A systematic comparison of commercially produced struvite: Quantities, qualities and soil-maize phosphorus availability

P-containing secondary raw materials such as struvite, the EU issued a revised fertilizer regulation in 2019.Acomprehensiveoverviewofthesupplyofstruviteanditsquality ispresently missing.This study aimed: i) to determine the current EU struvite production volumes, ii) to evaluate all legislated physicochemical characteristics and pathogen content of European struvite against newly set regulatory limits, and iii) to compare not-regulated struvite characteristics. It is estimated that in 2020, between 990 and 1250 ton P are recovered as struvite in the EU. Struvite from 24 European production plants, accounting for 30% of the 80 struvite installations worldwide was sampled. Three samples failed the physicochemical legal limits; one had a P content of <7% and three exceeded the organic carbon content of 3% dry weight (DW). Mineralogical analysis revealed that six samples had a struvite content of 80 – 90% DW, and 13 samples a content of >90% DW. All samples showed a heavy metal content below the legal limits. Microbiological analyses indicated that struvite may exceed certain legal limits. Differences in morphology and particle sizedistributionwereobserved forstruvitesourcedfromdigestate(rodshaped;transparent;82 mass% < 1 mm), dewatering liquor (spherical; opaque; 65 mass% 1 – 2 mm) and ef ﬂ uent from up ﬂ ow anaerobic sludge blanket reactor processing potato wastewater (spherical; opaque; 51 mass% < 1 mm and 34 mass% > 2


H I G H L I G H T S
• Struvite from 30% of the 80 worldwide installations were analyzed in detail.• All but three struvite samples meet all EU fertilizer directive standards.• Struvite granule size and shape highly depend on substrate type to recover from.• P availability and plant biomass yields were similar across struvite samples.• Current struvite production is limited, demonstrating large potential for growth.

G R A P H I C A L A B S T R A C T
a b s t r a c t a r t i c l e i n f o 1. Introduction Phosphorus (P) is an essential macronutrient for all living organisms, classified as a critical raw material by the European Commission (EC, 2020).About 90% of commercially available P is sourced from phosphate rock, a non-renewable and geographically restricted resource, with no meaningful reserves in the European Union (EU) (Chowdhury et al., 2017).To counter resource dependency and to realize sustainable P management, implementing P recovery is imperative.In society, the largest P losses occur through domestic wastewater (15% of total EU import) (van Dijk et al., 2016).In this regard, P recovery through struvite precipitation from wastewater (MgNH 4 PO 4 .6H 2 O) has received much attention.The reasons for this are that struvite has a theoretical P content close to that of phosphate rock (12.6% dry weight (DW)), it has been demonstrated to be an effective P fertilizer especially in acidic soils, and it is considered a slow release fertilizer that can reduce P losses to the environment (Everaert et al., 2018;Hertzberger et al., 2020).Furthermore, our research shows that struvite is currently mainly marketed at values between €0-100 per ton, but also at considerably higher prices of €350 per ton (Phoshorgreen) to €1000 (Pearl) per ton.This implies that in about 80-87% of the cases struvite is sold at lower prices than the estimated market value of its macro nutrients, namely €250-412 per ton (first value: literature De Vrieze et al. ( 2019) assuming €1900 t P fertilizer; own calculations: €549 t urea 47% N and 5.7% N in struvite and the value of €233 t single super phosphate (SSP) at 7.7% Pat farm price in Austria September 2020 excl.VAT -personal communication Timac AGRO).Finally, struvite production technologies have matured since their first application (Egle et al., 2016).Shaddel et al. (2019) identified 80 struvite producing plants world-wide, operating 19 different technologies.Three basic types of struvite precipitation processes can be distinguished (Fig. 1).Type I receives waste activated sludge (WAS and primary sludge if applicable) and/or digestate originating from anaerobic digestion (AD) in continuous stirred tank type reactors (CSTR).Type II struvite is precipitated on the dewatering liquor of the WAS digestate after a solid-liquid separation step (e.g.centrifugation).A key difference between the two streams is the lower total suspended solids (TSS) content in the influent to the struvite reactor (e.g.21-28 g TSS/L digestate vs.0.9-1 g TSS/L dewatering liquor, as measured in this study).In several cases, type II struvite reactors also receive a Prich liquor from WAS thickening prior to anaerobic digestion (Fig. 1).Type III struvite is sourced from agro-industrial wastewater treatment (i.e.potato processing, dairy).The main application is the potato industry (Abma et al., 2010;Moerman et al., 2009); in these applications struvite is precipitated directly from the liquid effluent of anaerobic digestion in upflow anaerobic sludge blanket (UASB) reactors.These liquid effluents typically are relatively low in TSS content (e.g.0.5 g TSS/L) (Muys et al., 2020).The different influent types and technologies implemented may affect struvite quality as struvite crystallization is governed by several interacting parameters including, amongst others, influent P concentration, crystal retention time, TSS concentration, viscosity, presence of colloidal substance, Mg:P ratio, pH and mixing conditions (Doyle and Parsons, 2002).As a result, the produced struvite is likely to vary in crystal size, presence of co-precipitates and inclusion of organics and other contaminants.However, currently there is no systematic comparison of struvite quality from different full-scale installations.Yet, such an analysis is imperative in the context of the recent publication of the new European fertilizer regulation that is setting EU-wide quality standards for struvite and hereby facilitating its EU-wide trade (EC, 2019).
The new fertilizer regulation defines 17 physicochemical and 5 microbiological parameters that struvite should meet to be utilized as a fertilizer or component material in N-P-K fertilizers (supplemental material (SM), Table S1).In addition to the regulation, it is proposed that physical parameters and the solubilization rate of struvite are critical for further commercial use of struvite (Huygens et al., 2017).Struvite granules must be of a shape and size which enable its utilization in modern application equipment or allow blending with N and K containing granules (Spiller et al., 2019).Alternatively, struvite with very small and heterogeneous granules may be more suitable for application in growing media or can be re-processed into an N-P-K fertilizer (Grunert et al., 2019;Spiller et al., 2019).Furthermore, it is known that struvite is not purely comprised of MgNH 4 PO 4 .6H 2 O, but does contain other precipitates (e.g.brushite, dittmarite, amorphous components).These precipitates do not only have a different N and P content, but also have different solubilization characteristics.Furthermore, it was shown that plants influence struvite P solubility through producing exudates (e.g.organic acids) and that therefore an investigation of P solubilization for the system comprising plant and soil is required (Talboys et al., 2016).
Answering to the knowledge gaps outlined, this study had three objectives.Firstly, to estimate the current EU struvite production volumes.Secondly, to compare the physicochemical and microbiological characteristics of struvite samples from 24 installations across the EU and to evaluate them against the new European fertilizer regulation.Thirdly, to provide insights on parameters not regulated for, that are known to affect utilization and fertilization characteristics of struvite, including mineralogical composition, particle size distribution and to study soilplant P-availability and the fertilization performance.

Struvite sampling
In this study, information about all known full-scale struvite installations was gathered through literature review, industrial contacts, web search and interaction at international conferences.In total 39 operational installations were identified (source separated decentralized installations were excluded, Table S2, SM Section 4) and 25 struvite samples were collected from 24 different installations across Europe (Table 1).All but one sample were collected in the period 2018-2019.For every sample collected, process engineers and technical staff were interviewed to gain insight into operational conditions of the struvite precipitation process.The samples are classified into the three main struvite recovery categories introduced above.

Physicochemical struvite characterization
Before the physicochemical analyses, all samples were oven dried, the drying was performed at a temperature of 40 °C (Huygens et al., 2017).Subsequently, samples where further prepared for the different analyses as described in the SM Section 3. Phosphorus was extracted according the EC, 2003/2003method 3 and 3.1.1 (EC, 2003) and measured according the spectrophotometric vanadomolybdate method (APHA et al., 2012).Ammonium was analyzed using the Kjeldahl method (SM Section 3).For the total organic carbon (TOC) analysis, the struvite was dissolved in distilled water and determined using an online TOC-V series Shimadzu analyzer (TOC-VCPH autosampler ASI).Heavy metals and cations were analyzed using Inductively Coupled Plasmaoptical emission spectrometry (ICP-OES; 5110 VDV Agilent Technologies, USA).Potassium and magnesium were extracted according the standardized method EC, 2003/2003method 4.1 and 8.1 (EC, 2003).Struvite bulk mineralogical composition was measured by X-ray diffraction (XRD; using CuKα radiation; Bruker D8 Advance diffractometer and PDF-4 crystal structure database of the International Centre for Diffraction Data).A mass balance was constructed integrating data from the different analysis outlined above (SM Fig. S1).Struvite content was estimated using both the mineralogical data from the XRD and data on the elemental composition (SM Fig. S1).A particle size distribution (PSD) of all struvite samples was made using sieve analysis (mesh sizes of 200,500,1000,1600,2000,2500,3150,3550 and 4500 μm) (norme Française; European norm, NF EN 1235).

Microbiological struvite characterization
Pathogenic microorganisms were determined for three fresh samples of Airprex 1, Pearl 1 and NuReSys 4 (Table 1).To explore the effect of storage on pathogen contamination the samples of Airprex 1 and Phosgreen 2 were kept for 9 months at room temperature in the dark.To determine the distribution of pathogens between struvite crystals and impurities, the visible debris of the fresh Airprex 1 was removed manually before a second pathogen analysis was performed.
The STRUBIAS document sets limits for selected pathogens, e.g.Salmonella spp., E. coli or Enterococcaceae, Clostridium perfringens, Ascaris sp.(Huygens et al., 2017) (Table S1 SM).The pathogens analyzed in this study deviate from the STRUBIAS indicators in that no analysis of Salmonella was performed and that the analysis of spores of sulphide reducing clostridia (SSRC) is used as an indicator for C. perfrigens, Ascaris sp.eggs and Cryptosporidium oocysts due to their persistence to wastewater treatment processes (Guzman et al., 2007).Furthermore, Fspecific RNA phages were used as an indicator for enteric viruses (Guzman et al., 2007).These analytical approaches were followed because defined methodological procedures were not available at the time of the study and no laboratories were available to analyze the required parameters.
For all biological contaminants, 7 g of sample was brought into 70 mL sterile tap water.The sample was sonicated for 2 min, after which the supernatant was separated in a sterile flask.Again 70 mL sterile tap water was added to the struvite and sonicated.These steps were repeated for 6 times in total.The collected supernatant was used for all microbiological analyses.SSRC was determined according NEN 6567 and F-specific RNA bacteriophages according ISO 10705 part 1, coliform bacteria, E. coli and Enterococcaceae were determined according NEN-EN-ISO 9308 part 1 and NEN-EN-ISO 7899 part 2, respectively, after membrane filtration.

Struvite P availability in soil-maize seedling pot tests
Phosphorus availability from struvite was measured in pot experiments with maize seedlings for eight struvite samples (Table 1).Controls with 12.9% (P) rock phosphate (RP) and 6.9% (P) Single Super Phosphate (SSP) and a negative control with no additional P were prepared in three pots (12 × 12 × 12 cm) with 1200 g of a natural loamyclayey soil (soil 8) with low available P (0.009 g/kg P, Olsen) and pH 5.8.To exclude effects of different granule sizes, fertilizers and struvite samples were ground to a powder and added to the soil to contain 0.017 g P/kg soil.Soils were vigorously mixed, transferred to pots and in each center a maize seed (Zea mays L. cv.'Dulcano') was placed at 2.5 cm depth.Each pot was ameliorated with 0.16 g (NH 4 ) 2 SO 4 -N/ kg soil and 0.33 g -K 2 SO 4 -K/kg soil dissolved in 50 mL of a tap water.Pots were kept in a glass house under a 16 h light/8 h dark regime, Table 1 Overview of analyzed struvite samples.Information on process/technology, origin, recovery category (see Fig. 1: type I, II and III) and operational parameters are given, while samples are indicated that were analyzed for microbiology and used in maize seedling tests.'-': Unknown.HHNK -Hoogheemraadschap Hollands Noorderkwartier, FBR -Fluidized bed reactor, CSTR -Continuously stirred tank reactor, (A full list of all currently known European struvite installations, including production volumes, is given in Table S3, SM). and temperatures between 19 °C (night) and 28 °C (day).Plants were regularly watered (max 200 mL), ensuring full adsorption by the soil.Soil samples for P analysis in diffusive gradients in thin films (DGT) were taken from each pot (six replicates in total) at 1, 3, 7, 14 and 28 days after germination.Soil samples were taken at the distance of 2 cm from the seed/plant using a 50 mL tube, which was pressed into the soil to the depth of 5 cm.The retrieved soil column was mixed, and 5 g soil were analyzed after drying at 37 °C for 24 h.Diffusive gels (polyacrylamide hydrogel) and the binding layer with ferrihydrite precipitated into a thin diffusive gel (ferrihydrite gel) were prepared according to Zhang and Davison (1995) and Santner et al. (2010), respectively.Soil paste of all samples with 90% water holding capacity (WHC) was placed on the DGT gels for 24 h under water-saturated air at 20 °C.After removing soils, P from gels were eluted with 2 mL 0.25 M H 2 SO 4 in 3 h and P was determined in the eluate with the molybdenum blue method and determined spectrophotometrically (Estefan et al., 2013).Data points were subdued to two-way ANOVA analyses followed by Tukey Honest Significant Difference Comparisons (adjusted p values <0.01).Maize biomass was determined after 28 days, by cutting all maize stems at the position where root outgrow starts, followed by drying for 48 h at 55 °C.Data were analyzed by ANOVA and Tukey Honest Significant Difference Comparison (adjusted p values <0.01).

Struvite installations and production volumes in Europe
In the present study 24 of a total of 39 identified operational European struvite installations were sampled, accounting together for between 59 and 67% of the struvite produced (excluding plants not operational at the time of the fieldwork but those to be commissioned in 2020, and those not separating struvite from solids, for details on methodology see SM Section 4 and Table S2).The installations identified include 29 municipal WWTP, 9 potato industry WWTP and one dairy WWTP.At these plants, 8835-11,100 tons of struvite are produced annually (Table 2), constituting an equivalent of about 1030-1350 tons P in total and between 1000 and 1250 tons struvite P that meets the legal requirement (i.e.excluding three sample as show in Section 3.2; for details on methodology see SM Section 4 and Table S2).Between 64 and 71% of struvite is recovered from municipal wastewater and 25-27% from potato wastewater.In Europe, the struvite market is currently dominated by the technologies of Airprex, NuReSys and Anphos, which are anticipated to produce 68-74% of the struvite in 2020 (including plants that do not separate struvite from sludge for Airprex).Most of the struvite is produced in the Netherlands (35-43%), Belgium (16-20%) and Germany (15%).

Struvite constituents (P, N, Mg, K) and organic carbon
Current legislation prescribes a minimum P content within the precipitated phosphate salt of 7% DW (EC, 2019).Only one sample does not comply with this limit (Anphos 3), while 12 out of 25 samples deviate not more than 3% from the theoretical value of 12.6% P DW (Fig. 2).The source of struvite influences the P content.Struvite recovered from the dewatering liquor of digestate (type II) contains more P with a mean of 12.1% DW (standard error of mean (SEM): 0.29), while digestate has a lower P content with a mean of 11.3% (SEM: 0.24).Struvite from UASB effluent has a mean value of 10.2%, but it shows large variability (5.6-12.9%;SEM: 1.14).The NuReSys 4 and 5 samples as well as the Anphos 5 sample show high P content of >12.3%, while the Phospaq and Anphos 4 samples only show values of 8.6% and 7.3%, respectively.Nearly all struvite samples that have a P content close to the theoretical value also contain N and Mg close to the theoretical value of 5.7% N and 9.8% Mg (DW) (Fig. 2) (Phosphgreen 1, 2 and 3; Pearl 1, 2 and 3; NuReSys 2, 3, 4 and 5; Anphos 1, 2 and 5).For potassium (K), samples from municipal wastewater treatment show values <0.2% DW (Fig. 4), whereas all struvite samples that originate from UASB effluent (type III) show at least a 4.5 times higher K content (0.9-1.25% DW).
Of the 25 samples, 22 have a total organic carbon (TOC) content below the legal limit of 3% DW (Fig. 3), with values between 0.12 and 1.6% DW.Highest values were observed for struvite recovered from UASB effluent (10.8-13.2%DW TOC for Phospaq, Anphos 3 and 4).

Heavy metal content
Struvite heavy metal concentrations are well below the legal limit and often even below the detection limit (Table 3).It is suggested that heavy metal contamination can be linked to total organic matter (TOM) content (Huygens et al., 2017).This can be confirmed for the samples Phospaq, Anphos 3 and 4.However, not all the analyzed samples follow this trend, as e.g.Airprex 1 and NuReSys 3 are low in organic carbon, but high in Zn, Pb and Cu content (latter two Airprex 1 only), compared to the other samples.Struvite originating from the dewatering liquor of digestate (type II), generally shows the lowest heavy metal concentrations.Whereas samples from type I have higher heavy metal concentrations, here Zn and Cu concentrations are up to an order of magnitude higher than in most type II samples.The heavy metal content of Type III samples varies widely, with, for example, Zn concentration ranging from 2.1 to 550 mg/kg.

Elemental and mineralogical composition
Based on the elemental analysis, the mass balances indicate that for 19 of the 25 samples, the struvite content (i.e.magnesium ammonium phosphate) is higher than 80% DW (six samples below this value).In contrast, the mineralogical results reveal that 10 samples are below this value (Fig. 4).Based on mineralogical data, 6 samples show a struvite content between 80 and 90% DW, while 13 samples have a struvite content of >90% DW.Samples with a lower struvite content often contain elevated amounts of quartz (SiO 2 ), which is an important fraction of sand.Seven out of eight quartz containing samples are Fig. 3. Total organic carbon (TOC), and total organic matter (TOM) content, for all struvite samples.TOM was calculated as TOC/0.56.Dotted line represents the legal limit of 3% DW.

Table 3
Heavy metal content of the analyzed struvite samples (Empty cells: measured content < limit of detection (LOD).N.A.:Not Applicable).Color fill of the cells indicates to which extend the regulatory limit is approached or if no regulation applies.Cr N.A. 1 2.9 1.6 3.3 4.8 2 6.2 7 2.5 6.8 13 2.7 7.5 9 5.4 5.7 Type I: digestate Type II: dewatering liquor of digestate Type III: UASB effluent produced on digestate (type I) (i.e.Airprex 1, 2, 3, 4 and 5; Nuresys 1 and HHNK 1 and Vecht).Furthermore, a P, N and Mg content close to the theoretical composition of struvite, does not guarantee a high struvite content, as in six samples (Phosgreen 3; Phospaq; Pearl 1, 2; Airprex 4; Anphos 4) a large fraction of amorphous phase is observed.Minerals containing Mg include periclase (3 samples), newberyte (1 sample), dittmarite (1 sample), eitelite (2 samples), nesquehonite (2 Samples) and magnesite (1 sample).Minerals containing Ca are not frequently encountered, only two samples contained monohydrocalcite in small amounts (Anphos 3: 7.4%; Anphos 4: 3.4%) and one sample contained brushite (NuReSys 1: 2.9%).Precipitates with Al are found in two samples of Pearl as rostite (Pearl 2: 6.5% and Pearl 3: 3.6%) and in the sample of HHNK (HHNK 1: 2.8%) as plagioclase.Potassium (K) was present as K-feldspar in one sample (HHNK).In other K containing samples no mineral precipitation of K could be detected.It is known that struvite can co-precipitate with K-struvite (KMgPO 4 .6H 2 O), albeit with preference for struvite because of its lower solubility (Xu et al., 2011).However, K-struvite could not be reliably be quantified as it could not be distinguished from struvite in the XRD analysis due to overlapping diffraction positions.Similarly, the integration of Kstruvite into the mass balance had either almost no effect due to low K values, or was not plausible due to a lack of Mg 2+ or PO 4 − in the mass balance.Therefore, while K struvite might be present (type I max 1.1%, type II max 1.2%, type III max 8.5%) a reliable quantification of its content in not possible.

Physical properties of struvite samples
Three struvite samples (Anphos 2, 4 and Naskeo; Fig. 5 and SM Section 6) exceed the limit for particle size distribution (PSD) of maximum 10% < 0.1 mm (note: not a legal limit defined in the revised fertilizer regulation, but proposed in an early version of the STRUBIAS working group report (Huygens et al., 2017)).Across all sampled struvite, the size range of <0.2 mm accounts for an average of 9.8% of the particles (standard error of mean (SEM): 3.9), and the majority of the particles have a size between 0.2 and 2 mm (mean: 78.5%, SEM: 6.1).Sizes above >2 mm accounted for an average of 11.7% of the particles (SEM: 4.8) (Fig. 5, SM Table S2).The largest granules are found for Anphos 5 with nearly 80% of the sample between 3.15 and 3.55 mm and Anphos 4 with 65% of the sample > 2 mm.In addition, also Phospaq contains a mass fraction of 44.5% of the sample above 2 mm.However, Anphos 3 and 4 as well as Phospaq are heterogeneous samples with no peak in the analyzed size range (Fig. 5, SM Section 6).
Investigation by struvite type shows that for digestate (type I) 81.9% of DW constitutes of granules smaller than 1 mm, while only 13.7% of the granules fall in the range of 1-2 mm and 4.4% are > 2 mm.The inverse can be found for type II, here 64.6% of the granules are between 1 and 2 mm, 33.1% of the material is <1 mm and the remainder of 2.3% is >2 mm.For type III struvite, 50.7% of the granules are smaller than 1 mm, 15.5% are between 1 and 2 mm and 33.9% are larger than 2 mm (Fig. 5, SM Table S3).The overall variability of particle sizes is highest for the type III samples, which show a distribution of grain size across the entire spectrum in many samples.Type II samples are more evenly distributed, while type I samples show the lowest variability (Fig. 5).
Visually, struvite can be classified into two types: transparent samples from digestate (type I) and opaque from the dewatering liquor of digestate and other origins (type II and III) (Fig. 5, SM Section 6).As reported also by Ping et al. ( 2016), we found that most struvite crystals  3. XRD data are presented in Supplementary Material Fig. S4, raw data presented in SM Table 4.

Microbiological quality
Most of the pathogens and indicator organisms are present at low levels in the analyzed struvite samples.Struvite recovered from digestate (type I) contains a similar amount of Enterococcaceae (2 log cfu/g) compared to struvite recovered from dewatering liquor (type II), while struvite originating from potato processing effluent (type III) has a concentration of Enterococcaceae of 4 log cfu/g, violating the limit of 3 log cfu/g struvite.For SSRC the fresh samples from type I and II struvite have concentrations between 2.8 and > 4.3 log cfu/g.However, since SSRC is only a proxy indicator, this does not necessarily imply that the legal limits for the bacterium C. perfringens (2 log cfu/g) and Ascaris sp.eggs (1.4 log cfu/g) are violated.
Comparing the values for Airprex with and without impurities, it can be observed that most of the E. coli (>90%) and SSRC (94%) are located on the impurities, while this does not apply to the Enterococcaceae (30%) (Table 4).Pathogen numbers in the struvite samples of Pearl 1 and Airprex 1 are several logs lower compared to the dewatering liquor and the digestate they are produced from (Table 4).Therefore, it can be concluded that struvite crystallization selectively excludes pathogens, leaving them in the water phase or kills them off during the crystallization process (e.g.elevated pH).Furthermore, the reduction of 0.8-2.5 log units in the 9 month-stored sample of Airprex compared to the fresh sample, indicates that storage can be a good strategy to reduce pathogens.Considering overall P availability, 7 of the 8 tested struvite samples are not significantly different from each other; while one struvite sample (Phosgreen 3) differs from at least 4 other samples by lower overall P availability (Fig. 6).It can be observed that P availability patterns are rather similar, with a continuous availability over the sample period (3.5-6.5 mg P/L of soil, mean: 4.8 mg P/L).As expected, P availability from SSP peaks at day one (9.09mg P/L) and then declines.Starting on day seven, P availability of struvite is not significantly different from SSP (SM Table S5).Rock Phosphate (RP) P availability is significantly lower than that of struvite and increases only marginally over the sampling period.Regarding maize seedling biomass, no significant difference is observed between struvite samples and SSP fertilization (Fig. 6B).However, due to the short sampling period significant differences between soil only, SSP and RP could not be achieved.Nonetheless, the data indicates a trend towards a better performance of struvite and SSP (SM Table S6).

Struvite installations and production volumes in Europe
Of the 39 operational European struvite production sites, an estimated equivalent of 1000-1250 tons P is available that meets legal requirements (Section 3.1).This corresponds to 0.5% of the total P theoretically contained in EU wastewater or 0.06-0.07% of the EU P imported for fertilizer use in 2017 (estimated based on a population of 508 million in EU-28, a discharge of 1.2 g P/person/day and an import of 1.725 million ton P/year (FAO, 2020)).In other words, currently recovered amounts are very low compared to the overall P flows and the demand of agriculture.Moreover, recovery of struvite from municipal wastewater has an efficiency (P struvite /P total-plant-influent ) of between 20 and 43% (Amann et al., 2018;Kazadi Mbamba et al., 2016).Assuming a maximum struvite recovery efficiency of 43%, this would supply about 13% of the P-fertilizer demand in the EU.

Elemental and mineralogical composition
This study demonstrates that 22 struvite samples have a chemical quality that is within the legal limits set by the new EU fertilizer regulation.The only legal infringements of the chemical composition can be found for the organic carbon content of three samples (Phospaq, Anphos 3, 4), of which one also failed the minimum required P content (Anphos 3).Another observation is that eight samples contain relatively high amounts of amorphous precipitates.Amorphous calcium phosphate in the crystals was described at high Ca concentrations in the solution of up to 4 mM (Le Corre et al., 2005;Pastor et al., 2010).Furthermore, above a reactor pH of 9.5, it was found that the struvite content decreased to <30% and amorphous P precipitates increased.In addition, it was observed that struvite crystal structure changes at temperatures above 64 °C (Doyle et al., 2002), potentially increasing the number of amorphous components.However, none of these reasons can consistently be aligned with our findings.The samples that contain the highest amount of Ca also contain amorphous substances (Anphos Fig. 6.A) Available P in days after germination in maize seedlings pots fertilized with struvite, rock phosphate (RP), single super phosphate (SSP) or without additional P fertilization (soil only).P collected in 24 h hours with DGT after the indicated time points are shown.P application was normalized to the P content of each sample so that in each pot the same amount of P was applied.B) Maize seedling dry weight after 28 days.Letter codes indicate significant differences between overall release patterns (i.e.different letters represent significant differences p < 0.01), for example letter combinations containing the latter 'a' are not significantly different from each other, while a combination not containing the letter 'a' is significantly different.4, Phospaq), but other samples with a similar Ca-content do not contain amorphous substances (Naskeo, Anphos 3).For the Pearl and the Phosgreen processes, struvite is heated (40-55 and 55-75 °C, respectively), but not all of these processes show amorphous components.
Values for the pH were generally well below 9.5 in the reactors.Another observation is that type III struvite has a 4.5 times higher K content than other struvite types, which is likely a result of the higher K content in the potato wastewater (around 1800 mg K/L), compared to municipal wastewater (10-30 mg K/L) (Arienzo et al., 2009).Similarly, quartz is abundant in the type I digestate.The quartz content is relatively high for these samples as in the other processes it is removed in the solid-liquid separation that is preceding the struvite reactor.Furthermore, it was found that there is a tendency towards a higher heavy metal content for type I struvite.Heavy metals are known to attach to organic matter, however, no correlation between heavy metal content and TOC was found in the struvite of type I. Nonetheless, it is plausible that struvite precipitated from a matrix that contains comparative high concentrations of organic matter would show higher values of heavy metals.

Physical properties
The particle sizes ranged from <0.2 to >4.5 mm, therefore larger samples (Anphos 5 only) were found in this research than previously described in literature (i.e.0.15 mm (Zhang et al., 2009) to 3.5 mm (Adnan et al., 2003)).Granules from the dewatering liquor of digestate are generally larger than those from digestate (type I).This may be explained by the higher TSS concentration in the struvite reactors of type I operating on digestate (25 ± 5 g TSS/L (Airprex 1) compared to dewatering liquor (1 ± 0.1 g TSS/L (Pearl 1).Ping et al. (2016) found that at elevated TSS concentrations, struvite crystals are smaller.They suggest that TSS interferes with the crystal growth by reducing the aggregation (i.e. the collision of two crystals and their adherence together) of crystals.The TSS content in UASB effluent is comparatively low (0.5 g TSS/L (Muys et al., 2020)) and should not prevent crystal aggregation for type III struvite.In fact, the largest struvite granules were found for this type of struvite, but the variability in size distribution was highest for these sample, ranging from <0.2 mm to >4.5 mm.Likely, complex interactions between other determining parameters play a role here including: agitation, crystal residence time, presence of foreign ions, pH as well as the Mg:P ratio.However, due to the interactions of different parameters and the limited availability, reliable conclusions about the effect of these processes on particle size cannot be made.
In addition to the crystal size, the visual difference between the shape and appearance (i.e.transparent and opaque) of type I, II and III were observed (Fig. 5).An explanation for the differences comes from Ye et al. (2018), they describe an evolution of crystal shapes in a fluidized bed reactor, from a primary nucleus followed by crystal growth to 'young' rod-shaped crystals and subsequent aggregation of individual crystals.This is further followed by agglomeration or the formation of clusters of aggregates in the middle section of the fluidized bed.Finally, clusters sink to the bottom of the fluidized bed where they are coated with primary crystallization nuclei (i.e.crystal coalescence).Throughout this evolution, crystals become more spherical with a smoother surface.As this study observed spherical struvite granules with a smoother surface from fluidized bed reactors, CSTRs and aerated tanks, it could be possible that, independent of the reactor type, similar processes as described by Ye et al. (2018) take place.This appears plausible as every type of agitation will create a fluidized bed of solid particles.However, the conclusion about the particle character should be first validated by providing a transect of struvite granules as done by Ye et al. (2018).With exceptions (i.e.samples mainly <0.2 mm), Type II and III would then correspond to agglomerated clusters and coated granules.Ye et al.'s (2018) classification could also suggest that type I struvite comprises crystals which may be unable to agglomerate or break apart easily once agglomerated as suggested by Ping et al. (2016).According to Ping et al. (2016) the reason for this attrition lies in the elevated TSS concentrations, which also has been observed in other studies (Tarragó et al., 2018;Ye et al., 2018).Another indication for reduced agglomeration has been associated to the lower viscosity of influent that contains a higher faction of colloidal substances.Colloids slow down the reaction kinetics and result in the formation of fewer agglomerated crystals (Capdevielle et al., 2016).

Biological properties
Struvite contamination with Enterococcaceae, E. coli, Coliform bacteria and F-specific RNA phages is low compared to the regulatory requirements and the influent they are produced from.Specially, for phages it has been described that they are mainly contained in the water phase and that adsorption to struvite granules does not occur (Decrey et al., 2011).It has been suggested that adsorption cannot occur because struvite, in a struvite saturated solution, is charged positive and, as phages are also positively charged, this is leading to electrostatic repulsion (Decrey et al., 2011).SSRC as a proxy indicator for Ascris eggs and Clostridium perfringens are of concern as their number exceeds the legal limits for these pathogens.Furthermore, the elevated SSRC level may indicated a low Cryptospridium oocyst removal efficacy.However, as SSRC are only a proxy indicator, reliable statements about removal and viability of these pathogens cannot be drawn from this study.In addition to pathogen removal in the struvite production process, it was observed that storage is a good strategy to reduce pathogen contamination.A similar observation was made by Decrey et al. (2011).These authors detected a 0.07 log reduction/d at 20 °C/93% relative humidity for phages and observed a comparatively low reduction of activity of 0.003-0.01log after three days at storage at 20 °C for Ascaris eggs.
In the present research, a higher reduction of 0.08 log of the proxy indicator SSRC in three days can be calculated (assuming a linear reduction over a nine months period).Since struvite recovered from digestate (type I) contained more pathogens compared to struvite from the dewatering liquor of digestate (type II), this type of struvite should be monitored more closely.The reasons for the elevated Enterococcaceae count in struvite originating from potato processing effluent should be explored.For SSRC, determination of C. perfringens should be performed to shed light on the actual risk, as this was not feasible in this research.After hand-picking macroscopic impurities and struvite washing, a significantly lower number of pathogens was observed by Huygens et al. (2017).The positive effect of hand-picking impurities on pathogen contamination was also confirmed in this study.Finally, heating struvite to temperatures just above 35 °C or a temperature at which it is expected that crystalline structure of struvite is not yet changing, showed to be an effective way to reduce the numbers of viable pathogens including Ascaris eggs and phages in struvite (Decrey et al., 2011).Therefore, a gentle drying process may further reduce pathogen contamination.

P availability and growth performance
The results showed that despite differences in the mineralogical composition of struvite, there is no significant difference between samples in terms of plant yields.Similarly, for all struvite samples a continuous P availability over the sample period was detected, indicating the often described 'slow release' character of struvite.In literature, a more heterogeneous release of different struvite types has been described.Rech et al. (2019) tested three different struvite samples and observed a P availability of 234-261 mg P/L on day one, for a laboratory sample originating from chicken manure (this sample contained more K than in any of the samples observed and contained only 7.6% P).This sample showed a P availability pattern similar to that of SSP and Triple Super Phosphate (TSP), with a high availability at the start and a continued decline.For the other two samples, Rech et al. (2019) found P availabilities comparable to the present study: mean 4.75 ± 0.68 mg P/L; this study 4.8 ± 0.5 mg P/L.Moreover, Rech et al. (2019) found that TSP and the conventional struvite obtained the same P availability after 17 days.
In the present study, this point was reached at day 7 when compared with SSP.The earlier occurrence of the point is likely a result of the more than two times higher P content of TSP as compared to SSP (i.e.6.9% (P) in the present study and 19.35-21% (P) TSP in the study of Rech et al. (2019).Literature indicates that these slow release characteristics can reduce P loss and the associated environmental impact (Everaert et al., 2018;Hertzberger et al., 2020).

Physical properties and fertilizer processing
The physical properties of struvite are of relevance to enable its agronomic use.Struvite granules must be of a shape and size which enable its utilization in modern application equipment.Small granules 1-2.5 mm are used for applications in rows, macro granules 2-3.5 mm are used for spreading and smaller granules may be used for specialized applications such as growth media (Grunert et al., 2019).Furthermore, there is also a market demand for component fertilizers that are containing N-P-K in suitable ratios.If struvite is to be used in N-P-K fertilizers, the P contained in struvite must either be incorporated into a complex granule or blended with N and K granules.Spiller et al. (2019) suggest type II struvite is most likely to be blended as it has reasonably large granules with a relatively homogenous distribution in the range between 1 and 2 mm.However, the authors also point out that the granules are still too small to blended as this is usually practiced at granules sizes between 2 and 3 mm.Whereas, type I and type III struvite is most likely to be reprocessed or used for specialized application (i.e.growing media) due to their irregular shape or small grain size.The results of this study support these conclusions but suggest that individual processes and sites provide larger and high-quality granules.This is demonstrated by the sample Anphos 5 (type III), with the majority of granules exceeding 3.5 mm.Therefore, a case by case assessment is required to determine potential struvite utilization routes.

Conclusion
• Up to 1250 ton of the 1350 ton struvite produced annually in the EU can be considered suitable for direct use as fertilizer or as secondary raw material for fertilizer production.Established processes in the EU produce struvite with a high quality and little variability in terms of chemical parameters.Only three struvite samples failed legal limits: one for a P content of minimal 7% DW (Anphos 3) and three (Phospaq, Anphos 3, Anphos 4) for an organic carbon content exceeding 3% of DW. • Current replacement of fertilizer P imports to the EU by struvite account for 0.5%.The theoretical maximum replacement is estimated to be 13% of fertilizer P imports.• The research demonstrated a rather uniform P availability of between 3.5 and 6.5 mg P/ kg soil per day over the experimental period.• It is suggested that struvite from different origins may be utilized in different application routes.Type II and III may be suitable for blending into N-P-K fertilizers due to their spherical shape, while type I struvite is more suited to be utilized for other applications (e.g. in growing media).• Analysis of the biological contamination of struvite indicates that struvite may exceed certain legal limits (e.g.Ascaris eggs, Clostridium perfringens based on SSRC as proxy), but also, that storage or removal of impurities can reduce contamination.However, at this point no conclusions about biological contamination and mitigation can be made since a more detailed analysis of biological parameters is necessary.

Fig. 1 .
Fig. 1.Three categories in which struvite recovery processes in Europe can be classified.Depending on the type of water/slurry and location of the struvite recovery process in a wastewater treatment plant (WWTP) I: on digestate of sewage waste activated sludge (WAS); II: on the dewatering liquor of sewage WAS digestate; III): after anaerobic digestion of agro-industrial wastewater.(UASB: upflow anaerobic sludge blanket).

⁎
Wet struvite samples (slurry) were obtained, and were dried before analysis at 40 °C.⁎⁎ 2 samples acquired at different time points.

Fig. 2 .
Fig.2.Phosphorus (P), Nitrogen (N) and Magnesium (Mg) content of the analyzed struvite samples as mass % DW, recovered from the digestate (type I), from dewatering liquor of digestate (type II) and from UASB effluent (type III).The theoretical values (12.6 mass % P; 5.7 mass % N; 9.9 mass % Mg) are indicated by the lines (P: dotted red line; N: full blue line; Mg: striped grey line; * samples below the legal limit of 7 mass % for P contents within a precipitated phosphate salt).
not regulated for in the ferƟlizer regulaƟon (mg/kg)

Fig. 4 .
Fig. 4. Mass balance based on elemental and mineralogical data for the analyzed struvite samples.Total metal fractionation is presented in Table3.XRD data are presented in Supplementary Material Fig.S4, raw data presented in SM Table4.

Fig. 5 .
Fig. 5. Mass fraction distribution per particle size for all analyzed struvite samples.Crosses indicate the mean particle size (mm).Numbers indicate the reactor type as follows: 1 Airlift reactor, 2 Continues stirred tank reactor, 3 Fluidized bed reactor, 4 Tank aerated, 5 Tank mixed.Histograms, descriptive statistics and pictures per sample are presented in the SM Section 6, raw data shown in SM Table4.

Table 2
Estimated amount of struvite produced in 2020 in the European Union (including plants with anticipated commissioning in 2020; *excluding sample that did not meet legal requirements).

Table 4
Pathogens and fecal indicator organisms in fresh and stored struvite samples and in struvite from digestate (type I), dewatering liquor of digestate (type II) and UASB effluent of potato wastewater (type III).E. coli: Escherichia coli; SSRC: spores of sulphite reducing clostridia.N.A.:Not applicable; cfu: colony-forming unit; pfu: plaque-forming unit.
⁎ Units are expressed per gram product (fresh weight without drying) for all samples.M. Muys, R. Phukan, G. Brader et al.Science of the Total Environment 756 (2021) 143726 3.3.Struvite P solubilization and maize plant biomass