recovery from ashes is only possible if the sludge is mono-incinerated or else the P content is diluted with other waste and the recovery is no longer economically feasible.The advantages of recovering P from sludge ashes include higher concentrations of P, elimination of organic compounds, and reduced residual waste after the recovery process.However, currently the limited availability of mono-incineration infrastructure poses a challenge for wide-scale implementation of Precovery from ashes.Therefore, this research work focuses on the recovery of P from sludge directly and the sludge remaining after P removal can then be used in any incineration as renewable fuel, also, e.g. in cement production.
P is present in sludge either as chemically bound organic P or as inorganic metal phosphates such as iron, aluminium and calcium phosphates.Wet chemical leaching under acidic conditions is commonly employed to extract P from sludge which mainly dissolves the inorganic P [11].The dissolution of these metal phosphates from the sludge depends on the pH and the corresponding metal concentrations.pH values below 1 may be required for complete dissolution of P depending on the fraction of ferric phosphate in sludge [12].However, the solid-liquid separation at pH < 3 presents a major challenge thereby limiting the feasibility of Precovery [13].Also, various heavy metals such as Cd, Cu, Cr, Pb, Zn, etc. that may be contained in the sludge will be co-leached under acidic conditions and eventually end up in the final product [8,14].The metals Fe and Al may not be toxic but their presence in the recovered product reduces the P solubility in soil.Metal separation during P-recovery from sludge has been carried out by metal complexation in the leach liquor by using complexing agents such as citric acid or by means of sulphide precipitation in previously developed processes [14].However, metal complexation leads to accumulation of metals in the process water and metal-sulphide precipitation may not work for all the metals at low pH.In case of P-recovery from sludge ash by leaching, ion exchange and solvent extraction techniques have been tested for metal separation [8,14,15], however these separation processes have not been investigated for P-recovery from sludge.
Considering the challenges encountered while recovering P from sludge, a new P-recovery process has been proposed as shown in Figure 1 which is named 'PULSE' (Phosphorus University of Liège Sludge Extraction) process.The concept of PULSE has been developed based on the Pasch process [15] and adapted to recover P from sludge.
While acidic leaching is frequently used in P-recovery processes from sludge, in PULSE process a drying step is included to support P leaching, solid-liquid separation, and downstream processing.Further, a reactive extraction step is utilised to separate and remove undesired metals from the leach liquor that are co-leached during the P dissolution.In reactive extraction, the targeted components are extracted into a water-immiscible organic phase containing a suitable extractant.The extraction of metals to organic phase may occur by different mechanisms such as ion-exchange, solvation or chelation depending on the extractant and the corresponding target components [16].After extraction, the loaded organic phase is stripped using a second aqueous phase, which allows further concentrating of the metals.In the final step, P is precipitated either as calcium phosphate (CaP) or magnesium phosphate (MgP) by raising the pH of the raffinate by addition of a base.While CaP can be directly precipitated by changing the pH, MgP precipitation requires fractional precipitation of calcium as calcium sulphate by adding (e.g.) magnesium sulphate before P can be precipitated as magnesium salt.Therefore, in the current study it was preferred to precipitate CaP salts directly.The development of the process, the assessment of the different unit operations at lab-scale, and the quality of the product obtained are discussed in this paper.

Materials and methods
In order to choose the best process options and parameters for the PULSE process, the cascaded optiontree (COT) methodology [17] has been used for decision-making support.Since the sludge composition varies from one wastewater-treatment plant (wwtp) to another and also seasonally, a chemical-equilibrium simulation tool has been developed to support optimisation of the process parameters which reduces the number of experiments required for optimisation.Experiments for the individual process steps were established at laboratory scale in order to generate information for process development, assessment of the unit operations, and decision making.The tools, materials, and methods used in the study are discussed in this chapter.

Cascaded option trees
COT is a decision-support tool in which the possible process options for an operation are evaluated against a set of relevant criteria and the best options can be chosen.The COT evaluation can be carried out at different hierarchy levels, for example, to evaluate and select a unit operation for a desired function or for more in-depth evaluations such as the chemicals, catalysts, equipment, process parameters, etc.The criteria are evaluated based on their criticality to the operation.The information required to evaluate these options is obtained from literature, laboratory experiments, or modelling.A matrix is formulated where in the criteria corresponding to the options are graded as good, acceptable, or not acceptable using numerical values or colour codes red (−1), yellow (0), and green (1) as shown exemplarily in Figure 2.Each option receives an overall ranking depending on the number of criteria that are graded as good or acceptable which is recorded in the left-most column of the matrix.The option with the best overall assessment is then selected.If there is a tie between different options, then one of them can be freely chosen.If any of the option-specific criteria is graded as not acceptable, the respective option is excluded from further consideration.In the example for selection of leaching acid, HCl was chosen for the process based on the positive ranking.At the level of evaluation presented in Figure 2, it is already clear that the options HNO 3 and HCl + HNO 3 cannot reach the positive evaluation of HCl.The detailed evaluation of criteria is discussed below.
Similarly, different options such as use of dewatered or dried sludge for leaching, reactive extraction of P or metals, reagents, and process parameters were evaluated using COT during development of the PULSE process.

Solid-liquid-liquid equilibrium (SLLE) simulation tool
The unit operations of the PULSE process are interdependent and sensitive to the pH among other operating parameters.Equilibrium simulations can be used to simulate the reactions occurring in a system and compute the equilibrium pH and the concentration of the various species in aqueous, organic, and solid phases.These simulations allow optimisation of process parameters and reagent requirements to achieve the desired results for each input sludge composition.Therefore, a tool has been developed in Matlab to compute the solid-liquid-liquid (SLLE) thermodynamic and chemical equilibria.For a chemical system, when the input components, temperature, and pressure are known, one of the ways to compute the equilibrium composition is by solving a system of nonlinear equations (NLE) consisting of the law of mass action (LMA), mass balances (MB), and the charge-balance (CB).Several numerical methods have been used to solve these NLEs such as the continuous fraction, simplex, and the Newton-Raphson (NR) methods [18].While the first two numerical methods are zero-order methods that are robust but converge slowly, the NR method makes use of first-order derivatives of the objective functions and converges quickly.The NR method is the most commonly used for computation of chemical equilibria and has been used in programmes such as PHREEQC [19].However, the NR method may lead to non-convergence if the initial guess used to solve the NLEs is far from the solution.This issue can be overcome by using a zero-order method to first generate an initial guess or an approximate solution and then employ the NR method for quick convergence to the final solution [20].
The SLLE tool developed is based on [21] and is coupled with the positive continuous fraction (PCF) method [20] to generate an initial guess.Furthermore, a data-fitting routine has been realised in the tool.
Consider that a chemical system consists of a total number of N species S. From this, it is assumed that N s species exist in the aqueous and organic phases and N p species are precipitating.For computation of equilibria, N m of these species are chosen as master species from all existing species such that they can describe all the reactions occurring in the system.The criteria for selection of the master species has been described in literature [22].
A chemical reaction for the formation of aqueous and organic-phase species can then generally be expressed as where y i,j is the stochiometric coefficient of the master species i with respect to the resulting species j.The master species S i on the left-hand side of Equation ( 1) are chosen such that all resulting species in aqueous and organic phase can be described by their linear combination.The equilibrium for Equation (1) can be expressed via the LMA: where K j is the chemical-equilibrium constant for the formation of S j , a i the activity of the master species i or the product species j, defined as with the reference concentration c 0 of 1 mol L −1 .The free-ion activity coefficient g i accounts for the non-idealities due to the interaction of the ions in solution, and c i is the concentration of species.
In the SLLE tool, g i can be evaluated using the Debye-Hückel or the Davies model [23,24], or set to 1, which corresponds to an ideal solution.For systems involving water-immiscible organic complexes due to reactive extraction, the LMA can be expressed analogous to Equation (2) using the corresponding reaction-specific chemical equilibrium constant.
In case of precipitation, it is assumed that a pure solid precipitate S p with an activity of 1 by definition is formed The LMA for Equation ( 4) is where K p is the equilibrium-precipitation constant.
In addition, the MBs have to be fulfilled: where C i is the total initial concentration of master species i in aqueous, organic or solid phase in mol L −1 which is equal to the input concentration of the respective component added to the system.Finally, chargeneutrality has to be ensured: where z j is the charge of the species S j .The LMAs for the different reactions along with mass and charge balances result in N non-linear algebraic equations which is same number as the total number of species in the system whose equilibrium concentrations are the variables to be determined thus resulting in a unique solution.
During the simulation, the speciation in the aqueous phase and the concentrations in the organic phase due to reactive extraction are first evaluated without considering solids precipitation using the MATLAB NLE solver 'fsolve' after the initial guess has been generated using the PCF algorithm [20].
In the next step, considering the partition of the master species between organic and aqueous phase the equilibrium activities of the master species in aqueous phase are used to determine, if precipitation of any solid species occurs.Precipitation or dissolution of species depends on the saturation index I S which is a function of the ion activity product IAP, with Based on I S , three cases can be distinguished for each solid speciesS p : . For I S , 0, the system is undersaturated and no precipitation will occur or the species will remain dissolved. .If I S = 0, the system is in equilibrium.
. For I S .0, the system is supersaturated and precipitation will occur.
If no solid phase precipitates, N p = 0, all precipitating-phase amounts are 0 and the liquid-phase concentrations are the final equilibrium concentrations.If I S for one or more solid species is positive, a new initial guess is computed using the PCF algorithm [20].The iteration for initial-guess generation is repeated until the I S for all the precipitation solid species is close to 0. The new initial guess considering the aqueous, organic, and solid species is then used to compute the final solution with the NLE solver.
The thermodynamic-data input to the SLLE tool are formulated as a matrix based on the Morel's Tableau [22].The thermodynamic data required for simulations for most of the aqueous-phase reactions and salt precipitations are well documented in literature [23,25].
In case of reactive extraction, the chemical equilibrium constants and the reaction stoichiometries may not be readily available as they are specific to the extractant system, which may include more than one extractant leading to synergistic effects.Therefore, in order to determine these variables, a data fitting routine has been coupled with the equilibrium tool.Results of equilibrium experiments with systematically varied concentration of extractants, pH, and salt concentrations are then used to fit the required parameters.

Sludge and chemicals used
The sludge used in laboratory experiments was obtained from Oupeye wwtp which has a capacity of 438,000 population equivalent and is located near the city of Liège in Belgium.The wastewater is subjected to conventional treatment, which involves primary, secondary, and biological treatment steps.The biological treatment step includes aerobic and anaerobic zones in order to enhance the removal and sequestration of P in the biological sludge.The wastewater then flows to the decanter wherein FeCl 3 is dosed to further decrease the P concentration in the effluent to meet the legislative standards.A portion of the activated sludge is recirculated, while the remainder is sent to a centrifuge for dewatering.Undigested and dewatered sludge with a dry matter (DM) content between 20% and 25% was collected after the centrifuge for the experiments.The dewatered sludge was either dried directly or stored in a refrigerator at 4°C until further use.

Instruments and analysis
Chemicals were weighed using digital balances OHAUS AS200 and Sartorius SIWADCP-1-16-S having an accuracy of 0.001 and 0.4 g respectively.Stock solutions of desired concentrations were prepared using de-ionized water.A single glass pore pH probe from Hamilton (sl.no.16925) was used for pH measurements and was calibrated using a 2-point calibration using buffer solutions for pH 1 and 9.A glass alcohol thermometer was used for temperature measurement.The density of liquids was measured using density and sound velocity metre DS 5000M (Anton Paar, Austria) with an accuracy of 0.001 g L −1 .The dry matter in the sludge was evaluated by drying the sludge in a Memmert UNB 400 oven at 105°C for at least 24 h and until no further change in sample mass was observed.The fraction of organic and inorganic P in the sludge was determined based on the standards, measurements, and testing (SMT) procedure for P fractionation in freshwater sediments [26].For analysis of sludge in the laboratory, sludge was dried at 105°C and calcinated at 550°C until no further change in mass was observed.The sample was then digested according to BS EN 16174:2012.Analysis of total P, total Fe, and Fe 2+ was carried out in the laboratory using the Hach quick tests and measured in the HACH DR 3900 spectrophotometer.Sample dilutions were made using deionised water to the desired range.Inorganic analysis of P and 16 other elements relevant for the study in both sludge and aqueous samples was performed by external laboratories, i.e.Celabor (Chaineaux, Belgium, ICP-MS and ICP AES), Eurofins (Nazareth, Belgium, ICP-MS) and the Faculty of Bioscience Engineering at Ghent University (ICP-OES).

Experimental procedure
Sludge drying and crushing: The undigested dewatered sludge was dried to a DM content of ≥95% at a temperature between 105°C and 120°C.Prior to drying, the sludge was extruded in the form of spaghetti to improve the drying characteristics which has been confirmed in previous work [27].Sludge samples up to 2 kg were dried in an experimental convective air drier described in [27].For larger samples, a batch forced air circulation drier (Brabender KG300/10) was used.After drying, the sludge was crushed in a laboratory grinder (Waring laboratory blender 3BL40) to a particle size of below 2 mm, stored in air-tight containers, and used for the different experiments.Before each experiment, the DM content of the sample was determined and the corresponding value was used for the evaluations.
Leaching: A weighed amount of sludge was transferred to the leaching vessel and the desired amount of acid solution added.Experiments were carried out in a borosilicate glass bottle in a 360°rotary mixer (Cenco Instrumenten B.V. 3426) at room temperature (between 19 and 23°C) and the leaching time was fixed at 1 h if not indicated otherwise.After leaching, the mixture was separated using a batch centrifuge (Eppendorf 5804) at 4000 rpm for 10 min and filtered using a syringe filter (1.2 µm) to remove any residual trace solids.The aqueous samples were then analysed for P and other elements.
Reactive extraction: Stock solutions of leach liquor and organic extractant phase were placed in a water bath at 25°C to equilibrate the temperature.The desired volumetric phase ratio of organic to aqueous phase (o/a) was transferred to a 250 mL bottle and mixed for 10 min using a magnetic stirrer (Heidolph MR 3001) at 300 rpm.After mixing, the phases were allowed to separate by gravity and then the two phases were carefully separated by means of a syringe with needle by withdrawing the two phases consecutively.The raffinate samples were then analysed to estimate the degree of metal extraction.When Alamine 336 was used as extractant, it was protonated using 2 mol L −1 HCl.The volumetric phase ratio between aqueous phase and Alamine 336 during protonation was fixed based on their molar ratio to 1:2.During extractant stripping or regeneration experiments, the stripped metals may precipitate in the aqueous phase depending on the aqueous-phase pH.Therefore, the regeneration experiments were conducted in a 50 mL centrifuge tube, which was placed in a 360°rotary mixer that was immersed in a temperature-controlled water bath at 25°C for 15 min.The samples were then centrifuged at 4000 rpm for 10 min and the organic phase was removed using a syringe with needle.After separation of the organic phase, 35 mL of 2 mol L −1 HCl solution were added to the aqueous phase in order to acidify the sample and to dissolve any precipitated metals before analysis.For testing the subsequent process steps, reactive extraction was performed with an extractant phase comprising of 10 vol% Alamine 336, 10 vol% TBP, and 3 vol% Exxal 10, dissolved in Ketrul D80.The extraction was repeated once to mimic a 2-stage crosscurrent extraction.
Precipitation of P salts: NaOH and Ca(OH) 2 were used to shift the pH of the raffinate obtained after the 2stage extraction for precipitating P as CaP salts.The precipitation experiments were carried out in a glass beaker and the mixing was achieved with the help of a magnetic stirrer (Heidolph MR 3001) at 200 rpm.NaOH was slowly added using a 50 mL burette or 25% Ca(OH) 2 slurry was added using a pipette until the desired pH was reached.Mixing was further continued for 20 min.Then the mixture was transferred to 50 mL centrifuge tubes and centrifuged at 4000 rpm for 10 min.After centrifugation, the supernatant was filtered using a syringe glass-fibre filter (1.2 µm), acidified using HCl (35%), and analysed.The temperature during the experiments was not controlled.Due to the exothermic neutralisation reaction, the temperature of the reaction mixture increased from room temperature to about 30°C.
Product for quality assessment: To produce a sufficient quantity of recovered P product using the PULSE process, a semi-pilot scale trial was carried out.A new batch of dewatered sludge was dried and crushed as described previously to obtain about 8 kg of dry sludge.A 15 L spherical glass reactor was used for carrying out the different steps of the PULSE process successively.The sludge was leached with 2 mol L −1 HCl at a solid-liquid concentration of 0.25 kg L− 1 for 1 h.A PTFE coated impeller (3 blades, 45°pitched) and an overhead stirrer (CAT R50) were used for mixing.The mixture was separated from the solids by vacuum filtration using a 150 µm polypropylene filter fabric (supplied by SIOEN).Then, 2 mol L −1 of H 2 O 2 (50%) solution was dosed into the liquor and mixed for 1 h.The metal extraction was performed in 2 successive extraction steps at an o/a phase ratio of 1.25.10 L of extractant phase composed of 10 vol% Alamine 336, 10 vol% TBP, and 3 vol% Exxal 10, and 77 vol% Ketrul D80 was prepared which was reused for the 2nd extraction after regeneration.Prior to extraction, the extractant was protonated by contacting it with HCl.The mixing time of the phases was fixed at 10 min after which the phases separated by gravity.The individual phases were withdrawn using the valve at the bottom of the reactor.After the first extraction stage, the extractant was regenerated using a solution containing NH 4 OH and NH 4 HCO 3 at a molar ratio of 1:0.42, a total ammonia concentration of 5.5 mol L −1 , and an o/a phase ratio of 4. To precipitate the P product after extraction, the pH of the raffinate was raised to around 7 by dosing NaOH and mixing for 20 min.The precipitate was separated by vacuum filtration using a grade 4 (Whatman) qualitative filter paper.The filter cake was dried at only 50°C in order to prevent evaporation of the crystal water as recommended by the analytical laboratory.The dried product was ground in a laboratory blender (Waring model no.3BL40) and the sample were supplied for analysis to the Faculty of Bioscience Engineering of Ghent University.
Pot Trials: The pot trials were performed at the Faculty of Bioscience Engineering of Ghent University.To assess the efficiency of the recovered PULSE product and compare it with a commercial mineral fertiliser, triple superphosphate (TSP), pot trials were established to grow perennial ryegrass (Lolium perenne) on slightly acidic (pH H2O = 6.7) river sand (substrate 1) and alkali (pH H2O = 8.5) artificial mineral substrate (substrate 2).The pots were filled with 2.4 L of substrates and amended with 33.7 mg P pot −1 of PULSE product or TSP and 44 mg of ryegrass seeds.The plants were grown for four months.As the plants grow at variable rate, initially the first cut was conducted after five weeks of growth when the plants had already formed the third leaf.For successive cuts, although it typically took one week for plants to reach the level of the pot (mowing height of 4 cm), the cuts were performed only every four weeks to allow the plants to reach a uniform growth stage.The amount of P uptake (mg P pot −1 ) at each cut was evaluated based on the P concentration of the plant shoot DM obtained from the pot.The relative agronomical efficiency RAE of the fertilisers at each cut was then assessed using following equation: RAE = shoot P uptake PULSE shoot P uptake TSP (10) The detailed procedure for the pot trials has been described by Bogdan et al. [28].

Drying and leaching
The sludge collected from Oupeye had an organic matter content typically between 65-70% DM.The average composition of inorganic components in sludge samples is shown in Figure 3.Besides the major components P, Fe, Al, and Ca, various metals are present, some at relatively high concentrations.
Leaching experiments were conducted with both dewatered and dried sludge to determine, which was the better choice for the process.The different appearance of dried and dewatered sludge is shown in Figure 4.The concentration of dried sludge or dewatered sludge (21% DM) in acid solution during leaching was 0.25 kg L −1 .No considerable difference was observed between leaching efficiency of P from dried and dewatered sludge, if leaching was performed at comparable pH around 0.5.Nevertheless, it was observed that dried sludge has the following advantages over dewatered sludge during leaching: . dried sludge is easier to handle and to store for longer periods without deterioration or odour generation, .less acid is required during leaching or more concentrated liquor is obtained with the same acid amount from dried sludge,  .the solids size can be controlled during crushing of the dried sludge, which eases the subsequent solidliquid separation, .the filtered solids after leaching from the dried sludge have a DM content of about 50%.Therefore, it is possible to directly feed them to incineration plants without additional drying since this is typically the DM content to which the dewatered sludge must be dried before incineration.On the other hand, after leaching of sludge that is dewatered but not dried, with DM content around 25% or lower, would require additional drying before incineration.Alternatively, the leached solids may also be used as fuel in kilns for example in cement industry.
Because of these advantages, the sludge is dried in the PULSE process before leaching.
Mineral acids such as HCl, H 2 SO 4 , HNO 3 , and H 3 PO 4 were tested for P dissolution from sludge.Also, here it was observed that leaching is only dependent on the equilibrium pH and not on the type of acid as shown in Figure 5.The maximum leaching efficiency at a pH around zero (2 mol L −1 HCl) was between 65 and 70%, which corresponds to the fraction of inorganic P in the sludge, suggesting that the majority of acid-soluble P is inorganic [11].The P-leaching efficiencies that were estimated using the SLLE tool considering the inorganic components of the sludge are in good agreement with the experimental data.The evaluation of criteria for the selection of leaching acid for PULSE process using COT is presented in Figure 2. HNO 3 and H 3 PO 4 are more expensive compared to HCl and H 2 SO 4 .While the use of HNO 3 for leaching does not have any added benefit, the use of H 3 PO 4 for leaching would be beneficial despite higher cost, as it does not contribute additional ions such as Cl -to the system.However, some of the leaching liquid is lost with the spent solids, leading to significant reduction in the P-recovery efficiency when leached with H 3 PO 4 or excessive washing steps would be required leading to significant P dilution.Therefore, both HNO 3 and H 3 PO 4 were not considered for further study.The leaching acid also influences reactive extraction of metals and, therefore, the compatibility of both HCl and H 2 SO 4 for metal extraction with different extractants was further evaluated and is discussed in the next section.
Experiments carried out at different leaching durations indicated that 90% of the inorganic P was leached in 1 h.Further, the addition of oxidising and reducing agents such as H 2 O 2 or Na 2 SO 3 during leaching with HCl at higher pH between 0.7 and 4, or leaching at elevated temperature of 35°C at a pH of 0.4 did not result in any noticeable improvement.The analysis of Fe 2+ and Fe 3+ in the leach liquor from dried sludge revealed that the fraction of Fe 3+ was ≥50%, which explains the considerable decrease in P leaching from pH 0-1, since P that may exist as Ferric Phosphate (FePO 4 ) has low solubility even at pH of 1 [12].
The concentration of metals and P in the liquor obtained from dried-sludge leaching with 1 mol L −1 HCl is shown in Figure 6.Similar results were obtained with H 2 SO 4 .As can be seen, the concentration of Fe is similar to that of P, hence a direct precipitation of P from the liquor would result in a major fraction of P precipitating as Ferric phosphate which has low soil solubility.Also, the heavy metals would be further concentrated in the product.Therefore, the removal of metals before precipitating P is necessary.

Reactive extraction
The candidate extractants Alamine 336, TBP, and D2EHPA were selected based on their extraction mechanism [16,29] and tested to find the best extractant.The degree of extraction of P and the undesired metals that were extracted from the HCl pregnant leach solution is shown in Figure 7.Only the extractant phase that contained Alamine 336 (A336) yielded significant extraction.In case of Alamine 336 dissolved in Ketrul D80, a third phase is formed after extraction which could be avoided by addition of a modifier that improves the solubility of the target component in the diluent.It was observed that both, TBP and D2EHPA, were able to suppress third-phase formation.However, experiments carried out at higher chloride concentration (>3 mol L -1 HCl) again lead to the formation of a threephase system which was found to be suppressed by addition of Exxal 10, a C10-alcohol, commonly used as a modifier with Alamine 336.
Based on the results, the extractant phase comprising of Alamine 336, TBP, and Exxal 10 in Ketrul D80 as diluent was chosen for further study.A similar extractant phase has also been used in previous studies for extraction of metals from sewage-sludge ash leach liquor [15].Extraction of a liquor obtained by leaching with H 2 SO 4 under otherwise identical conditions was carried out to assess the compatibility with metal extraction which was one of the criteria for COT evaluation as shown in Figure 2, the degree of extraction was less than 0.2 for most of the metals.Therefore, H 2 SO 4 was not considered further for leaching.
The extraction stoichiometry and equilibrium constants of different metal-complexes were determined by fitting them with the SLLE tool to data from equilibrium experiments for varied concentration of the components of extractant phase and leach liquor.During the initial experiments, it was observed that Alamine 336 only extracted Fe 3+ complexes and not Fe 2+ .As the addition of H 2 O 2 resulted in oxidation of Fe 2+ to Fe 3+ , all further experiments were carried out dosing 2 mL L -1 of H 2 O 2 (25%) to the leach liquor before extraction.An example of extraction of metals as function of leach-liquor HCl concentration is shown Figure 8.The increase in HCl concentration had both positive and negative effect on the metal extraction.This is because the metal-chloride complexes formed at different HCl concentrations have varying affinity towards the extractant [30].
On the other hand, the degree of extraction of metals was directly proportional to the concentration of Alamine 336.An example of the dependence of degree of Fe extraction on Alamine 336 concentration is shown in Figure 9 together with the fit obtained using the SLLE tool.It can be seen from the figure that stoichiometry of Fe extraction is a fraction indicating that possibly a mixture of iron-chloride complexes may be extracted.The formation of two iron-chloride and protonated Alamine 336 complexes had also been found in previous studies [31]: For extractant regeneration, Alamine 336, which is an amine-based extractant, was stripped with basic solutions [15,32] such as Na 2 CO 3 , NaHCO 3 , NH 4 HCO 3 , and NH 4 OH.Although stripping efficiencies were similar for the different solutions, it was observed that the equilibrium pH of the aqueous stripping solution was the critical parameter determining the stripping degree and that the ease of phase separation was strongly dependent on the concentration of the alkaline solutions.Salt  concentrations above 2 mol L -1 resulted in better phase separation.The phase separation with NaHCO 3 under different conditions was not optimum and Na 2 CO 3 has low solubility at room temperature.Therefore, it was decided to use combination of NH 4 OH and NH 4 HCO 3 in aqueous solution at a molar ratio of 1:0.42 for stripping.The higher concentration of salts in the stripping solution also allows to operate at higher organic-toaqueous phase ratio and recycle the aqueous phase in the process to generate a concentrated stream of metals.
How the stripping depends on pH is shown in Figure 10.In principle, the degree of stripping increases with increasing pH.The stripping efficiencies computed based on the fitted parameters using the SLLE tool were in good agreement with the experimental data.The SLLE-simulation data revealed that there was a complex formation between TBP and metals above pH 2. Therefore, stripping of Cd and Zn was relatively lower than that of the other metals even at high pH.
During stripping, a golden-brown precipitate was obtained at pH > 3, which was analysed to contain mostly Fe, while the other metals remained dissolved in the aqueous phase.Therefore, it is possible to further separate and valorise the metals after stripping.

Precipitation of P salts
The type of phosphate salt precipitated is a function of concentration of P, concentration of metal ions present in the solution, and pH.The concentration of metals in the raffinate is very low compared to P concentration except for calcium.Therefore, increasing the pH of the raffinate would result in precipitation of P predominantly as CaP.Different CaP salts can precipitate depending on the precipitation parameters such as Ca/ P molar ratio, pH, temperature, and reaction time [33].The precipitation of P and other elements as a function of pH using 5 mol L -1 NaOH solution is shown in Figure 11.
Almost all P was precipitated at pH ≥ 6.It can be seen that at pH 4 about 30% of P was precipitated, while the amount of Ca precipitated is very low.This can be attributed to the precipitation of aluminum phosphate, which precipitates already at low pH, as Al is not removed during reactive extraction.About 70% of P is then precipitated as CaP salts.It can also be seen from the results that Al starts to re-dissolve again at higher pH.So, it is possible to separate Al if required by fractional precipitation or dissolution [15].However, as its concentration in the sludge liquor was relatively low, this option was not evaluated in detail.The simulation of the CaP   precipitation using the SLLE tool was carried out by considering the Ca/P molar ratio and pH from experimental data.A good agreement between experimental and simulation data is obtained.Similar results were also obtained in case of Ca(OH) 2 .While the price for NaOH and Ca(OH) 2 are not very different, the use of NaOH may lead to undesirably high concentration of Na in the product.Therefore, it is preferable to use Ca(OH) 2 or a combination of NaOH and Ca(OH) 2 for precipitation.

PULSE product quality
The P-leaching efficiency during the semi-pilot trial was 69%.The recovered product had a P mass fraction of 0.11 in DM as compared to only 0.02 in dry sludge.The P content in the recovered is thus higher than the minimum P mass fraction of 0.069 required by the newly proposed limits for the secondary P salts under the new EU regulation [34].If also phosphate, calcium, and magnesium of the PULSE product are considered as nutrient, the product has a nutrient mass fraction of about 51%.
The reduction ratio of the inorganic component content per gram of P from sludge to product is presented in Figure 12.Significant decontamination was reached by reactive extraction for detrimental metals such as Fe, Cd, Cu, Pb, and Zn.The concentrations of heavy metals measured in the product were lower than the limits defined in the revised EU fertiliser regulation [35] with the exception of Cr.An unusually high amount of Cr was detected in the Oupeye sludge, which was also transferred to the product as it could not be separated in the reactive extraction process under the given process conditions.The Cr in sludge originates from an industry located in the Oupeye wwtp catchment.As this is an uncommon situation, the process was not specifically adapted to handle this issue although it would be possible with reactive extraction to selectively separate Cr as well.It is also observed that Ca, which is also a nutrient, is transferred to the product as it is not removed during the intermediate process steps.The content of Na increases in the product as NaOH is added to the raffinate during precipitation.Here, the product composition is reported before a final washing step, which significantly reduces the Na concentration.
To characterise the fertiliser quality of the PULSE product, its RAE is presented in Figure 13 showing values for both substrates and all samples around 100%, some even clearly above.This high RAE achieved in the present study suggests that the recovered P product consisted of more soluble primary and/or secondary P forms, available for plant uptake [28].The RAE of the PULSE product indicates that it is equally efficient as compared to TSP on both slightly acidic and alkaline substrate and confirms that the PULSE product has similar efficiency as commercial mineral fertilisers on various soils [28].

Conclusions
This study presents the new PULSE process for the recovery of P from sludge with varying metal content without compromising the recovery efficiency to produce a highquality secondary-P raw material that can be used as fertiliser or fertiliser constituent.The overall P-recovery with the PULSE process is mainly determined by P dissolution from sludge which is a function of inorganic P fraction and the leaching pH.Drying the sludge prior to leaching offers several benefits over leaching wet sludge directly such as ease of dewatering of leached sludge and reduction in acid consumption.The additional cost for sludge drying is partially compensated by the reduction  Reduction ratio relative to P (g g −1 P) between product and sludge and minimum reduction legally required for contaminants [35].
in acid consumption during leaching and can further be reduced by utilising green alternatives such biogas or solar drying.Further, the spent sludge that has a DM of about 50%, mostly containing organic matter can be used as renewable fuel for kilns, e.g. in the cement industry.
It has been shown through the lab experiments that metals such as Fe, Cd, Cu, Pb and Zn were from the sludge leach liquor prior to precipitation of P in the PULSE process via reactive extraction.Therefore, the process allows the selective removal of several metals with negligible P loss and can be further enhanced by increasing the number of extraction stages to handle higher heavy-metal loads or by using specific extractants to deal with the particularities of the sludge and obtain the desired level of metal depollution.Some of the extracted metals such as Fe, Cu, Zn etc. can be further separated and valorised from the concentrated effluent of the PULSE process.The CaP salt precipitated from the leach liquor after metal removal in the PULSE process has a plant availability that is comparable to TSP and a nutrient mass fraction of about 51%.The spent liquor after precipitation contains only trace levels of P and metals and may be recycled in the process again for leaching or sent to wwtp without the concern of accumulation of metals in the system.
The PULSE process was validated at pilot scale also with sludge from Belgium, Germany, and Scotland [9].The resource consumption evaluated based on the pilot scale results indicates that the P-recovery by the PULSE process would result in an additional treatment cost of about 5-10 €-cents per m 3 of wastewater, depending on P content of the sludge, not accounting for the economic benefits from selling the P, leached sludge as fuel to cement kilns, or other components and without further optimisation of the drying as indicated.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The research work was carried out as part of the PhosForYou project which received funding from the INTERREG VB North-West Europe Programme (2014-2020) under grant NWE292, as well as from the Service Public de Wallonie, and the Univeristy of Liège.

Figure 2 .
Figure 2. COT for evaluation of the choice of acid for leaching.

Figure 3 .
Figure 3. Composition of Oupeye sludge on dry matter basis.

Figure 5 .
Figure 5. Degree of P leaching from dried sludge with different acids as a function of pH.Sludge DM: 96.1%, solid-liquid concentration: 0.25 kg L −1 .

Figure 11 .
Figure 11.Precipitation of P and other metals from leach liquor using 5 mol L −1 NaOH.

Figure 13 .
Figure 13.RAE of the PULSE product after 4 months of plant growth in two substrates.

Figure 12 .
Figure 12.Reduction ratio relative to P (g g −1 P) between product and sludge and minimum reduction legally required for contaminants[35].