Optimizing and acidogenesis in order to dissolve and recover phosphorus in organic effluents upstream from production Simon Piveteau

To cite this version:

Simon Piveteau. Optimizing hydrolysis and acidogenesis in order to dissolve and recover phosphorus in organic effluents upstream from methane production. Organic chemistry. Université Rennes1, 2017. English. ￿NNT : 2017REN1S123￿. ￿tel-01800084￿

HAL Id: tel-01800084 https://tel.archives-ouvertes.fr/tel-01800084 Submitted on 25 May 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. ANNÉE 2017

THÈSE / UNIVERSITÉ DE RENNES 1 sous le sceau de l’Université Bretagne Loire

pour le grade de DOCTEUR DE L’UNIVERSITÉ DE RENNES 1 Mention : Chimie Ecole doctorale EGAAL

présentée par Simon Piveteau Préparée à l’unité de recherche OPAALE Optimisation des Procédés en Agriculture, agroALimentaire et Environnement IRSTEA

Thèse soutenue à Rennes Optimiser l’hydrolyse le 19 Décembre 2017 et l’acidogénèse pour devant le jury composé de : Mathieu SPERANDIO dissoudre et recycler Professeur, INSA LISBP/ rapporteur Xiomar GOMEZ Professeur associée, Université de Leon / le phosphore des rapporteur Eric TRABLY effluents organiques Ingénieur de Recherche, INRA-LBE / examinateur Julien BROCHIER en amont des unités Directeur, K-révert / invité Patrick DABERT de méthanisation Directeur de Recherche, IRSTEA / directeur de thèse Marie-Line DAUMER Ingénieur de Recherche, IRSTEA / co-directrice de thèse

I. ACKNOWLEDGEMENT

Je souhaite tout d’abord remercier José Martinez et Eric Dufour pour leur accueil au sein de l’UR optimisation des procédés en agriculture, agroalimentaire et environnement.

Je remercie également le projet VALODIM, BPIFRANCE et la Région Bretagne qui ont participé au financement de cette thèse.

Un grand merci à Patrick Dabert, mon directeur de thèse, et Marie-Line Daumer, ma co-directrice de thèse, qui m’ont choisi pour travailler sur ce sujet malgré mon parcours atypique. En espérant avoir été à la hauteur de leur confiance.

Je tiens également à remercier les membres du jury : Mathieu Sperandio et Xiomar Gomez en tant que rapporteurs et Eric Trably et et Julien Brochier en tant qu’examinateurs d’avoir acceptés d’évaluer ce travail et de faire partie de mon jury de thèse.

Merci également à Olivier Chapleur, Andrea Schievano, Claire Dumas, Michel Gautier et Dominique Wolbert pour leurs conseils lors des comités de suivi de thèse et la soutenance à mi-parcours.

Merci à tous les membres du projet Valodim avec qui j’ai pu échanger et progresser dans mes recherches.

Merci à Etienne Braak qui m’a mis le pied à l’étrier lorsque je suis arrivé à l’IRSTEA.

Merci à Patricia Mylène et Véronique pour leur bienveillance et leur patience.

Merci à Sylvie pour ses vastes compétences, ses conseils répétés inlassablement et sa disponibilité.

Merci à Julie, Guillaume et Jean-Claude, mais surtout Julie, sans qui aucun pilote n’aurait vu le jour.

Merci à Sophie qui a su s’y retrouver parmi mes échantillons et en extraire la substantifique moelle.

Merci au service d’appui et tout particulièrement Isabelle, Asma, Alexandra et Fabrice Egido car les RH, l’administratif et l’informatique ne sont pas mes points forts, surtout quand ils s’entremêlent.

Un grand merci à Younes, Romain C., Romain G. et Clément L. pour les discussions scientifiques particulièrement enrichissantes qu’on a pu avoir.

Un autre grand merci à Carlos et Henry parce que côtoyer des colombiens est surement ce qu’il peut nous arriver de mieux.

Merci à Axelle et Samuel pour les bons moments passés ensemble, aussi rares qu’ils aient pu être.

Merci à Stéphanie, Amandine, Benjamin, Simon et Lydie pour ces samedis mémorables même si parfois oubliés malgré moi. Mention spéciale à Amandine pour les duos chantés à tue-tête.

Merci à Emilien, Pascal et Clément S. pour les parties de Risk Napoléon, Game of Thrones, Seigneur des anneaux, futuriste, Godstorm et j’en passe. Re-merci à Clément S. pour le temps passé ensemble et tout ce qu’on a pu se raconter.

Merci à Lysiane que je retrouvais à Rennes après de trop longues années sans contacts. Merci à Romain de m’avoir fait découvrir le football gaélique et merci à tous les deux pour le petit Albert qui me manque déjà.

Un grand merci à Lucas, Loïc, Perrine, Yann, Coline, Kévin, Mathilde, Garance et tous les membres du Rouroux Club avec qui les moments partagés pendant ces trois ans furent aussi brefs que délicieux.

Merci du fond du cœur à mes petites sœurs, Marine et Elodie sur qui je pourrai toujours compter.

Un immense merci à mes parents, mamie Paulette, mamie Yvonne et papy Pierre, que je n’aurais pas vu beaucoup mais dont la confiance m’a toujours aidé à garder le cap.

Et finalement merci à Anne Laure, que le destin aura fait arriver le même jour que moi à l’IRSTEA. Son dévouement et sa patience n’ont surement aucun égal chez qui que ce soit, et il me faudra plus qu’une vie pour lui rendre la pareille.

PUBLICATIONS AND SCIENTIFIC COMMUNICATIONS

Article de revue scientifique à comité de lecture PIVETEAU, S., PICARD, S., DABERT, P., DAUMER, M.L. - 2017. Dissolution of particulate phosphorus in pig slurry through biological acidification: A critical step for maximum phosphorus recovery as struvite. Water Research, vol. 124, p. 693-701 BRAAK, E., AUBY, S., PIVETEAU, S., GUILAYN, F., DAUMER, M.L. - 2016. Phosphorus recycling potential assessment by a biological test applied to wastewater sludge. Environmental

Technology, vol. 37, n° 11, p. 1398-1407 GUILAYN,F., BRAAK, E., PIVETEAU, S., DAUMER, M.L. - 2017. Sequencing biological acidification of waste activated sludge aiming to optimize phosphorus dissolution and recovery.

Environmental Technology, vol. 38, n° 11, p. 1399-1407

Communication scientifique avec actes PIVETEAU, S., DAUMER, M.L., DABERT, P. - 2016. Phosphorus recovery from pig slurry through biological acidification and re-crystallization as struvite. Orbit 2016, Organic Resources and Biological Treatment, 10 th International Conference on Circular Economy and Organic Waste

25/05/2016-28/05/2016, Heraklion, GRC. 11 p.

Communication scientifique sans actes PIVETEAU, S. 2015. Biological acidification of pig slurry for phosphorus recovery as struvite, ManuREsource 2015. International conference on manure management and valorization, December 2- 3-4 2015, Ghent, Belgium

RÉSUMÉ

Le phosphore est un élément clé des mesures prises en faveur de l’économie circulaire en Europe car c’est un élément irremplaçable, indispensable à tous les organismes vivants et déterminant pour l’autonomie alimentaire du continent. Or, il est issu de ressources minières limitées fortement concentrées dans quelques pays et dont l’Europe est presque totalement dépourvue. L’approvisionnement et son coût sont dépendants du contexte géo-politique, soumis aux fluctuations de l’offre et de la demande mondiale. Paradoxalement, dans les régions d’élevage intensif, le phosphore importé massivement avec les aliments du bétail mais mal retenu par les animaux se retrouve épandu en excès avec les effluents d’élevage conduisant à l’eutrophisation des eaux de surface avec des impacts économiques importants sur la potabilisation de l’eau ou son utilisation pour la pisciculture ou les loisirs. C’est également le cas autour de certaines zones urbaines où le phosphore immobilisé dans les boues de station d’épuration est épandu plus que valorisé sur les terres agricoles. Au contraire, l’approvisionnement des régions de grandes cultures repose presque exclusivement sur les importations d’engrais chimiques. L’enjeu est donc de pouvoir recycler le phosphore des effluents humains et animaux, en excès par rapport au besoin des cultures dans certaines régions, sous une forme utilisable dans les zones de grandes cultures. Dans ces travaux, nous nous sommes intéressés spécifiquement au P des lisiers de porcs qui représente une source importante facilement mobilisable du fait de la concentration des élevages et de l’existence de procédés de traitement de l’azote et/ou de valorisation énergétique.

Des procédés de recyclage du P ont été développés produisant majoritairement de la struvite qui est un engrais à diffusion lente dont les performances agronomiques sont au moins équivalentes à celles des engrais minéraux fabriqués à partir de ressources minières. La plupart de ces procédés ont été développés pour recycler le P issu des boues de stations d’épuration urbaines de taille importante. Le plus souvent, ils ne concernent que le P dissous, ce qui limite fortement le taux de recyclage. Le prétraitement des effluents pour augmenter la fraction de P dissous est souvent coûteux particulièrement dans le cas des effluents d’élevage qui sont des milieux fortement tamponnés dans lesquels la dissolution du P majoritairement sous forme minérale particulaire exige d’importantes quantités d’acide. Cette étape de dissolution est le verrou au développement de la plupart des procédés.

Certains procédés biologiques comme l’ensilage ou la production d’hydrogène reposent sur le potentiel acidifiant de micro-organismes qui peuvent se développer dans certaines conditions dans les milieux fortement chargés en matière organique. La baisse du pH significative engendrée par ces métabolismes pourrait être substituée au prétraitement par acidification chimique pour dissoudre le P des lisiers.

Cette thèse accompagne donc le développement et l’optimisation d’un procédé innovant basé sur la l’acidogénèse, production d’acide in situ par les bactéries endogènes, permettant la dissolution d’une

large fraction du P total initial qui constitue la première étape du recyclage du P des lisiers de porcs. Après séparation, le liquide enrichi en P est ensuite soumis à l’étape de cristallisation pour produire de la struvite tandis que les boues appauvries en azote et en phosphore peuvent être utilisées comme apport de matière organique pour les sols déjà très fortement pourvus en P des régions d’élevage intensif. Du fait de l’hydrolyse acide de déchets organiques générée par ce procédé, il peut s’inscrire en amont ou en parallèle des procédés de valorisation énergétique par production de méthane dont le développement est fortement encouragé.

Objectifs

Les objectifs de cette thèse peuvent se résumer ainsi :

• Déterminer à l’aide d’un co-substrat modèle (saccharose) l’effet de l’acidogénèse sur le pH, la dissolution du P mais aussi des autres ions impliqués ou interférant dans la cristallisation de la struvite • Etudier l’effet de l’utilisation de divers déchets organiques comme co-substrats sur le développement de la flore acidogène et leurs métabolismes • Evaluer la possibilité de disposer d’un outil simple de prédiction de la baisse du pH en fonction du type de co-substrat et des caractéristiques du lisier. • Etudier la possibilité de développer une flore acidogène adaptée en travaillant sur un réacteur semi-continu. • Vérifier la faisabilité de cristalliser de la struvite à partir du liquide enrichi en P et la qualité du produit obtenu. • Etablir un bilan chiffré du recyclage et proposer des pistes d’améliorations.

Résultats

La présentation des résultats est organisée autour de trois publications dont l’une est publiée, l’une est soumise et l’autre prête à soumettre.

Les objectifs des travaux décrits dans le premier article sont de tester la faisabilité de l’acidification biologique et d’évaluer la quantité de sucre nécessaire en fonction des caractéristiques des lisiers (pH initial, pouvoir tampon …). L’effet de l’acidification biologique sur la dissolution des autres composés de la struvite (ammonium et magnésium) et du calcium qui peut interférer avec la cristallisation de la struvite a aussi été étudié. Pour cette première approche du saccharose (sucre blanc du commerce) a été utilisé. Les résultats obtenus avec le sucre permettront d’évaluer le potentiel de dissolution du P par voie biologique sur différents lisiers.

Quatre lisiers différents ont été utilisés. Les essais réalisés avec sept concentrations de sucre entre 0 et 60 g.L-1montrent que la baisse de pH obtenue est directement liée à la concentration en sucre ajoutée. Un pH de 3.99 a été atteint avec la concentration la plus forte. Le dosage des acides organiques a

montré que la production d’acide lactique est le principal responsable de la baisse du pH. Le faible pH atteint avec les concentrations en sucre les plus fortes permettrait d’inhiber les métabolismes de dégradation de l’acide lactique et de maintenir le pH stable. Les concentrations en phosphore, calcium et magnésium dissous sont liées à l’évolution du pH. Jusqu’à 95% du P total initial du lisier ont été dissous avec les trois concentrations en sucre les plus élevées.

La dissolution du P et du magnésium en fonction du pH n’est pas linéaire. Elle intervient principalement pour des pH compris entre 5.5 et 6.5. En revanche la dissolution du calcium en fonction du pH est quasi-linéaire entre pH 7 et pH 4.5. Cette évolution différente permet de définir une zone de pH comprise entre 5.5 et 6 pour laquelle la dissolution du Ca comparée au P et au Mg est moindre ce qui favoriserait la cristallisation du P sous forme de struvite lors des étapes ultérieures. Le fait d’opérer la séparation avant cristallisation à ce pH réduirait également la quantité de réactif nécessaire à la remontée du pH lors de la cristallisation. En revanche la faible stabilité du pH dans cette zone demande une gestion précise du procédé d’acidification.

Finalement la régression linéaire mise en œuvre sur les données obtenues avec les quatre lisiers a permis d’établir une relation bien ajustée entre le pH minimal, la quantité de sucre, le pH initial du lisier et son pouvoir tampon pour une zone de pH entre 4 et 8. Cette relation pourra être utilisée pour calculer la quantité de co-substrat à ajouter. Ces travaux ont montré le lien étroit entre la production d’acide lactique et le pH. Or l’acide lactique est le produit majoritaire de dégradation des sucres simples très facilement dégradables.

Les travaux décrits dans le second article ont pour objectif de s’assurer que l’acidification peut être obtenue avec les co-substrats riches en glucides complexes disponibles sur les unités de méthanisation et de tester la faisabilité d’un recyclage en parallèle de la méthanisation pour exploiter la présence d’une flore déjà accoutumée à la dégradation anaérobie de co-substrats complexes.

Un lisier brut et deux digestats issus du même digesteur mais prélevés à deux périodes différentes ont été utilisés. Avec le digestat, le pH décroît presque linéairement jusqu’à des concentrations de sucre de 30 g.L-1. La baisse est moins importante à quantité de sucre équivalente avec le lisier brut et 50 g.L-1 sont nécessaires pour obtenir la valeur minimale et ce malgré un pH initial plus faible. La présence d’AGV dans le lisier avant le test peut expliquer à la fois le pH plus faible et le pouvoir tampon plus élevé. Le lactate est l’acide organique majoritaire pour les fortes concentrations en sucre dans les deux matrices utilisées. Pour des concentrations inférieures, le lactate a été produit pendant les premières 48 heures mais consommé ensuite lors de secondaires qui ont pu avoir lieu grâce à un pH insuffisamment acide pour les inhiber. Un pH inférieur à 5 est obtenu avec les pommes, les carottes à 50g.L-1, les biscuits et la farine de blé noir stockés. Cette baisse de pH peut être attribuée à la production d’acide lactique. Aucune baisse de pH ni production d’AGV n’est observée pour la paille. Pour les autres co-substrats, le pH atteint des valeurs comprises entre 5 et 5.7. Les valeurs atteintes

sont corrélées à la teneur en glucides. La conversion de la matière organique en acides organiques est comprise entre 25 et 56% inférieures aux valeurs obtenues avec le sucre (44-78%).

Les communautés microbiennes impliquées dans l’acidification ont été caractérisées par séquençage haut débit des gènes des ARNr 16S. L’analyse des indices de diversité suggère une large diversité initiale avec quelques espèces dominantes. La diversité se réduit pour les essais pour lesquels les pH les plus bas sont atteints. Le faible pH et la concentration en lactate exerceraient une pression de sélection ne permettant de conserver que les micro-organismes tolérants aux conditions acides que sont les Lactobacillus. Ces essais montrent que la lactique avec des déchets organiques est possible et que la production d’acide lactique dépendra de la disponibilité de sucres facilement dégradables, qui favorisent le développement rapide des Lactobacillus, avec un effet inhibiteur empêchant les fermentations secondaires.

L’objectif des travaux décrits dans le troisième article est de définir un « équivalent sucre » pour quelques déchets issus de la production agricole régionale puis de vérifier la relation établie précédemment pour calculer la quantité de co-substrats à ajouter pour atteindre le pH désiré en fonction des caractéristiques du lisier. Enfin l’influence de l’acidification biologique sur la cristallisation du P sous forme de struvite et la qualité du solide final a été testée.

Quatre lisiers issus d’élevages industriels ont été utilisés et 5 co-substrats, dont le sucre, y ont été ajoutés à la concentration de 50 g de matière organique par litre pour déterminer l’équivalent sucre. Les 4 substrats sont des légumes ou des fruits habituellement produits en Bretagne et dont les invendus où les déchets de transformation alimentent les méthaniseurs. Ce sont des co-substrats riches en sucre mais avec des teneurs ou une accessibilité différente. Il s’agit des carottes, pommes petits pois et haricots verts. L’équivalent sucre est le rapport entre la quantité de sucre et la quantité de co-substrat permettant d’obtenir un pH équivalent. Des valeurs respectives de 0,74, 0,69, 0,42 et 0,42 ont été calculées pour les pommes, les carottes, les haricots et les petits pois. Les quantités pour atteindre des valeurs de pH de 4, 5 et 6 ont été calculées d’après l’équation citée plus haut et ont été ajoutées aux 4 lisiers. L’adéquation entre le pH obtenu et le pH calculé a été estimée par la moyenne des valeurs absolues des différences ou par la moyenne des différences positives ou négatives. L’adéquation est satisfaisante pour le sucre comme pour la plupart des co-substrats sauf les petits pois. Dans ce dernier cas, les acides majoritairement produits ne sont pas le lactate mais l’acétate, le propionate, le butyrate et le valérate. Ces acides ayant un pKa supérieur à 4.5, il n’est pas possible d’atteindre un pH de 4. La suppression des valeurs obtenues pour les essais dont l’objectif était d’atteindre ce pH permet d’améliorer l’adéquation entre les valeurs de pH attendue et mesurée. La différence est alors au maximum de 0,45 point de pH. Pour des valeurs de pH proches de la valeur optimale de 5.5-6 une fermentation secondaire intervient systématiquement ce qui confirme l’hypothèse émise précédemment selon laquelle le procédé devra être piloté avec un suivi rapproché du pH pour réaliser la séparation lorsque le ratio Ca/P est optimal. Comme lors des essais précédents, l’évolution de la

concentration en P dissous lors de l’acidification avec les co-substrats est étroitement liée à l’évolution du pH pour 3 des quatre lisiers. Un comportement différent a cependant été observé pour un des lisiers qui avait une concentration initiale en P total très faible. Dans ce cas et surtout lorsque les quantités de co-substrat ajoutées sont importantes l’assimilation du P pour la croissance de la biomasse n’est plus négligeable et réduit la concentration en P dissous observée. Si le co-substrat est riche en P comme c’est le cas des petits pois, la concentration en P dissous sera liée à l’hydrolyse progressive des formes organiques contenant le P sous l’effet de l’acidification.

Afin d’évaluer l’intérêt de conserver un inoculum acidogène adapté au co-substrat des essais en semi- continu ont également été réalisés. Le pH est descendu rapidement à 5.7 pendant les 4 premiers jours puis le pH minimum a baissé progressivement jusqu’à 5.4. La concentration en P dissous est passée de 150 à 680 mg.L-1 en quatre jours puis a augmenté progressivement pour atteindre 747 mg.L-1 soit 57% du P total du mélange au bout des 16 jours. L’effluent recueilli pendant la durée du test a été centrifugé puis analysé avant d’être utilisé pour des essais de cristallisation. La centrifugation a permis de récupérer 83% du volume sous forme liquide soit 47% du P total initial. La cristallisation a été réalisée en bécher agité. Environ 3.5g d’hydroxyde de magnésium ont été nécessaire pour atteindre pH 8. Le mélange a été laissé sous agitation pendant 2 heures. Plus de 99% du P dissous a été précipité. Le solide produit est composé à 69% de struvite, le reste étant probablement composé d’autres composés minéraux phosphatés, de l’hydroxyde de magnésium non dissous et de la matière organique.

Le faible temps de contact et le mode d’agitation ont conduit à l’obtention de cristaux de petite taille et hétérogènes. Aussi, malgré une précipitation presque totale du P dissous, le solide issu de la filtration à 100 µm n’a permis de recueillir que 26% du P total du mélange initial. L’obtention d’un pH légèrement inférieur lors de l’acidification aurait probablement permis d’améliorer le taux de dissolution du P et la réalisation de la cristallisation dans un équipement industriel optimisé d’obtenir des cristaux plus gros. Ces deux étapes sont déterminantes pour améliorer le rendement global du procédé. Si des équipements industriels développés spécifiquement pour la cristallisation de la struvite sont disponibles sur le marché, une bonne connaissance des co-substrats et notamment de leur équivalent sucre est déterminante pour choisir les cocktails qui favoriseront la production d’acide lactique et limiteront les fermentations secondaires. Des études complémentaires sur l’influence des paramètres de l’étape d’acidification biologique (temps de séjour, température …) devraient également permettre d’orienter les métabolismes.

Conclusion et perspectives

L’hypothèse de départ suivant laquelle l’utilisation du métabolisme acidogène de la flore endogène pourrait provoquer in situ la baisse de pH nécessaire à la dissolution du P a été vérifiée. Entre 40 et 90% du P ont été dissous suivant les co-substrats. Pour obtenir ce résultat, des co-substrats riches en sucres facilement dégradables favorisant le développement des bactéries lactiques devront être

privilégiés. La quantité à apporter pourra être calculée en fonction des caractéristiques du lisier avant ou après digestion grâce au modèle simple proposé dans ces travaux.

Les facteurs clés déterminant le rendement du procédé sont la concentration en P dissous au moment de la séparation et la taille des cristaux obtenus tandis que le coût du produit sera plutôt lié à la quantité de réactif nécessaire à l’augmentation du pH lors de la cristallisation.

Des travaux complémentaires sur les paramètres du procédé pour orienter la microflore et le métabolisme vers la production d’acide lactique permettraient de s’assurer d’un pH suffisamment bas et stable pour optimiser la dissolution du P. De plus la production majoritaire d’acide lactique réduirait le pouvoir tampon du lisier lié à la présence d’acides gras non dissociés ( acides organiques dont le pKa est plus faible) et ainsi la quantité de réactif nécessaire à ajuster le pH lors de la cristallisation de la struvite. La taille des cristaux pourra probablement être optimisée en utilisant un réacteur spécifique présentant des conditions hydrauliques favorables à la croissance.

Enfin, l’effet favorable de cette étape sur la production de méthane du fait de l’hydrolyse acide induite devra être vérifié.

TABLE OF CONTENTS

INTRODUCTION ...... 1

CONTEXT ...... 5

I. PHOSPHORUS, A NUTRIENT AT A CROSSROAD...... 5 1. Role of phosphorus in Nature and anthropogenic activities ...... 5 2. Phosphorus reserve, current and forecasted uses ...... 7 3. Mitigating measures envisioned to close the phosphorus cycle...... 10 II. PROCESSES FOR PHOSPHORUS RECOVERY ...... 12 1. Agronomic value of struvite ...... 12 2. Existing processes for phosphorus recovery ...... 14 a. Technologies for phosphorus recovery from municipal / industrial wastewaters ...... 14 Recovery from treated effluents ...... 15 Recovery from the liquid phase of ...... 16 Phosphorus recovery from municipal WWTP digested sludge itself ...... 19 Pre-treatments of municipal sludge ...... 20 Conclusion on phosphorus recovery from WWTP...... 26 b. Phosphorus recovery from Livestock effluents ...... 27 Phosphorus recovery from the liquid phase ...... 27 Pre-treatments prior to phosphorus recovery ...... 32 III. ACIDOGENESIS ...... 38 1. Catabolism of organic matter under anaerobic conditions ...... 41 a. Catabolism of amino acids ...... 41 b. Catabolism of ...... 43 c. Catabolism of ...... 43 2. fermentation ...... 45 a. Lactic acid ...... 45 b. Ensiling process ...... 47 c. production from lactate ...... 48 d. Other organic acids produced from lactate ...... 48 e. Fermentation of swine manure ...... 49 f. Lactic acid fermentation followed by methane production ...... 50 IV. CONCLUSION ...... 51

REFERENCES ...... 52

OBJECTIVES OF THESIS ...... 67

CHAPTER 1 ...... 68

DISSOLUTION OF PARTICULATE PHOSPHORUS IN PIG SLURRY THROUGH BIOLOGICAL ACIDIFICATION ...... 68

I. SUMMARY OF CHAPTER 1 ...... 69

II. ABSTRACT ...... 72 III. INTRODUCTION ...... 73 IV. MATERIALS AND METHODS ...... 75 1. Pig slurry ...... 75 2. Biological acidification of pig slurry in batch tests using sucrose as co-substrate ...... 75 3. Analysis...... 76 4. Statistics ...... 76 V. RESULTS AND DISCUSSION ...... 77 1. Effect of sucrose concentration on pH and fermentation product in slurry 1...... 77 a. pH change ...... 77 b. Evolution of lactic acid, VFAs and sucrose across time ...... 77 2. Phosphorus, magnesium, calcium and dissolution processes in slurry 1 ...... 80 3. Dissolution of P, Mg, Ca, N in the other slurries ...... 83 4. Potential for struvite crystallization ...... 84 5. Correlating lowest pH with initial sucrose concentration and buffer capacity ...... 85 VI. CONCLUSIONS ...... 87 VII. ACKNOWLEDGEMENT ...... 87 VIII. FUNDING SOURCE ...... 87 IX. REFERENCES ...... 88 X. APPENDIX ...... 92 1. Appendix A ...... 92 2. Appendix B ...... 92 3. Appendix C ...... 93 4. Appendix D ...... 93

CHAPTER 2 ...... 94

BIOLOGICAL ACIDIFICATION OF SWINE SLURRY: EFFECT OF VARIOUS ORGANIC CO- SUBSTRATES ON PH, ORGANIC ACIDS PRODUCTION AND BACTERIAL POPULATIONS ...... 94

I. SUMMARY OF CHAPTER 2 ...... 95 II. ABSTRACT ...... 98 III. INTRODUCTION ...... 99 IV. MATERIALS AND METHODS ...... 100 1. Characterization of pig slurry and digested pig slurry...... 100 2. Batch experiments...... 100 3. Physic and chemical analyses ...... 101 4. Microbiological analysis ...... 101 5. Statistical analysis ...... 102 V. RESULTS AND DISCUSSION ...... 102 1. Evolution of pH and organic acids concentration in batch tests with sucrose ...... 102 2. Evolution of pH and organic acid production in batch tests with organic waste as co-substrates ... 104

3. Microbial diversity detection using 16S rDNA high throughput sequencing ...... 106 4. Identification of the dominant species ...... 107 VI. CONCLUSION ...... 110 VII. REFERENCES ...... 112

CHAPTER 3 ...... 116

BIOLOGICAL ACIDIFICATION OF PIG SLURRY USING ORGANIC CO-SUBSTRATES: AN EFFICIENT PROCESS FOR PHOSPHORUS DISSOLUTION PRIOR TO STRUVITE CRYSTALLIZATION ...... 116

I. SUMMARY OF CHAPTER 3 ...... 117 II. ABSTRACT ...... 121 III. INTRODUCTION ...... 122 IV. MATERIALS AND METHODS ...... 124 1. Pig slurries ...... 124 2. Organic co-substrates ...... 124 3. Determination of the acidifying power for each co-substrate compared to sucrose ...... 125 4. Biological acidification of pig slurry in batch tests using complex organic co-substrates...... 126 5. Semi continuous reactor ...... 127 6. Crystallization process ...... 127 7. Analysis...... 128 V. RESULTS AND DISCUSSION ...... 128 1. Model fitness ...... 128 2. Dissolution of P, Mg, Ca and N ...... 131 a. Phosphorus ...... 131 b. Magnesium calcium and nitrogen dissolution ...... 134 3. Molar ratios of interest for struvite precipitation ...... 135 4. Semi-continuous reactor operation and phosphorus recovery ...... 135 VI. CONCLUSIONS ...... 137 VII. FUNDING SOURCE ...... 138 VIII. REFERENCES ...... 139

DISCUSSION AND PERSPECTIVES ...... 141

LIST OF TABLES

Table 1. rock reserves estimations ...... 7 Table 2. Amount of acid needed to dissolve P under various forms ...... 37 Table 3: pKa of organic acids produced during acidogenesis ...... 40 Table 4. Composition of the pig slurries (concentrations expressed per liter of raw manure) ...... 75 Table 5. recovery in R10 to R60 at each sampling time. Mean of three incubations and standard deviation are displayed. Carbon recovery at maximum lactic concentration shown in bold .. 78 Table 6. Characteristics of pig slurry and digested slurries (standard deviation displayed between brackets) ...... 100 Table 7. Composition of the organic waste used during biological acidification of pig slurry. Standard deviation of triplicate measures for TS and VS displayed between brackets...... 101 Table 8. Statistical results for nMDS analysis ...... 102 Table 9. Carbon recovery: C-mol organic acids produced / C-mol sucrose consumed (%)...... 104 Table 10. Conversion ratio of organic matter to organic acids ...... 105 Table 11. Composition of the pig slurries ...... 124 Table 12. Composition of the organic co-substrates ...... 125 Table 13. Sucrose equivalents of the co-substrates ...... 126 Table 14. Co substrates and sucrose concentrations in each batch test (g/L) ...... 127 Table 15. Average relative and absolute error of the model for sucrose and each co-substrate ...... 130 Table 16. Proportion of total phosphorus in each batch test originating from the co-substrate expressed as percentage of TP ...... 132 Table 17. Composition of acidified slurry and filtrate after crystallization, including mean value and standard deviation...... 136

LIST OF FIGURES

Figure 1: Production of phosphate rock in 1997 and 2016...... 8 Figure 2: Phosphate rock reserves in the world in 1999 and 2017 (USGS data) ...... 8 Figure 3: World phosphate rock production from 1960 to 2017 (USGS data) ...... 9 Figure 4: P fertilizer demand and expected growth until 2020 ...... 9 Figure 5. Access points for P recovery in municipal WWTP (adapted from Egle et al., (2015)) ...... 15 Figure 6. Aiprex process for struvite recovery (adapted from P-rex)...... 18 Figure 7. Combination process of Pearl and Wasstrip ...... 19 Figure 8. Seaborne process (adapted from Müller et al. (2007)) ...... 21 Figure 9. Stuttgart process ...... 22 Figure 10. Budenheim process ...... 23 Figure 11. Ash Dec process ...... 25 Figure 12. P-recovery process for pig manure. Adapted from Vanotti et al. (2007) ...... 29 Figure 13. metabolic pathways for the degradation of organic matter under anaerobic conditions. Adapted from Moletta (1993) ...... 39 Figure 14: proportion of protonated acid as a function of pH ...... 40 Figure 15: metabolic pathways during acidogenesis. Dotted lines correspond to secondary fermentation of primary ...... 41 Figure 16: Stickland reaction between alanine and glycine ...... 42 Figure 17: catabolic pathways in LAB for (adapted from Wang, 2015)...... 46 Figure 18: Shifts in bacterial population during the different steps of silage processing (Dunière et al., 2013) ...... 47

Figure 19: Conversion of lactate and to butyrate and CO2 (Duncan, 2004) ...... 48 Figure 20: Acrylate pathway for lactate fermentation to propionate by propionicum (http://microbiochem.weebly.com/propionate.html) ...... 49 Figure 21. Envisioned implementation of struvite recovery within an anaerobic digestion process: parallel biological acidification. Implementation of biological acidification prior to digestion is also possible ...... 51 Figure 22. Evolution of pH across time for each initial sucrose concentration (standard deviation displayed) ...... 77 Figure 23. Net production of VFAs and lactate, sucrose removed and pH across time. Standard deviation shown for the sucrose removed and the sum of VFAs and lactate. A: Low initial sucrose concentration (30 g/L). B: High initial sucrose concentration (50 g/L) ...... 80 Figure 24. A: Dissolved phosphorus across time for each sucrose concentration. B: Dissolved magnesium. C: Dissolved calcium. D: Ammonium ...... 82

Figure 25. A: Dissolved phosphorus expressed as percentage of the total amount, plotted against pH. B: Dissolved magnesium. C: Dissolved calcium. D: Ammonium ...... 83 Figure 26. A: Dissolved P expressed as a percentage of TP slurry 1, 2, 3 and 4. B: Dissolved Mg C: Dissolved Ca. D: Ammonium ...... 84 Figure 27. Measured pH plotted against model-predicted pH ...... 86 Figure 28. Organic acids produced and pH after 4 days. A: batch tests with 1. B: batch tests with raw slurry. *:not applicable ...... 103 Figure 29. Final pH and organic acids produced after 4 days ...... 105 Figure 30. Correlation between content and final pH ...... 105 Figure 31. Comparison of the microbial community diversity indexes (Shannon, Simpson and equitability) and pH of the different samples ...... 106 Figure 32. nMDS statistical analysis of samples microbial communities ...... 107 Figure 33. A: community structures and corresponding value of pH the swine slurry and digestate samples at the phylum taxonomic rank. B: inclusion of dominant Orders (Porphyromonadaceae), Families (Clostridiales, Lactobacilliales) and Genera (Lactobacillus, Streptococcus) ...... 110 Figure 34. Lowest pH reached at different initial sucrose concentrations. Example of sucrose- equivalent determination for apple ...... 126 Figure 35. Measured minimum pH plotted against model-predicted pH for sucrose, apple, carrot, pea and green bean batch tests ...... 130 Figure 36. Organic acids produced during biological acidification of slurry 2 with pea (A) and sucrose (B) as co-substrates ...... 131 Figure 37. Minimum pH reached during biological acidification using sucrose (A) and peas (B) as organic co-substrate ...... 131 Figure 38. Maximum proportion of co-substrate converted into organic acids during batch tests with high initial concentration of co-substrate (A). Mean proportion and standard deviation of initial VS converted into organic acids for each co-substrate and initial concentration (B) (slurry 4 excluded) 131 Figure 39. Phosphate concentration plotted against pH in slurry 1 (A), slurry 2 (B), slurry 3 (C) and slurry 4 (D) ...... 133 Figure 40. Percentage of dissolved P and pH plotted against time in slurry 1 at medium initial concentration ...... 133 3- 2+ 2+ + Figure 41. Percentage of PO4 , Mg , Ca and NH4 plotted against pH ...... 134 Figure 42. Phosphate concentration and pH plotted against time in the semi-continuous reactor ...... 137 Figure 43. Recovered crystals from the precipitation step ...... 137

LIST OF EQUATIONS

Equation 1: Formation of struvite...... 14 Equation 2: Formation of HAP and HDP ...... 14 Equation 3. Decarboxylation of amino acid ...... 43 Equation 4. Acid products from sewage sludge fermentation ...... 43 Equation 5. Metabolic pathways for degradation during acidogenesis (Yin, 2016) ...... 43 Equation 6. Pyruvate formation from glucose ...... 43 Equation 7. Acetate formation from glucose ...... 44 Equation 8. Butyrate formation from glucose ...... 44 Equation 9. Propionate formation from glucose ...... 44 Equation 10. Butyrate formation from lactate ...... 48 Equation 11. Butyrate formation from lactate and acetate ...... 48 Equation 12. Consumption of lactate via the acrylate pathway ...... 49

Equation 13. Conversion of lactate into acetate, CO2 and biohydrogen ...... 49 Equation 14. Calculation of initial sucrose concentration necessary to reach a targeted pH ...... 86 Equation 15. Calculation of buffer capacity of the slurry ...... 124 Equation 16. Calculation of lowest pH based on initial pH and sucrose concentration ...... 125 Equation 17. Calculation of initial sucrose concentration necessary to reach a targeted pH ...... 127 Equation 18. Calculation of initial co-substrate concentration based on its sucrose equivalent ...... 127

LIST OF ABBREVIATIONS AND ACRONYMS

AMP, ADP, ATP: Adenosine mono-, di-, triphosate ANOVA: Analysis of variance COD: Chemical oxygen demand CSTR: Completely stirred tank reactor DNA: Deoxyribonucleic acid EBPR: Enhanced biological phosphorus removal EMP: Embden-Meyerhof-Parnas FHL: Formate hydrogen lyase GAP: Glyceraldehyde-3P HAP: Hydroxyapatite HDP: Hydroxydicalciumphoshate HM: Heavy metals HPTH: High-pressure thermal hydrolysis HRT: Hydrolyc retention time LAB: Lactic acid bacteria LCFA: Long chain fatty acids LDH: mEq: milliequivalent MgO: Magnesium oxide MDS: Metric multidimensional scaling MT: Million tonne MW: Microwage N: Nitrogen NAD: Nicotinamide adenine dinucleotide OTU: Operational taxonomic unit P: Phosphorus PAO: Polyphosphate accumulating organisms PFL: Pyruvate formate lyase PFOR: Pyruvate ferredoxin POP : Persistent organic pollutants QIIME: Quantative insights into microbial ecology RNA: Ribonucleic acid SCWO: Super critical water oxidation SSA: Sewage sludge ash T: Total TS: Total solids TSS: Total suspended solids UASB: Upflow anaerobic sludge reactor VS: Volatile solids VSS: Volatile suspended solids VFA: Volatile WAO: Wet air oxidation WAS: Wasted activated sludge

INTRODUCTION

Phosphorus (P) is a crucial element for life. It can be found in molecules as important as Deoxyribonucleic Acid (DNA, Ribonucleic Acid (RNA) and Adenosine Triphosphate (ATP). Phosphorus is also one of the limiting nutrients in agriculture with nitrogen (N) and potassium (K) (Sattari et al., 2014). As a result, those nutrients need to be added to the soil using mineral fertilizers and animal manure (Vitousek et al., 2009). Historically, crop production relied on natural levels of soil-P and addition of locally available organic matter like manure and human excreta (Mårald, 1998). In the twentieth century, increased food demand due to rapid population growth has led to the massive application of phosphate rock based mineral fertilizers to crop soils (Smil, 2000). This “green revolution”, which combines technical and agronomic progress with large application of mineral fertilizer enabled major improvements in crop yields and associated food production but also led to the accumulation of P in soils and a significant amount can be lost due to runoff to surface waters (Carpenter, 2005). Nutrient over enrichment in fresh water bodies can lead to excessive biomass production, a harmful phenomenon known as eutrophication. In this situation, phytoplankton community may shift to toxic algal bloom (Smith and Schindler, 2009). Its decomposition leads to foul odors and oxygen depletion, killing the fish population (Carpenter, 2005). Other problems associated with eutrophication include the presence of toxins (Lawton and Codd, 1991), extirpation of native plants (Gleick, 1998), and loss of biodiversity (Maitland, 1993). In fresh water undisturbed by anthropogenic activity, P is usually the limiting nutrient. As a result, increase in P concentration is the main cause of eutrophication (Foy and Withers, 2002; Schindler et al., 2008). State regulations have been enacted during the last 20 years in the developed countries to limit P concentration in municipal and industrial wastewater effluents in order to fight against eutrophication. Indeed, P concentration is relatively high in municipal wastewaters, coming mostly from human feces and detergents. Municipal and industrial wastewaters are now fully treated under stringent discharge limits, reducing drastically the “point sources” of P in fresh waters. As a result, it is now necessary to redirect the efforts towards the management of “diffuse sources” of P, i.e. agriculture (Dupas et al., 2015; Grizzetti et al., 2012; Némery et al., 2005; Van Drecht et al., 2009). Among agricultural activities, intensive livestock breeding can be a main cause of diffuse P emissions as a result of high application rates of animal waste and the resulting enrichment of soils in phosphorus (Heckrath et al., 1995; Ringeval et al., 2014). The recent focus on modern agriculture’s role in diffuse P emissions and associated eutrophication of freshwater bodies has led to modest but significant improvements with lower P fertilizer application, reduced transport to aquatic ecosystems thanks to retention basins (Bennett et al., 2001) and individual or collective treatment systems. These efforts have recently been heightened by the realization that the only resource for P fertilizer manufacturing, phosphate rock, is a fossil material under threat of exhaustion in a near future. A still growing world population as well as a phosphate rock resource of lower quality, less accessible, with higher processing and shipping cost has led to a

1 steady increase in fertilizer price and consequently higher food price, while depleting the resource at an increasingly faster pace (Smil, 2000). The threat to worldwide food security makes phosphorus a new global sustainability challenge intertwined with other major threats such as climate change, water scarcity and energy production (Cordell and Neset, 2014). Therefore, reducing the diffusion of P to freshwater ecosystems to limit eutrophication is now part of a larger goal that targets a sustainable management of the P resource to guaranty long term food security. One of the main impediments to optimal phosphorus usage is the disparities between geographical areas in term of agricultural needs for P on one hand and production of P waste flows on the other hand. Crop-oriented agricultural regions rely heavily on mineral fertilizers while areas with intensive livestock farms struggle to find outputs for their P-rich manure. At a larger geographical scale, economic and development disparities also affect the utilization of phosphorus: while Western Europe continues to over apply phosphorus in agricultural soils with harmful impacts on the environment, countries in Eastern Europe have steadily relied on an undersupply of P to the crops due to their economic downfall following the collapse of the Soviet Union, diminishing progressively the crop yields (Csatho and Radimszky, 2009). A possible solution to these disparities could be to recycle phosphorus from livestock waste streams to produce concentrated, easily transportable P-based mineral fertilizers to be used in agricultural areas in need. Brittany, France is a good example of an agricultural area with excess P that has to be exported away. With a local climate enabling intensive livestock breeding and a soil poorly suited for crops, Brittany became the leading region in France for pig breeding during the mid-twentieth century and maintained its position until now, with 60% of the swine production on only 6% of the arable land in the country. The discrepancy between the amount of P-rich waste streams produced and the crop needs as well as the progress in domestic wastewater treatment make agriculture the main cause of P emissions to fresh water in the area, with 60% of dissolved P and 80% of particulate P (Colmar, 2017), despite an 80% decrease in P fertilizer in the last 20 years. Numerous treatment units have been implemented, mainly activated sludge systems initial solid liquid separation and treatment of the liquid phase with nitrification denitrification to convert ammonia into nitrogen gas. Since the beginning of the 21st century, 51 digestion units have also been erected, co-digesting animal manure with crop residues and industrial organic waste. Because phosphorus in pig slurry is mostly under inorganic, solid forms, it necessarily ends up in the solid phase at the end of either treatment pathways. In Brittany, this P rich solid phase is usually composted with the intent to be exported as an organic fertilizer in cereal producing regions, such as “la Beauce” in the south west of Paris. However, the product obtained is far less concentrated in phosphorus than mineral fertilizers and is usually sold at a very low price. Pig slurry is a suitable effluent to precipitate struvite (Kataki et al., 2016), a P rich, easily transportable mineral fertilizer. Recovering P from pig slurry as struvite would lower the transport costs compared to compost and maintain the organic matter locally. In order to maximize the amount of P recovered, it is necessary to dissolve it first by lowering the pH before a crystallization step at alkaline pH. Several organic acids have been tested for acidification, such as formate and acetate, leading to the dissolution

2 of 75-80 % of TP (Daumer et al., 2010). However, the cost of the acids was too high to compete with conventional mineral fertilizer and the environmental assessment was very poor, with chemical acidification as the main influence in the Life Cycle Analysis. Such acids can be avoided by taking advantage of a biological process known as acidogenesis: under anaerobic conditions, acidogenic bacteria in the slurry are able to produce organic acids from organic matter, leading to a pH decrease. In the context of anaerobic digestion development in Brittany, it could be interesting to place an acidogenic process prior to (or in parallel of) the digester in order to recover phosphorus and produce organic acids that would subsequently be converted to methane. The objective of this thesis was to study the biological acidification of pig slurry with various organic co-substrates and investigate the precipitation of phosphorus as struvite.

The first part of this manuscript is dedicated to the context of the study. The problematic of phosphorus regarding eutrophication, resource use and waste is investigated in a first section. In a second section, the agronomic effect of struvite as fertilizer is reviewed, as well as the existing processes for P recovery from domestic and livestock wastewater. In a third and last section, the biological process of acidogenesis is investigated, describing the acidogenic bacteria involved and their metabolic pathways, with a focus on lactic acid fermentation, a favored reaction regarding efficient phosphorus dissolution and recovery as struvite.

In chapter 1, biological acidification of pig slurry was performed in 96 h batch tests using sucrose as a model organic co-substrate and four different slurries. The evolution of pH, the organic acids produced and the dissolution of P, N, Mg and Ca were studied. A mass balance was calculated and expressed as carbon recovery. A simple model was set up to predict the lowest pH based on the slurry’s characteristics (initial pH and buffer capacity) and initial sucrose concentration.

In chapter 2, the same biological acidification protocol in batch reactors was tested on raw and digested swine slurry using sucrose and various organic wastes (potato, biscuit, buckwheat flour, apple, carrot, bean, pea and wheat straw). A microbiological study was conducted prior to fermentation and at the end of the batch tests in order to evaluate the changes in term of bacterial diversity and dominating groups. The co-substrates were characterized in term of acidifying capacity in order to expand the model predicting the lowest pH reachable.

In chapter 3, biological acidification of raw pig slurry was conducted in batch tests using carrots, green beans, peas and apples to verify the pH model validity. These co-substrates were chosen because they can easily be found as agricultural, agro-industrial or municipal waste in Brittany. Dissolution of P, Mg, N and Ca was investigated again. A semi-continuous reactor was set up to investigate the potential adaption of the biomass towards higher acidification and phosphate dissolution, using a mix of carrots and peas as organic co-substrates. The effluent was used to precipitate phosphorus and the dried solid obtained was dissolved in an acidic solution to study its ionic composition.

3

Finally, in the last part of the manuscript, the main results obtained are discussed and additional considerations are provided in order to improve upon this work and make the process fully implementable.

4

CONTEXT

II. PHOSPHORUS, A NUTRIENT AT A CROSSROAD

1. ROLE OF PHOSPHORUS IN NATURE AND ANTHROPOGENIC ACTIVITIES

Phosphorus is a fascinating nutrient at the nexus of several local, national and global problematics. It is at once critical to life on earth, severely limiting primary production in most ecosystems, paradoxically the main cause of eutrophication in fresh waters and, as recently realized, a rather scarce resource that could impact global food security in the mid-term (Elser, 2012).

Phosphorus is involved in the cell membrane through the bilayer of phospholipids, which make up 5- 10% of the cellular mass (Sterner and Elser, 2002). Even more importantly, adenosine monophosphate (AMP), diphosphate (ADP) and triphosphate (ATP) are the universal molecules acting as energy carrier in any living organism. Finally, phosphorus is involved in the genetic material and its expression through DNA and RNA. In humans, phosphorus constitutes a major part of bones and teeth under various calcium phosphate salts, or 0.7 Kg of total body mass (Ashley et al., 2011).

3- On Earth, two natural phenomena are responsible for (PO4 ) fluxes: weathering and leaching. Weathering corresponds to the chemical and physical processes responsible for exposed rock to decompose, while leaching is the phenomenon by which molecules are carried away from the soil by water flowing through. Weathering and leaching are relatively slow processes that represent the main constraint on terrestrial productivity (Lambers et al., 2008). Indeed, P is the least mobile and available nutrient in most soil conditions due to its high reactivity with other molecules causing a strong retention (Gull et al., 2004). As a result, P constitutes a limiting nutrient in agriculture, along with nitrogen and potassium (Cleveland et al., 2013). Some natural processes occurring locally have enabled sufficient replenishment of soil P to allow agriculture to flourish early in human history, such as the flooding of Nile and Indus basins (Nixon, 2004; Weil, 2006). However, under regular conditions, the phosphorus removed from soil through the harvest of crops needs to be replaced artificially (Ngoze et al., 2008). Humans understood that as long as 5000 years ago when they started to apply their excreta as well as animal manure to the fields (Mårald, 1998). Other secondary sources of phosphorus such as ashes and harvest residues were also applied (Cordell et al., 2009). Erosion, a major source of P loss, was controlled through terracing as early as 2000 BC in Arabia and later in all inhabited areas, including remote pacific islands (McNeill and Winiwarter, 2004). Complex sewer systems were implemented during Roman and Greek era, enabling both high hygiene levels in the cities and organic matter collection points outside the urban areas for reuse in surrounding agricultural fields (Dersin, 1997) This very efficient use and reuse of P made for a closed cycle that started to “leak” during the Middle Age with often a dysfunctional sewer system and a partial diversion of sewage towards nitrate production for gun powder instead of field application (Kirchmann et al., 2005). As a result of soil deterioration and deadly famines in Europe during the 17th and 18th century,

5 new sources of P were actively sought outside of the traditional and local resources (Emsley, 2000). Thus crushed bones were used as P-fertilizer in England and US in the early 19th century (Ashley et al., 2011). Shortly after, other concentrated sources of P were discovered such as guano and phosphate rock. Guano consists in a dense mix of sea bird excrement, feathers and bones, first found in Peru and south pacific islands (Hutchinson, 1950). Worldwide trade of guano and phosphate rock grew rapidly, transforming what had been a very local phosphorus loop into a global, linear flow. At the same time, widespread contamination of drinking water in large cities led to a sanitation/hygienist revolution, during which the implementation of modern sewer systems and water closet shifted the flow of human excreta from land application to water-based disposal. If landfilling and incineration of solid waste and development of wastewater treatment largely reduced the health hazard inherent to the waste, it also deprived arable land of a major phosphorus input (Mårald, 2002). Industrial, economic and scientific development, as well as the discovery of phosphate rock deposits and the Haber Bosch process for artificial nitrogenous fertilizer production have all enabled the green revolution to take place. Through massive irrigation, chemical fertilizer and pesticides application and new varieties of crops, agricultural yields doubled and sustained a 50% world population growth during the 20th century (Tilman, 1998). However, this new agricultural paradigm also led to a massive waste at every step of the food production chain. While population on earth increased by half, P fertilizer consumption was multiplied by 5 compared to pre-industrial times, which is far more than carbon (+13%) and nitrogen (+100%) (Falkowski et al., 2000). Put another way, only 20% of the P extracted to produce food actually ends up in our plates (Cordell et al., 2009). Two major leaks in the P cycle were originating from excessive application of chemical fertilizers on one hand and poor treatment of human excreta on the other hand. Widespread availability of cheap chemical fertilizer led to overuse in the developed countries while wastewater treatment initially focused only on reducing the oxygen demand without regard for nutrients. As a result, the P content of freshwater systems worldwide is now at least 75% greater than preindustrial levels, and oceans face an inflow of P from 8 MT per year to 22 MT per year (Bennett et al., 2001). As described in the introduction, the shift of water bodies from oligotrophic to eutrophic with regard to phosphorus and nitrogen has major and costly consequences. Beyond the physical, chemical and biological changes induced by eutrophication, there are also significant socio- economic impacts on health, commercial fisheries, tourism, recreational activities and mitigation measures undertaken (Anderson et al., 2000). The economic impact in the US alone was estimated very conservatively at 2.2 billion US$ annually (Dodds et al., 2009). It is through the prism of environmental damage and economic loss that new regulations regarding wastewater treatment were taken in the late 20th century in developed countries, focusing mostly on domestic wastewater (point source pollution) and much less on nutrients runoff from agricultural fields (diffuse source pollution). Ashley et al. (2011) described the livestock manure treatment processes as having a medieval level of sophistication. In any way, reducing P input to freshwaters from either agriculture or domestic waste through conventional treatment processes without return to arable soil does not disrupt the linear flow

6 of phosphorus and associated waste. The notion that phosphorus was a finite and irreplaceable resource threatening food security only came to the forefront after the major price hike (+700%) in 2007-2008 (Cordell et al., 2009). In the last ten years, rigorous research have eventually been conducted to assess the resource stock, quantify P flux, forecast future needs, identify causes of waste and suggest various initiatives to close again the phosphorus cycle.

2. PHOSPHORUS RESERVE, CURRENT AND FORECASTED USES

Large uncertainties remain regarding the amount and quality of phosphorus deposits in the world (Elser and Bennett, 2011). Beyond the inherent uncertainties regarding yet to be discovered underground resources, a major cause for this lack of knowledge stems from the low transparency of producing countries (Gilbert, 2009). Now that P has become a strategic commodity, producing countries are reluctant to disclose their reserves.

Current estimations have been re-evaluated upward, varying from 5.2 to 38.8 GT-P (Table 1) depending on the methodology and the type of resource considered (e.g. including or not additional resource not yet discovered, currently inaccessible with current technics or too costly for now). The reference in the field, the United State Geological Survey (USGS) gave its last estimate at 9 GT-P, considering only the P that can be extracted/produced economically under the current situation (Ober, 2017). The initial prediction made by Cordell et al. (2009), forecasting a P production peak in the short term (2035) followed by a sharp decrease due to a depleting resource has eventually been judged unlikely (Heckenmüller et al., 2014; Van Vuuren et al., 2010). Indeed, at the current extraction rate of 34 MT-P/year, the reserve could last for 265 years.

Table 1. Phosphate rock reserves estimations

P reserve (GT-P) Reserve Reserve base Source 5.4-12.9 7.1-38.8 Van Vuuren et al, 2010 7.9 Cooper et al., 2011 5.2 6.1 Sverdrup et al., 2011 7.3 Cordell et al., 2009 9 USGS (Ober, 2017)

Currently China is the main producer of phosphate rock, exceeding US production since 2006, with major changes in the repartition of producing countries in the last 20 years (Figure 1). There is still a large diversity of countries producing significant amounts of P rock. However, the extreme concentration of underground P resources in very few countries (even more than oil) or one could say, in one country alone (Morocco, Figure 2) could cause huge tensions for access to the commodity and major price hikes due to monopoly/oligopoly pricing (Elser and Bennett, 2011; Scholz et al., 2013; Taylor and Moss, 2013). Economic growth in a country would not depend anymore on access to

7 affordable oil, but affordable P, making the world a P-based economy instead of oil-based (Gilbert, 2009).

20 18 16 14 12

10 1997 8 2016 6 4

P-rock production (MT-P/year) 2 0 US Morocco China rest of the world

Figure 1: Production of phosphate rock in 1997 and 2016

10 9 8 7 6

5 1999 4 2017 3

P- rock reserves (GT-P) 2 1 0 Morocco China rest of the total world

Figure 2: Phosphate rock reserves in the world in 1999 and 2017 (USGS data)

As described earlier, phosphate rock production has increased steadily since the beginning of the green revolution, and is now more than 5 times what it was in 1960 (Figure 3). Indeed, 80 to 90% of the phosphate rock extracted is used to produce P fertilizers (Heifer and Prud'homme, 2014; Koppelaar and Weikard, 2013; Prud'homme, 2010; Smil, 2002), making fertilizer demand the main component for the increase in P rock extraction. Minor uses for P-rock are feed additives (7%), food additives (2%), detergents (5%) and industrial applications such as lightning and electronics (Scholz et al., 2014). In all these minors applications, P could actually be substituted with other compounds (Scholz et al., 2014). Indeed, the use of zeolite in laundry detergent has already decreased P consumption by half in that domain (Shinh, 2012). The slight recess in P rock production observed in 1990 corresponds

8 to the fall of the USSR. Since then, most of the P-fertilizer demand comes from China and the developing countries in Latin America (Figure 4), while the western world has reduced its use after 50 years of over application.

40

35

30

25

20

15

10

5 Phosphate rock production rock Phosphate (MT-P/year)production 0 1950 1960 1970 1980 1990 2000 2010 2020 year Figure 3: World phosphate rock production from 1960 to 2017 (USGS data)

20 8

18 7 16 6 14 12 5 10 4 demand growth rate MT-P/year 8 3 6 2 4 annual growthrate2015-2020 2 1 0 0 Latin america Central/East Africa Asia Oceania North Am. + world Eur.+Russia West. Eur.

Figure 4: P fertilizer demand and expected growth until 2020

Future demand for P-fertilizer depends on several factors: population growth, changes in food diet, evolution of agri-systems (intensive/extensive, specialized or mixed crop-livestock agriculture), trade fluxes (globalization vs. regionalization of the economy), residual-P content in soils, availability of fertilizers and whether or not active measures are taken to reduce the waste and recycle P. Since the beginning of the millennium, available phosphate rock (after beneficiation) has increased from 19.8 MT-P in 2000 to 20.5 MT-P in 2005, 22.8 in 2009 and 25.2 in 2016 (Heffer and Prud’homme, 2016). Until 2050, population is expected to grow by 30% (medium scenario) to reach 9.2 billion (FAO,

9

2013), requiring agricultural production to grow by 60-110% with crop demand increasing by 100- 110% (Alexandratos and Bruinsma, 2012; Tilman et al., 2011). Such demand might not be met due to underachieving gains in crop yields, well below what would be necessary (Ray et al., 2013). In any case, in order to sustain this population growth and associated need for larger food production, P- fertilizer consumption is expected to grow steadily until 2050. Koppelaar and Weikard (2013), having demonstrated the link between population growth and P fertilizer demand during the last 50 years, projected an increase of 40% to reach 30 MT-P produced per year in 2060 and stabilize afterward, considering a scenario “business as usual” with sufficient P rock resource. Due to the lower quality of the remaining phosphate rock, prices would increase to twice their current value, regardless of the reduction measures taken. Van Vuuren et al. (2010) also predicted a quick rise in P production until 2050 followed by a stable phase (23.8-42.7 MT-P/year), due to population growth, economic development and associated increased in kcal per person, as well as soil-P deficit in poor countries requiring rapid remediation. Reserves were forecasted as diminishing by 10 to 65% in 2100, indicating the absence of critical shortage but prices were expected to double or triple due to increasing extraction and transport costs. Scholz et al. (2013) estimated that despite the absence of a physical scarcity for P in the short term, economic scarcity (i.e. difficult access to unaffordable P fertilizer) is already a growing concern in poor countries, and increasing prices would only exacerbate it (Cordell and Neset, 2014). Others have found that no such increase in P production would be necessary to meet the demand for doubling current food production in 2050. Indeed, by taking into account the cumulative amount of P applied on crop soil and the cumulative amount of P exported with the crops during the period 1965-2007, Sattari et al. (2012) calculated that residual soil-P accumulated during those years would offset the increased higher crop yield and increased P fertilizer addition, resulting in a stable global P-fertilizer use of 16.8-20.8 MT-P/year, similar to the 18.5 MT-P/year consumed in 2016 (FAO, 2017).

3. MITIGATING MEASURES ENVISIONED TO CLOSE THE PHOSPHORUS CYCLE

As described above, P loss to the environment is both a threat to ecological systems (and humans when toxins are produced) through eutrophication of freshwaters and to food safety via a depleting stock. Closing the P cycle would mitigate both adverse consequences of P wasted to the environment. In order to do so, it is necessary to identify the mains flows of P and where they could be either reduced or re-directed toward safe crop fertilization. Replacing P with equally efficient compounds in detergents and other industrial use would have a significant impact on overall P consumption and this process is already underway in the European Union and the US (Scholz et al., 2014). However, as discussed above, the overwhelming majority of P is used in agriculture through P fertilizer and spreading of manure. As a result the main levers to improve P efficiency and reduce the amount of P extracted from the earth are to be found in the domain of agriculture. Similarly, within the P cycle, human excreta and landfilled matter represent respectively 2-5.7 and 2.5-3.8 MT-P/year (Reijnders,

10

2014; Scholz et al., 2014). In comparison P fertilizer and manure represent 14.2-20.4 and 5-28.2 MT- P/year respectively (Liu et al., 2008; MacDonald et al., 2011; Sheldrick et al., 2003). These are both the main flows of phosphorus within the system and the less efficient. Annually, 10-13.7 MT-P/year are lost from arable land into freshwaters while only 12-13 MT-P/year are taken up by crops (Koppelaar and Weikard, 2013; Scholz et al., 2014). Several explanations exist for this inefficient use of P. Firstly, the global amount of P mineral fertilizer used worldwide does not reflect the major disparities in application rates among countries and even within a country. Indeed, 30% of worldwide cropland is facing a P-deficit (Africa and South America) while 12.8% of arable land (located in western Europe, USA and East China) receives more than 50% of P fertilization (MacDonald et al., 2011; Potter et al., 2010). Excess fertilization results from the increase of livestock production, specialization and intensification of animal breeding and spatial disconnection between livestock and crops (Reijnders, 2014). The disparities at global scale are also found at domestic scale. Indeed, areas specialized in intensive livestock production often cannot find arable land within an (economically) reasonable radius to apply their manure due to high transport costs for a low fertilizing content (Kellogg et al., 2000; Schipanski and Bennett, 2012). Bateman et al. (2011) described the case of England, in which despite an overall soil P deficit and excess manure, full reuse is not achieved due to physical distance and time disconnection between crop need and manure production. A similar situation occurs in France too, with regions specialized in crops such as cereals and other focused on intensive animal breeding. The result is a high heterogeneity in soil P balance, which generates the need for imported mineral fertilizer and reduces the recycling of P from manure to crops (Senthilkumar et al., 2012). A local example would be Brittany, where swine producers struggle to find outputs for their manure, giving it for free or converting it to economically non-competitive organic fertilizers such as compost.

Part of the solution would be to come back to integrated crop-livestock systems with sufficient cropland available, where manure reuse enables high P efficiency (Russelle et al., 2007). This would require agricultural policy changes at national and continental level. At the technological level, a promising possibility would be to recover the P contained in animal manure under an easily transportable form, adapted to the needs of crop producers, and cheap enough to compete with mineral P fertilizers. This would reduce the over application of manure in intensive breeding areas, limit runoff and leaching to freshwater and diminish the need for imported mineral fertilizers, thus slowing down the depletion of P reserves.

Scientific research on P treatment from industrial, domestic and livestock wastewaters has long been focused on its removal in order to discharge clean water and limit eutrophication. However a major shift toward P recovery instead of mere removal occurred in the last 20 years. In the next sections, those technologies are investigated.

11

III. PROCESSES FOR PHOSPHORUS RECOVERY

As discussed in the previous section, recent concerns about P resource depletion have led the scientific community institutional and industrial stakeholders to shift their aim from phosphorus removal for clean water discharge to P-recovery under reusable forms for agriculture. A wide range of feedstock is suitable for P-recovery, with domestic wastewater and animal manure as the most abundant effluents and most adapted for P-recovery. Several industrial wastewater types with high P content also make for worthy sources to recover P from. Struvite (magnesium ammonium phosphate hexahydrate) is overwhelmingly the main product targeted, followed by calcium phosphate and K-struvite (in which magnesium is replaced by potassium). In this literature review, the agronomic value of struvite is examined, before describing the existing P-recovery processes for struvite as well as other P-based mineral fertilizers. This review highlights the need for phosphorus dissolution in pig slurry in order to achieve a significant recovery rate. Because the strategic option for P dissolution developed in this PhD is a biological acidification under anaerobic conditions using sucrose and organic waste as co- substrate, the final section of this review focuses on the metabolic processes involved in acidogenesis and particularly lactic acid fermentation.

1. AGRONOMIC VALUE OF STRUVITE

Struvite precipitation from waste streams requires very specific conditions that can be both difficult to attain and financially expensive (Hao et al., 2013). As a result, the added value of struvite has to be high enough in term of practical use, safety for the environment and agronomic value in order to justify its production.

So far, some advantages have been observed when struvite was applied compared to conventional fertilizers. The main quality of struvite is its low solubility leading to slow assimilation in arable soil. As a result, nitrogen leaching is severely reduced compared to conventional fertilizers (Rahman et al., 2011). Struvite is particularly suited for plants grown in pots since a lot of water is drained in this situation and a slow release fertilizer prevents nutrients runoffs (Antonini et al., 2012). Another advantage of the slow nutrient release provided by struvite is to prevent the chemical burning of crop when used in a single, large application (Fattah, 2010). Magnesium present in struvite also improves P uptake compared to conventional P fertilizer thanks to a synergistic effect on P absorption by the plant (Li and Zhao, 2003). Struvite effect on crop yield depends on the type of soil and nutrient content. Thus, better fertilizing effects were observed compared to chemical fertilizers in slightly acid soils (Barak and Stafford, 2006; Ghosh et al., 1996; Kataki et al., 2016), as well as P depleted soils (Vogel et al., 2015).

Many studies found at least similar performances compared to conventional P fertilizers (Achat et al., 2014a; Achat et al., 2014b; Gell et al., 2011; Li and Zhao, 2003; Liu et al., 2011; Pérez et al., 2009;

12

Plaza et al., 2007; Thompson, 2013) yet sometimes observed slightly lower yields (Ackerman et al., 2013; Ganrot et al., 2007). These lower performances seemed to be related to the alkaline pH of the soil, a condition that reduces the availability of struvite’s nutrients compared to conventional fertilizers (Ackerman et al., 2013). A potential long term issue with struvite application is the potential accumulation of magnesium, noted by Gell et al. (2011). Indeed, if natural variations of the Mg:Ca ratio don’t affect crop yields (Schulte and Kelling, 1985), an artificially large ratio induced by struvite application would change soil conductivity and aggregates stability, negatively impacting crop yields (Zhang and Norton, 2002). Besides, excess magnesium also reduce Ca uptake by certain plants, which could lead to calcium deficiency (Stevens et al., 2005). Finally, the low N:P ratio of struvite (1:1) is not perfectly suited for plant growth, which requires usually an N:P:K ratio of 4.5:1:0.94. Consequently, struvite should be used in P depleted soil or in combination with other fertilizers to adjust the NPK ratio applied to the crop.

Beyond its intrinsic characteristics, struvite recovered from waste needs to meet certain physical and chemical criteria in order to find a market. Thus, high crushing strength is absolutely required in order to ensure durability and prevent the loss of fine material during manufacturing, transport and field application. Highest strength was found in crystals of 2-2.5 mm (Fattah, 2010), while Forrest et al. (2008) advised for larger crystals (3-5mm) with sufficient strength as demonstrated by Adnan et al. (2003). When the crystals are too small and too brittle, only low economic returns can be expected (Forrest et al., 2008). Another critical aspect factoring into marketability of struvite recycled from waste is its content in heavy metals (HM) and pathogens. There is still a lack of information regarding the transfer of HM to the precipitate based on the waste stream used, yet most some studies indicate that the HM content of struvite was well within the European norm for fertilizers (Gell et al., 2011) and below the HM values found in phosphate rock (Wollmann and Möller, 2015). Pathogens have been found in struvite recovered from human urine, but improvement in the drying process (temperature, air moisture content) could mitigate this transfer (Decrey et al., 2011).

From this review it appears that a strong emphasis should be put on improving struvite crystallization process to meet the physical characteristics required by the fertilizer/agricultural sector. The low HM content compared to traditional chemical fertilizer should help in commercializing struvite recovered from organic waste. Further studies are needed to determine the best climate and edaphic conditions to maximize struvite efficiency (Kataki et al., 2016). Even though recovered struvite can be sold directly to large fertilizer manufacturers and incorporated in their process (El Rafie et al., 2013), the buying price would be low if the final product is not marketed as environment-friendly. Due to its relatively high cost of production compared to conventional P fertilizer from P rock (at least in the short term) , the most likely outputs for struvite will remain high value added crops such as golf turf, tree seedlings, ornamentals, vegetables and flower boards (Munch and Barr, 2001).

13

2. EXISTING PROCESSES FOR PHOSPHORUS RECOVERY

Phosphorus recovery technologies for municipal effluents and to a lesser degree industrial wastewaters have been largely developed since the 90’s, mostly as a mean to save money on maintenance, dewaterability and disposal costs of sludge. Meanwhile, recycling of P from livestock waste streams remains at laboratory and pilot scale. Before delving into the variety of stream sources and technologies used to recover P, a common aspect among all those processes is the precipitation of phosphate ions into low soluble salts that can be recovered. The most common precipitate targeted is struvite, formed according to Equation 1. In the case of calcium phosphate, different forms are possible: hydroxyapatite (HAP), hydroxydicalciumphoshate (HDP) (Equation 2), brushite or amorphous calcium phosphate. Usually Mg2+ and Ca2+ are the limiting ions for full P precipitation and are added as magnesium chloride and calcium chloride. Because precipitation only takes place at pH above 8 for struvite and above 9 for calcium phosphate, caustic soda (NaOH) or alkaline salts can be used to increase pH. Magnesium oxide (MgO), magnesium hydroxide (Mg(OH)2) or their calcium equivalents enable both pH and stoichiometric adjustment. Carbonate stripping through air bubbling is also a common technic used to elevate the pH.

Equation 1: Formation of struvite

Equation 2: Formation of HAP and HDP

a. TECHNOLOGIES FOR PHOSPHORUS RECOVERY FROM MUNICIPAL / INDUSTRIAL WASTEWATERS

In municipal wastewater treatment plants (WWTP), a significant fraction of dissolved P (30-40%) is taken up during biomass growth thus ending up in the sludge. In order to meet the discharge requirements, iron and/or aluminum addition leads to the co-precipitation of P, trapping it in the sludge phase too. A biological process known as Enhanced Biological Phosphorus Removal (EBPR) or Bio-P uses Polyphosphate Accumulating Organisms (PAO) to naturally store P in the sludge through the alternation of anoxic and aerated phases. Because the biomass can be inhibited by toxic compounds and variations of load and temperature, Bio-P is often used in combination with chemical precipitation to reach discharge limits. Use of EBPR is increasing in order to reduce the cost associated with Fe and Al addition. Besides, P recovery is much more difficult once P is bound to metals. Technologies have been developed to recover phosphorus at different location within a WWTP (Figure 5): - Final effluent (1) - Liquid phase of anaerobic thickener (2)

14

- Digested sludge (3) - Anaerobic centrate (4) - Thickened sludge (5) - Sludge ashes (6)

These processes are able to recover phosphorus initially present in the liquid phase, in the sludge or in sludge ashes.

Figure 5. Access points for P recovery in municipal WWTP (adapted from Egle et al., (2015))

Recovery from treated effluents

Only one technology exists to recover P from treated effluents of a municipal WWTP. The REM-NUT process uses two ion exchange units to remove phosphate, ammonium and potassium from low concentration effluents (PO4-P = 5-10 mg/L) (Liberti et al., 2001). Sodium chloride (NaCl) as brine from seawater is used to regenerate the ionic exchangers. pH is raised to 9.5 caustic soda and MgCl2 is added to form struvite and K-struvite. Up to 90% of influent P can be recovered using this process, yet low selectivity from the ion exchange unit for P leads to the adsorption of undesirable compounds, frequent regeneration of the exchanger (as often as every 3 hours (Brett, 1997)) and lower recovery (50-70%). This selectivity problem and its consequences severely impact the economic sustainability of the process. Logically this technology has been evaluated as the most expensive compared to the amount of P recovered (Egle et al., 2016).

15

Recovery from the liquid phase of anaerobic digestion

Most technologies for P recovery are designed to precipitate the dissolved P from the supernatant of anaerobic digestion (Desmidt et al., 2015) (stream 4 in Figure 5). These are the simplest and cheapest processes since they focus only on the dissolved phosphate of the liquid phase, but the recovery rate is usually very low, in the 10-25% range compared to P in WWTP influent (Egle et al., 2015; Melia et al., 2017). Indeed, most of the phosphorus is located in the sludge after anaerobic digestion (Pastor et al., 2008). Beside the low cost of these technologies, the main interest is to reduce the amount of P returned to the head of the treatment plant and thus reduce incrustation and scaling in pipes. Perhaps because the finality of these processes is not to reach a high recovery rate of incoming phosphorus to the WWTP, the scientific literature offers very little data on this parameter while describing extensively the proportion of phosphate (not total-P) recovered (Abma et al., 2010; Cullen et al., 2013; Moerman et al., 2009). Several versions of this type of process have been commercialized, some of which possess relevant improvements. Four examples are given below. The first describes the very basic technology for P recovery from digester supernatant. In the second one, a similarly simple process is applied to industrial wastewater. In the third example, the crystallization step is placed before centrifugation of the digestate, improving massively the dewaterability of the sludge. The fourth and final example presents a dual process recovering P from both digester supernatant and undigested EBPR sludge, improving the recovery rate significantly and reducing incrustation/scaling.

• Basic recovery from anaerobic supernatant

The oldest phosphorus recovery process on full scale is the Phosnix technology developed by Unitika in Japan in 1987 (Munch and Barr, 2001). The supernatant from anaerobic digestion containing 110mg of dissolved P per liter is fed at the bottom of a fluidized bed reactor. The bed contains fine particles of struvite acting as seed material on which newly formed crystals can grow. Mg(OH)2 and NaOH are added to reach a stoichiometric ratio of Mg:P and increase the pH to 8.2-8.8 respectively. Air stripping provides the necessary mixing to maintain the particles in suspension. When the crystal size reaches 0.5-1mm, the pellets are removed from the reactor while the liquid containing too fine crystals is returned to the reactor. The pellets are stored until the water content is less than 10% before being transported to a processing plant and used as a raw material to produce a fertilizer equilibrated in N, P and K (Ueno and Fujii, 2001). During three year study, Ueno could demonstrate that 90% of dissolved P entering the fluidized bed reactor was effectively recovered, leading to a 50% P-recovery at the scale of the WWTP. This high recovery value is unusual for such a process. Indeed, on the schematic displayed by Ueno, 55% of TP is dissolved in the digester supernatant, a surprisingly high value, compared to other digesters, e.g. <10% (Wild et al., 1997), 15% (Marti et al., 2008), 19% (Pastor et al., 2008), 20% (Martí et al., 2017), 23% (Cullen et al., 2013).

16

• Struvite recovery from potato processing and dairy effluents

The NuReSys® process is a completely stirred tank reactor (CSTR) with prior CO2 stripping treating anaerobic effluent from an upflow anaerobic sludge reactor (UASB) (Moerman et al., 2009). Sodium hydroxide and magnesium chloride are used to adjust pH to 8.5-8.7 and Mg/P ratio to 1-1,2. At a full scale dairy plant, no struvite precipitation occurred initially despite a 70% phosphate removal. The high calcium concentration compared to phosphate (140 vs 42 mg/L) caused the concurrent precipitation of an amorphous solid containing Ca, P and Mg. When the P content of the UASB effluent increased, calcium phosphate precipitation ceased, replaced by 100% struvite crystals of 2- 6mm, with 78% of TP removed. The same process was tested at pilot scale on a UASB effluent of a potato processing facility, reaching up to 88% TP (117 mg-TP/L and 15 mg/L in the influent and effluent). Very similar results (76% TP removal) were obtained on potato processing anaerobic effluents with the PHOSPAQ technology, a similar process using a CSTR reactor and integrated air bubbling for CO2 stripping (Remy et al., 2013). Therefore, struvite recovery from anaerobic effluent appears particularly suited to industrial effluents with high dissolved P and low particulate P. However, effluents with interfering ions, such as calcium in the case of dairy effluents might not the most suited effluent for P recovery as struvite.

• Precipitation of struvite prior to sludge dewatering: the AirPrex™ process

At the WWTP of Waßmannsdorf, Berlin (Germany), severe scaling was noticed in the centrifuge dewatering digested sludge as well as in the pipes leading to it and exiting it (Heinzmann and Engel, 2006). This resulted in operational disturbances and frequent interruptions (mechanical cleaning every two weeks). Uncontrolled precipitation of struvite occurred due natural CO2 degassing once the anaerobic sludge had exited the digester. Indeed, CO2 release increases the pH, lowering the solubility product of struvite. In order to reduce the incrustation problem, a controlled precipitation reactor was 2+ installed prior to sludge dewatering (Figure 6). MgCl2 provides the necessary Mg ions for full precipitation of phosphorus. CO2 is stripped on purpose through fine bubble aeration, raising the pH to 8 and creating an internal recirculating flow, allowing the struvite crystals to grow. The crystals are removed at the bottom of the air lift reactor while sludge flocs, lighter, are washed out toward the dewatering unit. 90-95% of phosphates are precipitated (Kataki et al., 2016). Struvite solids go through a sand washer for cleaning and purification, leading to low organic contamination. The product is marketed as a mineral fertilizer called “Berliner Pflanze”, sold since 2008 within the Berlin and Brandenburg region in retail stores or shipped to farmers/traders. The fertilizer contain a

N:P2O5:MgO ratio of 7%:21%:8%. The P recovery rate corresponds to 18% of total P entering the WWTP, while saving 474 000 € per year. Commercialization of the fertilizer represents only 10% of the savings, reduced maintenance 15% (less incrustation) and better dewaterability / lower disposal cost 75%. Five Airprex™ units have been built so far, in Germany and in the Netherlands.

17

Figure 6. Aiprex process for struvite recovery (adapted from P-rex)

• Enhanced recovery from anaerobic supernatant of digested EBPR sludge

When Bio-P is used on the mainstream line of the WWTP, phosphates are released by the PAO during anaerobic digestion but re-precipitate immediately (Doyle and Parsons, 2002; Marti et al., 2008). As a result, this uncontrolled precipitation leads to higher maintenance costs to remove the scaling, decreased dewaterability and lower P available for recovery (Cullen et al., 2013). Martí et al. (2017) showed that dissolved P concentration decreased during anaerobic digestion due to the release of poly- P by PAO and re-precipitation, leading to only 9% of TP entering the WWTP available for recovery. The Wasstrip® process (Figure 7) was developed to solve this problem by diverting a large fraction of the P flow from the digester to the crystallization stage (Cullen 2013). The struvite reactor (Pearl®) is a fluidized bed treating anaerobic supernatant of digested domestic EBPR sludge. Developed by the company Ostara, it was first installed in Durham, Oregon (USA) in 2009. In 2011 the aerated holding tank for EBPR sludge prior to digestion had its aeration turned off and volatile fatty acids (VFA) were added from the anaerobic primary sludge thickener. Under anaerobic conditions and in presence of VFAs, PAO release their stored phosphates and magnesium while taking up VFAs. Because ammonium concentration is still low compared to the digester where mineralization of organic matter leads to high ammonium concentration, no struvite precipitation takes place in the holding tank. The liquid phase, rich in P and Mg is sent to the struvite reactor while the impoverished sludge is sent to digestion. The anaerobic centrate from digestion, rich in ammonium, is then sent to the struvite reactor. The fluidized bed for struvite crystallization has several compartments of decreasing sizes from top to bottom, where MgCl2 is added. This allows a good segregation of crystals, with the largest at the bottom and smaller particles higher the reactor where they have longer contact time with the

18 liquid and more surface area. Besides, the high flow velocity at the bottoms leads to the wash out of remaining sludge solids that were not retained during prior centrifugation. Consequently, the struvite obtained as a low contamination in terms of organic content and pathogens. The product obtained is marketed as a mineral fertilizer “Crystal Green®” with an NPK content of 5:28:0 and 10% Mg (Britton, 2009). The P flow from the holding tanks represented 70% of all incoming P to the struvite crystallizer during a three year study, leading to a 60% increase in struvite production (Cullen et al., 2013) and an improved dewaterability of digested sludge (Britton et al., 2015). However, postulating a 7 mg-P/L in the influent to the WWTP (the actual value was not given), the recovered P represented only 16% of the total P flow. Egle et al. (2015) estimated the recovery rate at 20%. The Ostara Company has now 14 Pearl® reactors, 10 of which equipped or planned to be equipped with the Wasstrip® process.

Figure 7. Combination process of Pearl and Wasstrip

Phosphorus recovery from municipal WWTP digested sludge itself

Municipal digested sludge possesses all the necessary nutrients for plant growth (N, P, K, Ca and Mg), yet it also contains significant quantities of heavy metals (Ignatowicz, 2017; Lake et al., 1984), persistent organic pollutants (POP) (Jelic et al., 2011) such as pharmaceutical compounds and endocrine disruptors pesticides (Stamatis et al., 2010) and pathogens (Sahlström et al., 2004). Presence of these harmful compounds as well as uncertainties regarding plant availability of P (Kahiluoto et al., 2015; Kidd et al., 2007; Krogstad et al., 2005) make municipal digested sludge application in agricultural fields heavily regulated and suspicious by citizens of European countries (Ott and Rechberger, 2012). As described above, phosphorus in municipal digested sludge is mostly in the solid phase, making its extraction a necessary step if high recovery rate is desired. The existing processes

19 for P recovery from municipal sludge rely on energy and/or chemical intensive technologies associated with challenging post treatments for residual by products.

Pre-treatments of municipal sludge

• Thermal pre-treatments of municipal sludge

Wasted activated sludge (WAS) from WWTP using chemical precipitation for P is not impacted by thermal treatment. Indeed, Phosphorus is mostly present as an insoluble salt, e.g. FePO4 and cannot be solubilized when treated at high temperature (Appels et al., 2010). However, most of the P is released during thermal treatment of EBPR sludge, at temperature and duration as low as 70°C and one hour (Takiguchi et al., 2003; Xue and Huang, 2007). Thermal hydrolysis is always performed on thickened WAS and primary sludge, primarily to enhance sludge dewaterability and methane production during subsequent anaerobic digestion (Graja et al., 2005; Hii et al., 2014; Kepp et al., 2000; Pickworth et al., 2006). While moderate temperatures (70°C) can be used as pasteurization/hydrolysis treatment of WAS, most commercial technologies implemented on full scale are 30 minutes high-pressure thermal hydrolysis processes (HPTH) with temperature between 130 to 180 C and pressure in the 6–12 bar range (Morgan-Sagastume et al., 2011). HPTH is an efficient pre-treatment that has been enabled a near doubling in destruction of volatile content of waste activated sludge by anaerobic digestion (Norli, 2006). The increase in methane production could potentially compensate the energy necessary to produce the steam (Kepp et al., 2000). HPTH treatment of EBPR sludge combined with the Airprex process is currently under development in Germany (Egle et al., 2015). Laboratory studies have shown that thermal pre-treatments, by hydrolyzing organic matter into VFAs, can lower the pH and prevent re-precipitation of P. Yu et al. (2017) obtained an increase in dissolved P from 370 to 580 mg/L, or a 25% increase relatively to TP, at 220°C during 60 minutes.

• Wet chemical treatments

Contrary to thermal pre-treatment, wet chemical treatment is used on digested sludge where complex organic matter has already been reduced. One full scale wet chemical treatment technology for P recovery and two pilot scale units using different processes have been implemented so far.

The Seaborne process has been built on full scale in Gifhorn, Germany. The digested sludge is acidified with sulfuric acid, down to an unknown pH, leading to the dissolution of phosphorus and part of the HMs (Figure 8). After centrifugation the solids are incinerated. Dissolved HM are removed from the supernatant thanks to the hydrogen sulfide contained in the from the digester, thereby purifying it to be converted into heat. The HM-free liquid can then undergo P recovery using magnesium hydroxide and sodium hydroxide in two consecutive steps with pH adjusted to 9. Severe problems were encountered regarding the removal of precipitated HM, due to colloid formation passing through the filter and interfering with the subsequent P recovery step (Müller et al., 2007). The

20 process was modified: now, immediately after acidification, iron and other interfering ions are precipitated by adding sodium hydroxide and sodium sulfide to reach a pH of 5.7, and removed with the solids during centrifugation (Egle et al., 2015). Depending on the pH targeted during acidification with sulfuric acid, 30 to 90% of TP can be dissolved. P recovery rate is not described but Egle et al. (2015) estimated it at 50% of incoming TP to the WWTP.

Figure 8. Seaborne process (adapted from Müller et al. (2007))

In the Stuttgart process pilot installed at a WWTP in southwest Germany, the digested sludge is acidified using sulfuric acid down to pH 4 and centrifuged (Figure 9). Contrary to the Seaborne process, the dissolved metals are removed using citric acid as chelating agent (Antakyali et al., 2013). Magnesium oxide is added first to the supernatant in a batch reactor to increase magnesium concentration and sodium hydroxide is used to adjust the pH to 8.5. When the targeted pH is reached, the liquid is transferred to a conic reactor to allow sufficient retention time for crystal growth. The liquid is recirculated from top to bottom for mixing. The final effluent after struvite precipitation is returned to the mainstream treatment line of the WWTP. 700-800 mg-P/L were obtained from the acid dissolution of digested sludge, corresponding to 75% of TP in the sludge. 98% of the dissolved P was subsequently precipitated, leaving only 5-10 mg-P/L in the effluent. Even though the pilot did not treat all the digested sludge produced at the WWTP, it could recover approximately 35-45% of incoming P to the plant, two to three times as more than processes focusing on dissolved P in digested supernatant. However, the effluent returned to the mainline lead to accumulation of heavy metals in the sewage sludge and the P precipitate obtained had high contamination with Al, Fe and Ca. The chemical costs of the Stuttgart pilot process (construction and energy costs not taken into account) amounted to 2-3 € per Kg of struvite produced, or 15.8-23.7 € per Kg P. These costs are far above any potential selling

21 price for struvite fertilizer but once again the main interest of P recovery in municipal WWTP is to reduce the costs of maintenance and sludge disposal.

Interestingly, large variations of dissolved P occurred during acidification even at identical pH (e.g. 72% and 52% TP at pH 4.2). Antakyali et al. (2013) hypothesized that it could be due to differing quantity and type of phosphorus compound. Such phenomenon also occurred during the work undertaken in this PhD thesis, and similar hypothesis were made.

Figure 9. Stuttgart process

The final wet chemical technology for P recovery, currently at pilot scale, is the Budenheim process

(Figure 10). Contrary to the Seaborne and Stuttgart process, no sulfuric acid is used. Instead, CO2 gas is applied at a flow rate of 400 L/h at a pressure of 10 bar during 30 minutes, lowering the pH down to 4-5. The pH range represents a compromise between P recovery rate and dissolution of interfering ions and HMs. At pH 5, dissolution of P is not maximal but the HM remain in the sludge phase. The process enables 50% dissolution of TP without any need for chemical addition. 90% of dissolved P is then recovered as di-calcium phosphate by adding lime (CaOH).

As it could be expected, P recovery from municipal sludge using wet chemical treatment leads to significantly higher recovery rate. However, a balance has to be found between maximum P recovery and minimization of issues arising from the byproducts. Acidification below pH 4 led to foaming problems, the acidified sludge needs additional treatment (corrosion induced by high sulfur content), return of HMs to the mainstream treatment line and contamination of the product with iron.

22

Figure 10. Budenheim process

• Wet air oxidation (WAO) and supercritical water oxidation (SCWO)

Wet air oxidation involves aqueous phase deconstruction of organic and oxidizable inorganic components at sub-critical temperature and pressure (150–325°C and 20–175 bar) in the presence of oxygen The process leads to: (1) solubilization and removal of organic compounds, (2) reduction of mass and volume (>90%) and (3) destruction of pathogens and sterilization of material. One of the advantages of wet air oxidation is that dewatering of sludge is no longer necessary (Baroutian et al., 2016; Bernardi et al., 2010).

The level of solubilization and oxidation of organic matter increase with the amount of oxygen added, temperature and pressure (Chung et al., 2009). Under moderate conditions, wet air oxidation leads mostly to the formation of dissolved volatile fatty acids (Suárez-Iglesias et al., 2017). No dissolution of P occurs when chemical P sludge is used since iron ions re-precipitate concomitantly with phosphates to form FePO4 (Niewersch, 2013).

The Phoxnan technology avoids the precipitation of FePO4 by using a prior wet chemical treatment identical to the Seaborne and Stuttgart technology: acidification with sulfuric acid, this time at a pH as low as 1.5-2. The acidified sludge then undergoes a low pressure wet air oxidation treatment (160- 200°C, 20 bar) using the LOPROX® technology. Residual solids naturally settle while the liquid is filtered by ultrafiltration and nanofiltration in order to remove HMs, Fe and Al. Despite the use of advanced technology, only 55% of TP is obtained in the filtrate due to poor selectivity during filtration.

23

SCWO functions at pressure and temperature above 220 bar and 375°C, conditions in which water is above its critical point. The properties of water change regarding density, viscosity, diffusivity and solvating ability (water becomes a non-polar solvent). In this state where gas and liquid cannot be distinguished, mass transfer resistance disappears, favoring faster reactions (Goto et al., 1998; Svanström et al., 2004). When dioxygen is added, the organic matter is completely oxidized within a few minutes leaving only inorganic ashes and a clean liquid phase. Phosphorus needs to be extracted from the ashes using acid or alkaline treatment. 80-100% of TP was dissolved using acid leaching but iron was also leached, making FePO4 the only recoverable product (Levlin et al., 2004). Alkaline leaching enables the selective precipitation of phosphorus but initial dissolution is significantly lower. Consecutive acid and alkaline leaching was tested, leading to an increase in P dissolution compared to alkaline only treatment (+23%, reaching 53% of TP). The cost of chemicals remained acceptable (1.2 €/Kg-P) but did not include the costs for subsequent precipitation (Levlin, 2007). Two processes have already been tested at pilot scale, yet no upscaling has apparently been planned (Egle et al., 2015).

• P-recovery from sewage sludge ash (SSA)

Due to concerns about pollutants in sewage sludge, a large fraction of it is now incinerated (100% in the Netherlands).The sludge is first dewatered to >25% dry solids content and introduced in a fluidized sand bed at 800–900 °C in presence of oxygen (Figure 11). Organic matter, including organic pollutants is oxidized to CO2, NOx and SOx gases. The ashes generated contain all the P initially present. Fe and Al, as well as significant quantities of HM similarly end up in the inorganic ashes (Adam et al., 2009). P concentration is usually around 60 g-P/Kg, within a range going from 50 to 110 g/Kg (Donatello et al., 2010). The mineral form and proportion of P in SSA is actually similar to that of low grade phosphate rock (average phosphate rock contains 130 g-P/Kg). This phosphorus is mostly present as whitelockite (Sturm et al., 2010) a mineral insoluble in water and therefore not available to plants. As a result, further processing is necessary to remove the HMs and convert P to a water soluble form. A process that can be used is actually one of the technologies developed for fertilizer production from phosphate rock: sulfuric or phosphoric acid is added to SSA, producing single super phosphates (Ca(H2PO4)2) (Donatello and Cheeseman, 2013). The RecoPhos® process produces single super phosphates by adding phosphoric acid to SSA. The final product can be sold as a mineral fertilizer with a P content of 16.6%, demonstrating similar effect on crops compared to conventional fertilizer, a logical result since the product is very similar. The contamination with HM is low compared to phosphate rock but mostly because phosphoric addition in the process reduces the relative concentration of HM compared to P. The fate of Fe and Al is not described, but they are likely still present in the fertilizer product. This type of P extraction from SSA enables a 99% P recovery, corresponding to 85% of P entering the WWTP (Egle et al., 2015).

24

In order to remove the HMs from the final product, a stepwise increase in pH after dissolution enables the segregated precipitation of P and HMs (Petzet and Cornel, 2011). In the Sephos process, sodium hydroxide is added to reach pH 3.5, precipitating aluminum phosphate while trace metals remain in solution (Schaum et al., 2007). Al-P can be used as a raw material by the fertilizer industry. After further alkaline treatment on the aluminum phosphate product, a re-precipitation of P with calcium is possible (Schaum et al., 2005).

Another strategy to separate trace metals from P is to use ion exchangers. Xu et al. (2012) used hydrochloric acid to dissolve P from SSA rather than sulfuric acid alleging its poor solubilizing capacity (Cohen, 2009), though others found similar capacity between HCl and H2SO4 (Biswas et al., 2009). By applying 25mmol HCl per gram SSA during 2 hours, 95% of TP was dissolved, along with significant amounts of Cu, Pb, Zn, Cr and Ni. Then a cation exchange resin was added to the liquid phase to remove the trace metal elements. Struvite was then precipitated with the addition of MgCl2,

NH4Cl and NaOH. The final product exhibited 97% struvite purity.

Finally, the AshDec® process, operated at pilot scale in Austria, is able to remove part of the HM by feeding the SSA into a rotary kiln heated up around 1000°C with the addition of calcium, potassium or magnesium chloride. The HMs combines with chloride transfer to the gaseous phase, leaving the system with the exhaust before being removed during flue gas cleaning (Petzet and Cornel, 2011). A removal rate of 70-90% was reached for Cd, Cu, Pb, Zn, Mo and Sn, while As, Cr and Ni concentrations remained unchanged (Adam et al., 2009). The product obtained depended on the chloride source: Chlorapatite (Ca5(PO4)3Cl) when CaCl2 was used and stanfieldite (Ca4Mg5(PO4)6) when MgCl2 was used (Peplinski et al., 2009). A recent modification using sodium sulfide led to the precipitation of rhenanite (CaNaPO4).

Figure 11. Ash Dec process

25

Conclusion on phosphorus recovery from WWTP

Phosphorus flow to WWTP in Europe (0.4 MT-P/year) corresponds to 20% of imported P or 37% of chemical P fertilizer (Withers 2015). Recovering P from this stream or “secondary deposit” would have a major impact in term of resource efficiency and advance toward closing the P cycle. However a little less than 30 full scale plants for P-recovery currently exist in Europe (50 in the world). All of them are located in Western Europe (and the US). Most only recover phosphorus from the supernatant of digested sludge. The main incentive for implementing such processes is related to the cost associated with maintenance (less incrustation/scaling), chemical demand (flocculants and Al/Fe for chemical P removal) and disposal of residual sludge. Indeed, the market for mineral P fertilizer derived from municipal sludge does not really exist yet, due to a heterogeneous legislation regarding this type of product and the uncompetitive price compared to conventional P fertilizer. However, the large benefits regarding cost saving enabled by such technology should already appeal to many treatment plants. Because the product obtained has very good fertilizing value, less metal contamination than conventional fertilizer, and limited chemical and energy costs, the P-recovery technologies from digester supernatant will likely see a large development in the near future. The processes recovering P from the sludge itself or its ashes could theoretically reach much higher recovery rate compared to dissolved P technologies, yet much optimization is needed. The primary issue in this type of process is the high contamination in HM, requiring complex technology and high costs for removal. Beyond the technical and economical barrier, the environmental impact of these processes remains largely unknown, or if such assessments have been conducted, they are rarely in the public domain. One life cycle analysis was conducted on struvite recovery from human urine provided by source-separated sanitation system (Remy and Ruhland, 2006). A positive balance was found compared to conventional treatment of phosphorus, yet such study should be conducted in every case using a uniformed method in order to be able to compare the processes together. This task was very recently undertaken by (Egle et al., 2016) leading to the following conclusions:

• Recovery processes for dissolved P in digested sludge obtain relatively clean products (HMs, POP, pathogens), significantly less contaminated than the original source, and with a pollution potential inferior to that of a commercial fertilizer. They are plant available and perform as well as chemical P fertilizer, depending on the soil properties and climate. Nevertheless, the recovery rate of these processes is low relatively to the P flow to the WWTP (25% maximum). • Recovery from sewage sludge requires significantly more complex technology, enabling nonetheless high removal of HMs and pathogens, with medium recovery rate (40-50%). • Sewage sludge ash appears as the most promising source of P in municipal WWTP, with processes reaching 60-90% recovery rate. Because sludge incineration is centralized, regrouping several WWTP, economy of scale is possible. Pathogens and organic pollutants are obviously removed during the process, yet the costs associated with this type of process

26

strongly depend on the SSA quality, i.e. HM contamination. Synergy with other industrial processes nearby is strongly advocated as a mean of cost saving.

These recommendations are quite reasonable in the context of large, semi-public WWTP situated in heavily populated areas. Beyond the lack of environmental assessments for advanced, energy and technology intensive processes, their cost and the practical implications of implementing such large units in agricultural areas make them irrelevant for the treatment of livestock effluents. Dissolution of solid P is indeed a necessity to achieve a good recovery rate but chemicals and energy intensive processes are not adapted. Existing lab scale and full scale unit for P recovery from animal waste streams rely on the same type processes, either simple technology applied to dissolved-P only or chemical/energy intensive technology to access solid P. These processes are described in the next section.

b. PHOSPHORUS RECOVERY FROM LIVESTOCK EFFLUENTS

In Europe, P flow from animal manure represents 1.5 times the amount of chemical P fertilizer applied (1.6 and 1.1 MT-P/year) (Withers et al., 2015). In France, both flows are approximately equal (0.31 and 0.29 MT-P/year) ((Senthilkumar et al., 2012). The P balance in soils of Europe and France is still largely in excess despite a significant decrease in the last 30 years. Indeed, 42% of P currently applied is not exported and accumulated every year (21% in France). As described previously, these mean values hide the massive spatial variations in P applied and soil P accumulation/deficit across Europe and between regions of a same country (Csatho and Radimszky, 2009; GisSol, 2011), mostly due to livestock density and associated manure application in the fields nearby. Therefore, the challenge is to recover manure P from intensive livestock regions under a form that can be transported easily and at low cost to regions with soil P deficit. Intensive pig breeding is the primary livestock production in need of manure processing (Schoumans et al., 2015) due to its enclosed nature, without enough arable land for spreading. Currently, excess pig manure is exported mostly as compost for the solid phase, while the liquid phase undergoes nitrification/denitrification for nitrogen removal. Very few technologies for P recovery from pig manure or any livestock effluent currently exist on full scale. Therefore, this section will describe both these full scale processes and the one only tested at lab scale. The types of P recovery techniques for animal manure are similar to the ones in municipal WWTP. Thus P can be recovered from the liquid phase of digested or undigested manure, P can be recovered after pre-treatment of manure to extract P from, or concentrate in, the solid phase, and finally P can be converted into ashes through incineration.

Phosphorus recovery from the liquid phase

Two full scale units currently exist to recover phosphorus from animal manure. The first one treats 115,000 m3 of denitrified calf manure in Putten (Netherlands) since 1998, producing K-struvite. The process consists in 3 consecutive CSTRs, a clarifier and a struvite buffer tank (Schuiling and Andrade,

27

1999). pH is increased to 9 using MgO. Few details are known regarding the influent or effluent characteristics. Since the manure has already been denitrified, ammonium concentration is very low. As a result, the rise in pH leads to the precipitation of K-struvite, a process that would not occur at high ammonia concentration due to the lower solubility product of N-struvite compared to K-struvite. No information is given concerning the proportion of dissolved P vs. TP or the recovery rate. The solid recovered had low calcium or ammonium contamination (<2%) and a relatively low amount of organic matter (5.5%). Suspended solid concentration had a negative impact on the process when above 1 g/L, a phenomenon also noticed by Shen et al. (2011) and Zeng and Li (2006). Cattle slurry has a very high suspended solid content, and even after centrifugation the supernatant can still contain 10 g-TSS/L or more (Zeng and Li, 2006). The negative impact of suspended solids on struvite precipitation has also been observed in the case of pig slurry, with a decrease in the kinetics of precipitation, contamination of the solid obtained and smaller crystals (Capdevielle et al., 2016; Cerrillo et al., 2015; Zhang et al., 2012)

The second full scale unit is operating in North Carolina, recovering phosphorus as calcium phosphate from the supernatant of centrifuged raw swine manure since 2003 (Vanotti et al., 2007) (Figure 12).

The liquid phase undergoes nitrification denitrification in order to remove nitrogen as N2 gas and reduce alkalinity. Then, lime (Ca(OH)2) is added to adjust the pH to 10.5-11 and promote calcium phosphate precipitation. The solid obtained is withdrawn from the settling tank afterward and filtered with the help of an anionic . N removal and associated alkalinity consumption helped reduce the amount of lime needed to 567mg/L, for 118 mg-P/L were removed. The crystallization step enabled a removal rate of 95% of soluble P entering the reactor. However, most of the phosphorus has been removed during the initial solid liquid separation step. As a result, only 20% of TP contained in the raw slurry is recovered as calcium phosphate. P contained in the solid phase is composted at a centralized solid processing plant, and therefore can be applied on nearby arable soil. However, compost is not as concentrated in P as a mineral fertilizer and cannot be transported very far away. The solid obtained contained 10.7%-P and a moisture content of 77%. No details were given regarding the use or commercialization of the product.

28

Figure 12. P-recovery process for pig manure. Adapted from Vanotti et al. (2007)

• Lab studies on supernatant of digested manure

No full scale process currently exists to recover P from the liquid phase of digested manure. Most lab studies on struvite precipitation from these effluents focus on ammonium removal to meet discharge standards. Because nitrogen is mineralized during anaerobic digestion while phosphorus remains solid, ammonia is in large excess compared to phosphate in the supernatant. Consequently, when ammonium removal is the goal, a source of phosphorus has to be added, on top of magnesium a chemicals from pH adjustment. The P recovery rate is therefore irrelevant in these studies and rarely mentioned.

Karakashev et al. (2008) tested a multi steps process in which swine raw manure was first digested, the supernatant used for P recovery before a second chemical oxygen demand (COD) removal stage in UASB reactor and finally N removal using partial nitrification denitrification. The raw manure had 1.6 3- g-TP/L, the supernatant after digestion 600 mg-TP/L including 160 mgPO4 -P/L. MgO was added in excess of P by 30% (molar ratio Mg:P) and pH adjusted to 8.5. 96% of phosphate precipitated, + corresponding only to 26% of TP in the supernatant and 10% of TP in raw slurry. 245 mg NH4 -N/L were removed, an amount in excess of what struvite precipitation stoechiometrically leads to. Based + on the phosphate removed, it can be inferred that 175mg NH4 -N were lost as ammonia emission. The excess of magnesium severely disrupted the UASB reactor, and struvite recovery was discontinued. Inhibiting effect of cations on anaerobic digestion has been noticed elsewhere in the scientific literature (Chen et al., 2008; Soto et al., 1993). In this PhD, magnesium hydroxide is used to adjust the pH and steer the precipitation toward struvite rather than calcium phosphate. As a result, magnesium concentration after crystallization can be in large excess (up to 100 mg/L) and its effect on subsequent anaerobic digestion should be investigated.

29

Because magnesium is always the limiting nutrient regarding struvite precipitation, MgO and MgCl2 were often added as a magnesium source (Cerrillo et al., 2015; Perera et al., 2007; Perera et al., 2009). Bittern, a residual product of seawater evaporation after halite removal was also used as a cheap source of magnesium removal (Ye et al., 2011). NaOH was systematically added for pH control, sometimes in combination with CO2 stripping through air bubbling (Cerrillo et al., 2015; Song et al., 2011). Air agitated reactors can reduce the cost of chemicals, it also results in massive ammonia stripping, as high as 90% of ammonia removed (Song et al., 2011). Sequencing batch reactors and continuous feed reactors have been tested for phosphorus precipitation. Retention times in SBR were lower than in continuous reactors, with 0.6-4 h (+ 30 minutes settling) and 6-20 h respectively. Interestingly no fluidized bed reactors have been tested for P recovery from supernatant of digested animal manure. When mixing is brought only by a rotating propeller, there is a high risk that precipitation will occur onto it (Le Corre et al., 2009).

Metals such as iron and aluminum are in low concentration in pig manure, even more so in the digester supernatant. Zinc and copper content can be relatively high in the remaining suspended solids of the supernatant, especially when compared to phosphorus, but the solids recovered after P recovery had a contamination below legal limits in all the studies that measured it. Thus (Cerrillo et al., 2015) found that Zn and Cu accounted only for 0.3% and 0.006% of the recovered solid, while struvite represented approximately 50%, due to organic contamination. To improve the purity of the crystals, solid accumulations devices can be added, usually made of stainless steel wire mesh. Perera et al. (2009) found that such device did not increase the P recovery rate, but the 25% of recovered P accumulated on it were far less contaminated in organic matter and free of trace metals (Cu, Zn) compared to the settled solid recovered at the bottom of the reactor. (Song et al., 2011) obtained a higher proportion of recovered P on the accumulation device compared to Perera et al. (2009), with 40% of all recovered P on the stainless steel mesh, during both SBR and continuous operation. Overall, clean struvite from the collector represented 30% of phosphate initially contained in the supernatant. He also found that the struvite had less impurity on the accumulating device, with struvite representing 90% of the solid recovered on the accumulating device versus 67% in settled solids. Therefore, accumulating devices situated above the settling zone offer the possibility to recover a purer product. It should be noted that no such device exists in full scale processes for digester supernatant such as Pearl, Phospaq or Nuresys. It would have to be mechanically extracted from the reactor and scrapped, which does not seem suitable for a full scale process.

All the studies on pig slurries found an optimal pH at 8.5-9.5. Concurrent precipitation of calcium phosphate as amorphous calcium phosphate (Ye et al., 2011) and calcium carbonate (Song et al., 2011) was observed in several studies, especially when high pH (10 or above) was tested or an insufficient amount of excess Mg was introduced, regardless of the initial P:Ca ratio.

30

In the case of digested dairy manure supernatant without pretreatments, several studies have been conducted so far for P recovery. Contrary to the full scale unit in the Netherlands treating calf manure, they investigated struvite precipitation. Uludag-Demirer et al. (2005) exclusively focused on ammonia removal and added 4.3 g-P/L as Na2HPO4 (300mg-TP/L initially present, unknown proportion of phosphates), for only 500 mg NH4-N/L. With the addition of MgCl2, logically more than 95% of ammonium precipitated thanks to the large excess of Mg and P. Calcium is not mentioned but it is likely that the solid obtained contained some calcium phosphate.Zeng and Li (2006) noticed that simply by raising the pH to 9, 30% of the phosphates precipitated, without any Mg addition, concluding that a mix of struvite and calcium phosphate precipitated. An enormous excess of Mg compared to P was necessary to recover a high percentage of P (73% of P recovered at a molar Mg:P ratio of 22:1). This severe inhibition of P precipitation likely resulted from a very high ionic strength, suspended solid concentration and COD content. Similarly only 50% of ammonium precipitated when P:N ratio was adjusted to 1:1. The absence of K struvite precipitation confirms that N-struvite has a far higher formation reactivity preventing potassium removal as struvite when ammonium is in excess.

Most studies on P recovery from the supernatant of digested swine manure had a phosphate concentration between 22 and 160 mg/L and a TP concentration (though often not given) below 160 mg/L, except for Karakashev et al. (2008) cited above. Phosphorus content in digested dairy manure 3- supernatant was slightly higher with 100-450 mg-P04 -P/L. Phosphorus concentration in raw manure can vary widely, depending on the type of livestock, handling process at the farm and resulting suspended solid content. However, it can be safely assumed that phosphates available for precipitation in the supernatant of digested manure do not represent more than 20% of TP (Barnett, 1994; Daumer et al., 2008; Toor et al., 2006). It appears that the supernatant of digested swine manure is more suited to struvite recovery than cattle/dairy manure due to the high ionic strength in the latter, as well as the difficult solid/liquid separation and associated high suspended solid content.

• P recovery from raw swine manure (recovery of dissolved P only)

Multiplies studies have been conducted on raw swine manure, with differing objectives. P-recovery as struvite was often seen as a mean of decreasing the dissolved phosphate or ammonium concentration. Thus, Ryu and Lee (2010) focused on ammonium removal: they used a CSTR reactor with mechanical mixing and MgCl2, K2HPO4, NaOH to adjust the N:P:Mg ratio to 1:1:1 and increase the pH to 9 with an hydraulic retention time (HRT) of 16-48 h. A nitrification denitrification reactor received the effluent from the struvite reactor. Overall, complete ammonium removal was achieved, with 80-90% during struvite precipitation. Because the raw slurry had 2-5 g-TSS/L and no struvite accumulation device was used, the solid obtained had a high contamination by organic matter (35%). TP content of the slurry was relatively low, 100-200 mg-TP/L, with only 30% as phosphate. 95% of TP was effectively removed from the effluent, but the solid obtained potassium, sodium, calcium and silicon

31 on top of the organic matter making it unsuitable for processing as a mineral fertilizer without further treatment. Similarly, Rahman et al. (2011) achieved a 92% PO4-P recovery rate (no TP given) in an aerated reactor with MgCl2 addition and an HRT of 18 hours. However, due to the settling of half of the suspended solids, struvite represented only 6.6% of the solid recovered. Air agitation also led to the removal of 30% or 1.2 g/L of ammonia. With the same reactor configuration Liu et al. (2011) observed the co-precipitation of half of dissolved Na, K, Zn, Cu, and Ca in the influent. Taddeo et al. (2016) obtained relatively pure struvite without a dedicated accumulation device in an aerated batch reactor with Mg(OH)2 addition, an aeration time of 1.5 hour and settling of 30 minutes. However, the phosphate concentration had been multiplied by 75 (from 64 to 4776 mg/L) to reach a molar ratio N:P of 1:1. As a result, the amount of struvite is artificially inflated compared to the suspended solids initially present and compared to the co-precipitation of phosphorus with calcium.

Suzuki et al. (2005) used an aerated reactor with a settling zone and a stainless steel mesh wire recovery device, with no addition of magnesium. The effluent contained 50% less phosphate 65% less TP. The struvite recovered on the accumulated device was very pure (98%) without Zn and Cu, but represented only 4.5% of TP introduced in the reactor. In a follow-up study Suzuki et al. (2007) added bittern to reach a Mg:P ratio of 1.6:1 and improved the recovery device. During a 70 days period, the process achieved a 66% phosphate removal, a similar TP removal, with 50% of TP in the settled solids and 18%-TP in very pure struvite (95%, no Zn or Cu).

All the studies on P recovery from raw swine manure never used truly “raw” swine manure and systematically screened it through a sieve with 0.5-1.5 mm mesh opening. The studies conducted on actual raw slurry spiked large amounts of phosphorus to match ammonium concentration. As a result, despite high P removal rate, the actual amount of P recovered was either artificially inflated or representing a relatively low amount compared to the true initial TP of the slurry. Compared to the supernatant of digester, the overall recovery of phosphorus is largely higher with raw slurry when the settled solid are taken into account (50-70% of TP). However without a dedicated struvite collector the settled solids are similar to the solid phase obtained from a centrifuge or a screw press, with high organic matter and suited only to composting. With struvite collector, the maximum P recovery obtained was 18% TP. In order to achieve higher recovery rate, it is necessary to dissolve the inorganic P first.

Pre-treatments prior to phosphorus recovery

As discussed above, dissolving P is a necessity to achieve decent P recovery rates. Several pre- treatments of dairy and swine manure have been tested at lab scale with the objective to evaluate the extraction of P from the solid phase. They involve wet chemical treatments (chemical acidification), and thermochemical processes.

32

• Acidification of animal slurry

Pig slurry Daumer et al. (2010) studied the recovery of P as struvite from pig manure using organic acids (formate and acetate). Because the P recovery step would occur prior to anaerobic digestion, sulfuric acid was not used due to its inhibitory effect while organic acids could be converted to methane. The pig manure had already undergone nitrification denitrification in order to decrease the buffer capacity. 7 and 20 g of formic and were necessary to decrease the pH to 4.5-5, leading to an increase in ortho-P from 267 mg/L to 920-930 mg-P/L, or 77% of TP. Because simple decantation was used to recover the supernatant, the mass balance indicated that “only” 50% of initial phosphorus was available in the liquid phase despite 77% dissolution. The ratio of acid to dissolved P was approximately 7.5 g-formate per gram P. Supernatant of settled untreated pig manure was then added to the acidified slurry to provide the necessary ammonium for struvite precipitation.

Szögi et al. (2015) developed a process called “quick wash” consisting in applying citric or hydrochloric acid to the solid fraction of swine manure (213-300g-TS/Kg) to lower the pH down to 3-

5 and dissolve P. After centrifugation, Ca(OH)2 is then added to the liquid fraction rich in P and poorer in N (partly undissolved during acidification) to increase the pH to 8-10 and precipitate P as calcium phosphate. An organic anionic polymer is then added to enhance the precipitation. Dissolution of P was a function of the amount of acid added, and citric acid was slightly more efficient than HCl. Two g of manure and 50 mL of acid were mixed together, leading to dissolution of 87% TP (or 3.5 g-TP per Kg of manure) with 10 mmol/L citric acid and 40 mmol/L HCl. Lime at different concentrations was added to the supernatant obtained after centrifugation of the acidified manure, at a ratio of 2% w/v

(Ca(OH)2). This ratio seems to indicate that 20 g of supernatant were added into 1L of lime solution. After lime and polymer addition, 87% of TP initially contained in the manure was recovered in the precipitate at pH 8. The precipitate contained 5.6% P. Overall the process enables the recovery of 80% TP or 5.7 Kg-P per ton of “solid” manure with, according to the author, 4.8Kg of citric acid, 40 Kg of lime and 0.35 Kg of anionic polymer. However, the amounts of citric acid and lime seem to be ten times lower than what was used in the experiments. Indeed, 50 mL of citric acid at 10mmol/L mixed with 2 g of manure corresponds to 0.25 mol/Kg manure, and with a molar mass of 192 g/mol, one obtains 48 g-citrate/Kg manure or 48 Kg per ton of manure, not 4.8 Kg per ton. As a result, the economic analysis seems flawed. It claims that to process one ton of manure, the cost of chemicals would be 11.93 $, while the product has a fertilizer value of 22.43 $ on a P2O5 basis. That would be a profit of 50% considering a selling price equal to that of regular fertilizers, a very conservative hypothesis. Considering the more likely amount of citric acid necessary, the selling price would cover 60% of the chemical cost. No analysis of heavy metals, organic pollutants or pathogens has been conducted a prerequisite for commercialization.

33

Dairy manure Zhang et al. (2012) postulated that most of the P in dairy manure is under an inorganic, solid form, bound to calcium. As a result, to recover P as struvite it is necessary to separate it from calcium by acidic dissolution. HCl was used to lower the pH to 4 in the supernatant of digested dairy manure, dissolving 90% of TP (dissolved P increased from 20 to 250 mg/L). In order to re-precipitate P as struvite, a chelating agent was added to bind the dissolved calcium and MgCl2 was introduced in large excess to promote struvite formation. The tests confirmed that P is initially bound to calcium and that adding EDTA and Mg enable struvite precipitation. However, the amount of EDTA and Mg necessary is too large to make the process economically viable.

• Thermochemical pretreatments

Pyrolysis of swine manure Pyrolysis is the thermal degradation at ambient pressure of inorganic material under anoxic conditions and at temperatures between 400 and 800°C (Furness et al., 2000). Before pyrolysis the material needs to be dried (dry matter > 95%).Azuara et al. (2013) studied the extraction of phosphorus by chemical acidification after pyrolysis of dried pig manure. The manure was first pressed and then dried. The dried fraction contained 94% of TP initially contained in the manure. The solids were introduced in a -1 fluidized bed reactor at a rate of 0.2 Kg.h . A N2 flux was used to maintain anoxic conditions. The retention time was set at 90 minutes, and three temperatures were tested: 400-500-600°C. 92-96% of P contained in the dried manure remained in the solid phase, called “char”, at the end of pyrolysis. In order to extract phosphorus from the char, the solids were dissolved in a solution of sulfuric acid at 200 mmol.L-1. 90% of the P from the char was leached, 60 to 75% as orthophosphates. The author also used incineration on the dried slurry and found that all of the P could be dissolved as orthophosphate with 5 Kg-H2SO4 per Kg-P. Consequently, he advised that pyrolysis should be used to obtain a liquid fraction rich in dissolved organic (called “oil”, with the properties of diesel oil), while the char would be incinerated and P leached from the ashes. Indeed, leaching from the char after pyrolysis required 100 Kg of acid per Kg of P. Another solution is to use directly the char as a fertilizer.

Hydrothermal treatment and acid leaching of P from manure These pretreatment can be conducted on the raw manure itself and it does not require concentration/drying processes. Ekpo et al. (2016) studied the effect of temperature and type of acid or base on P dissolution in batch experiments lasting one hour, at 120, 170, 200 and 250°C. Without acid, temperature had a negative on P dissolution, with 500 and 10 mg-P/L after the 120°C and 250°C treatment respectively. The best results were obtained at 170°C with 0.1M H2SO4.79% of TP initially present in the manure was dissolved, 92% of it as ortho-P. The ratio of sulfuric acid to dissolved P was

5.6 Kg-H2SO4/Kg-P, a value similar to what is obtained by direct acidification on raw manure. It

34 should be noted that in this experiment, the swine manure was dried at 60% and re-dissolved afterward to undergo thermal treatment. Consequently, it was not direct use of raw manure. In a similar experiment, Heilmann et al. (2014) tested hydrothermal treatment on pig slurry during 40 minutes runs at 200-260°C. Hydrochloric acid was not added before treatment, but used instead on the char obtained. 90% of the P contained in the char was extracted jointly with calcium, indicating that most of the calcium in the pig manure was initially present as calcium phosphate. The acid to P ratio to reach this level of extraction was 108 Kg-HCl / Kg-P. Struvite was tentatively precipitated from the + acid extract by adding stoichiometric amounts of Mg and NH4 , but only calcium and phosphate precipitated. It appears that extraction of P from biochar is too expensive and the product obtained does not possess sufficient added value.

Hydrothermal treatment of cow manure was evaluated in two consecutive studies by Dai et al. (2015,2017) with the objective to immobilize P in the char in the first experiment, and extract it with hydrochloric acid in the second experiment. Hydrothermal treatment alone (200°C, 4-24 h) enabled a large increase in phosphorus content in the solid phase and resulted in a lower extractability. This resulted from a conversion of water-extractable P and Ca into apatite. Depending on the objective for manure processing, hydrothermal treatment could be used as a mean to concentrate and immobilize phosphorus into the char. In the second study, Dai obtained the full extraction of TP with a 2% HCl solution. More than 90% of dissolved P was present as ortho-P. The acid to P ratio was strikingly low, with 2.2 Kg-HCl per Kg-P. There might be some ambiguity regarding the volume in which dissolved P is measured. No explicit value for TP in the initial manure was given. If the acid to P ratio is exact, the manure would have a P content of 4.2% P (% DM), a surprisingly high value, since most cow manure have a P content around 1-2% of dry matter.

Microwave heating (MW), acidification and wet oxidation of dairy manure with hydrogene peroxide

Pan et al. (2006) applied H2SO4 + MW + H2O2 pre-treatment to dairy manure based on its success on domestic sludge. The results indicated that MW in itself does not affect P dissolution, contrary to what occurs with EBPR sludge. Cell lysis induced by MW is efficient to release P from PAO in EBPR sludge but has no effect on inorganic P as calcium phosphate in dairy manure. Only temperature at

170°C, H2SO4, and H2O2 treatment led to the release of phosphates. MW only seemed to slightly accelerate the effect of H2O2 on P dissolution. No details were given regarding the exact amount of sulfuric acid used, making it impossible to calculate the ratio of acid to dissolve phosphorus.

Qureshi et al. (2008) similarly found a positive effect of temperature but no effect of hydrogen peroxide on P dissolution during MW treatment in combination with H2SO4. Unfortunately, the effect of H2SO4 alone is never presented in any of the publications on MW treatment of cow manure. Again, the exact concentration of H2SO4 was not given. However, pH was decreased to 1.4, indicating a rather large amount of acid and making acidification the main reason for P dissolution. Phosphate concentration increased from 40 mg/L (or 20% TP) to 160 or 86% at 170°C, with and without H2O2.

35

Less efficient results were obtained by Jin et al. (2009). MW treatment alone did not lead to any dissolution of P. With 2% H2SO4 + MW, dissolved P increased from 20%-TP to 39%. A similar result was obtained with a lower amount of H2SO4 (0.5%) and the addition of H2O2 (1%), indicating that both acidification and wet oxidation increase P dissolution. It should be noted that if the H2SO4 added at 2% v/v is pure acid, then the acid to dissolved P ratios are extremely high, with 80-320 Kg-H2SO4 per Kg of P. The pre-treatment was used for both struvite precipitation and anaerobic digestion. MW alone did not increase the methane production, and H2SO4 logically inhibited anaerobic digestion due to sulfur inhibition.

Cow manure is extremely challenging regarding solid liquid separation (Hotaling, 2006), as described earlier. The resulting high suspended solid concentration severely impedes struvite crystallization. In order to obtain a liquid with low SS content and high ortho-P/TP ratio, Zhang et al. (2015) tested several pre-treatments on liquid cow manure. An unknown amount of sulfuric acid was added initially to dissolve P by lowering the pH to 3-4. The acidified manure underwent a microwave (MW) process with the addition of H2O2. Solid and liquid were separated through a simple clarification stage. Microwave and oxidation massively lowered the SS content by solubilization/destruction and by improving the settleability of the remaining suspended solids. By clarification, 70% of the volume was recovered as SS-free supernatant. Initially, 40% of the 220 mg-TP/L were ortho-P. With acidification to pH 4, the ratio of ortho-P / TP increased to 60%. With acidification plus MW-H2O2 (0.1-0.3%

H2O2), the ratio increased to 90-95%, indicating that the increase in dissolved P was due in equal part to acidification and MW-H2O2. Oxalic acid was added to the supernatant to bind to calcium and prevent re-precipitation of phosphorus with it. A ratio of 2 moles oxalic acid per mole Ca was enough to remove 90% of Ca without binding with Mg. In the final stage, struvite was precipitated without addition of Mg since P was the limiting nutrient. 95% of P was removed from the liquid. No details were given regarding pH adjustment. The product was very pure, between 91 and 97%, with very small amounts of Ca, Na, and K.

MW, hydrothermal treatment and pyrolysis have positive effects on transforming the organic matter into a useful liquid product. However, on top of the relatively advanced technological level involved in these pre-treatments, their effect on P dissolution in very limited or absent. Only chemical acidification has a significant and positive effect on P dissolution with as much as 95% TP solubilized into ortho-P. H2SO4 was slightly more efficient than HCl in certain studies and both were equivalent in others. The exact amount of acid was not described in most studies on MW pre-treatment but it seems to have been used in very large excess.

A recapitulative table (Table 2) sums up the amount and type of acid used, as well as the ratio of acid to P dissolved during acidification of swine manure and sewage sludge ashes. It appears that ashes from swine manure and sewage sludge give similar results in term of acid to P ratio. Acidification from char or raw manure requires significantly more acid. While no hypothesis was made by Azuara et

36 al. (2013) regarding the higher need for acid during P extraction from char compared to ashes, the reason for higher acid need in raw slurry is quite straightforward: many compounds in raw slurry can act as pH buffer and some of them have been removed from the ashes are transformed into other compounds without buffering capacity. In raw slurry, the ratio of acid to dissolved P is minimized when TP concentration is high and on the contrary the ratio is larger at very low TP concentration, since inorganic P is not the only compound reacting during acidification.

Chemical acidification, contrary to more advanced processes, does not require large investment costs, massive energy consumption, and skilled operators. Microwave, hydrothermal carbonization, pyrolysis and incineration are processes suited for very large treatment plants. In rural areas with low population densities, P recovery processes should be kept as simple as possible to be installed on site or regrouping other livestock breeders living nearby. As a result, chemical acidification is the most low tech process, relatively easily applied locally. However, it still constitutes the main post for expenses for struvite recovery from swine manure. In Szögi et al. (2015), after correcting the amount citric acid necessary, acidification represented 76% of chemicals price. (Daumer et al., 2010) found that acidification represented 85% of chemical costs and 65% of the overall process counting investment, energy and maintenance.

Table 2. Amount of acid needed to dissolve P under various forms

Source of P Type of substrate acid mmol H+/g-P g-acid/g-P

Sewage sludge SSAa HCl 455 16.6 (Schaum et al., 2007) SSA HCl 145 5.3 SSA HCl 200 7.3 (Montag and Pinnekamp, 2009) SSA HCl 142 5.2 SSA HCl 222 8.1 (Xu et al., 2012)

SSA H2SO4 271 13.3 (Franz, 2008)

SSA H2SO4 171 8.4

SSA H2SO4 194 9.5

SSA H2SO4 87 4.3 (Takahashi et al., 2001) SSA H2SO4 232 11.4 (Biswas et al., 2009)

Swine manure Ashes H2SO4 102 5.0 (Azuara et al., 2013)

Char H2SO4 367 18.0 Raw formic 152 7.0 (Daumer et al., 2010) acetic 333 20.0 Solids (30% DW) citric 35 6.7 (Szögi et al., 2015)

Raw H2SO4 180 8.8 This author

H2SO4 352 17.2

H2SO4 268 13.2 a SSA: sewage sludge ash

37

Biological acidification of pig slurry using organic co-substrate could reduce significantly the cost of P recovery from pig manure by replacing the chemicals by free (or cheap) organic waste. Biological acidification of pig slurry has never been studied with the objective of P recovery, however it has been used in lab study for other purposes such as two-stage hydrogen and methane production or ammonia emission control.

Strong acidification down to pH 1-2 is not necessary since a large fraction of TP in swine manure can be dissolved at pH 4-6 (Christensen et al., 2009; Daumer et al., 2010; Sharpley and Moyer, 2000). Such pH can be reach during biological acidification. Hjorth et al. (2015) obtained a pH of 4.3 with 30 g-glucose per liter of swine manure, as a mean of reducing ammonia loss. Schievano et al. (2012) obtained a pH between 5-6 using swine manure and market biowaste as organic co-substrate. He similarly reached a pH below 5 with vegetable market waste at a ratio of 35% (weight waste / weight manure). As described by Tenca et al. (2011), this fermentation process facilitates the subsequent methane production step. Therefore, a struvite recovery step could be put prior to anaerobic digestion, with dissolution of P and hydrolysis or organic matter in the first stage, solid liquid separation to recover P as struvite, and a return of the solid fraction and liquid phase after P recovery to an anaerobic digester.

IV. ACIDOGENESIS As discussed above, biological acidification (or anaerobic acidogenesis, or fermentation) of pig slurry using organic co-substrates appears as a cheap, low-tech process yet efficient enough to bring the pH down to 4-6, a range in which P dissolution is relatively high.

Acidogenesis is a microbial process in which organic matter is hydrolyzed and converted into organic acids, alcohols, hydrogen gas and CO2 under anaerobic conditions. This process takes place naturally in soil, human/animal gut, silage and fermenters. It is usually presented as the second step of anaerobic digestion, after the hydrolysis step of macromolecules, and producing the VFA and hydrogen and CO2 gas that are used thereafter in the acetogenic and methanogenic steps to produce methane (Figure 13). The amount and type of product obtained depends on the organic substrate (carbohydrate, , and content), the present and the environmental conditions (pH, substrate concentration, temperature).

38

Figure 13. metabolic pathways for the degradation of organic matter under anaerobic conditions. Adapted from Moletta (1993)

Manure and sewage sludge have been used for biohydrogen and methane production in what is known as a two-stage process. Biohydrogen production takes place under slightly acidic conditions and occurs concomitantly with conversion of complex organic matter into VFA (Xing et al., 2010), a process similar to what would be needed to dissolve P in swine manure. Because the P recovery process investigated in this PhD is conceptualized as a pre-treatment for anaerobic digestion, the existing two-stage processes are of interest, by providing hindsight into the optimal conditions for acidification and its consequences in the subsequent methanisation stage. Fermentation of waste with or without anaerobic inoculum has also been the subject of many scientific investigations in order to produce biofuels or molecules with high added value such as lactate, succinate, propionate and butyrate. These scientific studies can provide information regarding the types of metabolite produced for each kind of waste and the conditions applied.

One process uses acidogenesis with the same purpose as this PhD research: ensiling. This process consists in the storage or crops under anaerobic conditions in order to preserve it all year long and use it as an energy rich feed for ruminant livestock (Dunière et al., 2013). Silage relies on lactic acid fermentation of water soluble carbohydrates leading to a quick drop in pH, inhibiting global bacterial activity thus preventing further degradation of organic matter. This fast acidification is exactly what is sought for P dissolution in pig slurry. The fact that lactic acid bacteria (LAB) are found in large amounts in pig slurry (Snell-Castro et al., 2005), food waste and municipal organic waste (Abdel- Rahman et al., 2013; Bonk et al., 2017; Probst et al., 2015; Tang et al., 2016; Tang et al., 2017; Yin et

39 al., 2016) tends to indicate that lactic acid fermentation could be possible in swine manure. Hjorth et al. (2015) effectively obtained a high lactate concentration and low pH using glucose as organic co- substrate in swine slurry. Lactate as an (almost) exclusive end product of acidogenesis has two advantages: high acidification efficiency and low buffer capacity during pH adjustment for struvite crystallization. The efficient acidification is due to the low pKa of lactate compared to other organic acids (Table 3). Malate, succinate and formate are often transient and used as intermediaries, not end products contrary to lactate. From this low pKa results a low buffer capacity, i.e. a low proportion of protonated acid in the targeted acidification range (4-6). Only 5% of lactate is protonated at pH 5, versus 36-43% for acetate, propionate, butyrate and valerate, the other major products of acidogenesis (Figure 14). During pH adjustment to promote struvite precipitation these protonated acids would act as acidic buffers, releasing their proton into solution and increasing the need for alkali to increase the pH. Consequently, this section on acidogenesis is oriented toward lactic acid fermentation.

Table 3: pKa of organic acids produced during acidogenesis

pKa lactate 3.86 succinate 4.2-5.6 malate 3.4-5.2 formate 3.77 acetate 4.76 propionate 4.88 butyrate 4.82 valerate 4.82

Figure 14: proportion of protonated acid as a function of pH

40

1. CATABOLISM OF ORGANIC MATTER UNDER ANAEROBIC CONDITIONS

Carbohydrates, lipids and amino acids can be used by bacteria under anaerobic conditions to produce simpler organic compounds and energy (as ATP) for maintenance and growth (Figure 15). In absence of oxygen acting as final electron acceptor, the reactions mediated by co- such as nicotinamide adenine dinucleotide (NAD) require a neutral electron balance. appears a major hub in the acidogenesis process from which most final compounds originate. Interestingly, lipids, polysaccharides and amino-acids can all be converted to pyruvate and subsequently organic acids or through a various number of redox reactions. It should be noted that several major product of acidogenesis such as lactate, acetate, propionate and ethanol can be reused, transformed and combined to generate other compounds. Some of those reactions will be described in a dedicated section below.

Figure 15: metabolic pathways during acidogenesis. Dotted lines correspond to secondary fermentation of primary molecules

a. CATABOLISM OF AMINO ACIDS

During acidogenesis, two processes are used by bacteria to degrade amino acids. The Stickland reaction consists in a paired deamination redox reaction between two amino acids. As shown in Figure 16, an amino acid, e.g. alanine is oxidized to form pyruvate while the electrons are used to reduce two glycines. Ammonia is released and energy is recovered through formation of three acetate molecules

(Thauer et al., 1977). The second process is the decarboxylation of the amino acid, releasing CO2 and an amine (Equation 3). The balance between deamination and decarboxylation is effectuated as a mean

41 to adjust the pH toward neutrality. Production of organic acids by deamination takes place at alkaline pH and tends to lower it. Decarboxylation occurs at low pH, the amines released having an alkaline effect on pH (Stancik et al., 2002). Stickland reactions can represent up to 90% of protein degradation (Nagase and Matsuo, 1982) leading to a majority of acetate as end product. Thus Liu et al. (2012) found that 80% of the organic acids produced during fermentation of protein rich sewage sludge were acetate (Equation 4). Similarly Yin et al. (2016) found that 70% of the COD from the proteinaceous substrate (peptone) converted to organic acid was acetate, using sludge from a UASB reactor as seed. The pH was adjusted to 6 throughout the batch test study.

Protein conversion to organic acids seems to be partial most of the time, contrary to carbohydrates which are often completely converted to organic acids. Thus, Yin et al. (2016) had protein conversion to VFA of 40-70% while glucose conversion was 88%. Similarly, Liu et al. (2012) found that 40-50% of sludge were converted to VFA despite a thermochemical pre-treatment of the domestic sludge. Interestingly, despite an incomplete protein conversion to organic acids (65%) a high lactate concentration was obtained by Shen et al. (2017) during the fermentation of egg white with UASB sludge as inoculum, pH adjustment at 6 and a temperature of 30°C. 22% of the proteins (in COD) were converted to lactate by Leuconostoc and Lactobaccillus. The incomplete conversion of proteins to organic acids could simply be due to ammonia toxicity, since protein degradation releases ammonium. + NH4 -N concentration was as high as 6 g/L in Shen’s batch experiment.

Figure 16: Stickland reaction between alanine and glycine

42

Equation 3. Decarboxylation of amino acid

Equation 4. Acid products from sewage sludge fermentation

b. CATABOLISM OF LIPIDS

Hydrolysis of lipids leads to glycerol and long chain fatty acids (LCFA). Long chain fatty acids can be converted to acetyl-CoA and an acyl-CoA with two less. The reaction is continued until all the LCFA is converted to acetyl-CoA. This reaction is thermodynamically unfavorable and thus requires ATP. Butyrate and acetate can then be produced. Glycerol enters as dihydroxyacetone phosphate, which leads to phosphoenolpyruvate, itself converted to a of pyruvate. Then, as described in Figure 15 pyruvate can theoretically be converted into a wide variety of products such as acetate, propionate, butyrate lactate, ethanol and valerate. However, Yin et al. (2016) obtained mostly propionate, acetate and butyrate (Equation 5), with COD loss corresponding well with the theoretical biohydrogen production. Once again it should be noted that pH was maintained at 6, which necessarily impacts the metabolic pathways occurring in the mixed culture.

Equation 5. Metabolic pathways for glycerol degradation during acidogenesis (Yin, 2016)

c. CATABOLISM OF GLUCOSE

Under anaerobic conditions, most bacteria are using the Embden-Meyerhof-Parnas pathway (EMP, or glycolysis) to oxidize glucose to pyruvate. In the process two moles of ATP and two moles of reduced Nicotidamide Adenine Dinucleotide (NADH) are produced (Equation 6).

Equation 6. Pyruvate formation from glucose Glucose + 2 ATP + 2 NAD+ + Pi → 2 pyruvate + 4 ATP + 2 NADH

Pyruvate can then be converted to lactate thanks to the lactate dehydrogenase . Another pathway is the cleavage of pyruvate to Acetyl-CoA and formate by the Pyruvate Formate Lyase (PFL). Formate is converted further to molecular hydrogen and by the Formate Hydrogen Lyase (FHL) complex. This pathway is present in Enterobacter for example (Converti and Perego, 2002). In the absence of an exogenous final electron acceptor, the different pathways used for the degradation of pyruvate and the relative proportion of each final compound are driven by the need for

43 the bacteria to regenerate NAD+ for glycolysis and maintain the redox balance. Thus, this can be achieved by producing two moles of lactate per mole of glucose (homolactic fermentation, see dedicated section below) or a 50/50 mix of acetate and ethanol (Alam and Clark, 1989; Redwood et al., 2008). pH also affects the relative proportion of end-products. The conversion of formate to H2 and

CO2 depends on pH and formate concentration (Hallenbeck, 2009; Rossmann et al., 1991; Zhang et al., 2005). Below pH 7, the FHL is progressively activated to reach full activity at pH 6, countering the acidifying effect of formate.

Clostridia, a bacterial group present in large proportion in pig slurry can degrade pyruvate to a mix of acetate and butyrate during acidogenesis (Equation 7 and 8).

Equation 7. Acetate formation from glucose

C6H12O6 + 2H2OH → 2C 3COOH + 2CO2 + 4H2

Equation 8. Butyrate formation from glucose

C6H12O6 H→ C 3CH2CH2COOH + 2CO2 + 2H2

Pyruvate is converted to acetyl-CoA and CO2 thanks to the pyruvate ferredoxin oxidoreductase enzyme (PFOR), which reduces ferredoxin. A can then convert two protons to hydrogen by re-oxidizing the ferredoxin. This assures the production of two moles of hydrogen per mole of glucose consumed. Under low H2 partial pressure, the acetyl-CoA is converted to acetate, producing one ATP and allowing the NADH produced during glycolysis to be re-oxidised to NAD+ while reducing ferredoxin via an NADH:ferredoxin oxidoreductase enzyme (Hallenbeck, 2005). The reduced ferredoxin is used subsequently to produce two extra H2 per mole of glucose (Jungermann et al., 1973). Under higher H2 pressure, acetyl-CoA is converted to butyrate (Chatellard et al., 2017;

Hallenbeck, 2009), in which case NADH is used in the process and no extra H2 is produced. The maximum VFA and hydrogen production occurs during the initial exponential growth phase (Lay, 2000). pH once again affects the distribution of end product by modifying the enzyme activity in the different catabolic pathways (Zhu and Yang, 2004). Similarly to Enterobacter, the redox balance through the relative proportion of the reduced and oxidized form of NAD is also a crucial parameter to explain the changes in the distribution of end products (Wang et al., 2012).

Propionibacterium (Papoutsakis and Meyer, 1985), Bacteroidetes (Macy et al., 1978) and several Firmicutes pertaining to the class of Negativicutes (Reichardt et al., 2014) are able to produce propionate and acetate (Equation 9) from glucose using the succinate pathway (Figure 15).

Equation 9. Propionate formation from glucose

3 C6H12O6 → 4 CH3CH2COOH + 2 CH3COOH + 2 CO2

44

2. LACTIC ACID FERMENTATION a. LACTIC ACID BACTERIA

Lactic acid bacteria comprise a diverse group of Gram-positive, non spore-forming bacteria. They occur as cocci or rods and generally lack catalase. They are chemo-organotrophic. LAB are present in high amounts in foods (dairy products, fermented meat, beverages), on plants (epiphytic bacteria responsible for lactic acid fermentation in silage), in sewage, and finally in the genital, intestinal and respiratory tracts of mammals (Holzapfel and Wood, 2012). LAB are facultative anaerobic or micro aerophilic organisms able to grow in a wide range of temperature, from 5 to 45°C, and a large range of pH, 3.2-9.6 (Wang et al., 2015). They require rich and complex nutrients (amino acids, vitamins, minerals) and rely on carbohydrates for growth. Most known LAB species belong to the genera Lactobacillus, Lactococcus, Pediococcus, Aerococcus, Carnobacterium, Oenococcus, Tetragenococcus, Vagococcus, Weisella, Leuconostic, Streptococcus and Enterococcus. Two metabolic pathways exist for the catabolism of carbohydrates: homolactic and heterolactic fermentation (Figure 17). In homolactic fermentation, hexoses are oxidized to pyruvate through the EMP pathway while pentoses are converted to -6P in the pentose phosphate pathway and then enter regular glycolysis to be converted to pyruvate. Pyruvate is reduced to lactate by an NAD linked lactate dehydrogenase enzyme (nLDH). By producing two moles of lactate per mole of glucose, or 5 moles of lactate for 3 moles of pentose, NAD is regenerated and available for glycolysis. In heterolactic fermentation, glucose is converted to a pentose, ribulose-5P and CO2, leading to glyceraldehyde-3P (GAP) and acetyl-P thanks to the phosphoketolase enzyme. Acetyl P can be converted to acetic acid or ethanol while GAP can be fully converted to lactate. As a result, the yield on glucose is only 1 mol of lactate, half of the homolactic fermentation pathway. The proportion of acetate and ethanol depends on the redox potential in the bacteria, ethanol production regenerating NAD while acetate does not (but generates ATP). LAB can be classified as obligate homofermenters, producing only lactic acid from hexose but unable to degrade pentose. Facultative heterofermenters use hexose and pentose to produce lactic acid and small amounts of acetate/ethanol. Obligate heterofermenters only use the phosphoketolase pathway (Oude Elferink et al., 1999). nLDH does not exist solely in LAB, as seen above in the case of Enterobacter. Two types of nLDH lead to two isomers, L and D lactic acid (Garvie, 1980). LAB can form one or the other, or a racemic (1:1) mixture. Conversely, certain homolactic fermentation LAB have the capacity to synthetize other enzymes and adapt their catabolism routes to environmental changes. Thus in presence of excess pyruvate from outside redox reactions (e.g. from citrate), Lactococcus lactis is able to produce diacetyl and acetoin using the enzymes diacetyl synthase and acetolactate synthase, when pH is low and carbohydrates are in limited supply (Salminen and Von Wright, 2004). Other LAB such as Lactobacillus casei possesses the PFL system evoked earlier in the case of Enterobacter. This pathway is used under substrate limitation and long retention time, leading to a change from

45 homolactic fermentation to mixed acid fermentation with the production of lactate, acetate, formate and ethanol.

The assertion that LAB can only grow on carbohydrates is not considered absolute any longer. Thus ATP can be obtained from glutamate, arginine and histidine (Christensen et al., 1999; Fernández and Zúñiga, 2006). Pyruvate can be formed by LAB from Aspartate, cysteine, methionine, serine and alanine, meaning that lactate could be produced from the catabolism of these amino acids (Fernández and Zúñiga, 2006). This could be an explanation for the high lactate concentration obtained from egg white by Shen et al. (2017).

Figure 17: catabolic pathways in LAB for carbohydrates (adapted from Wang, 2015).

Enzymes 1, galactokinase; 2, arabinose isomerase; 3, xylose isomerase; 4, mannose phosphotransferase system; 5, hexokinase; 6, glucose-6-phosphate dehydrogenase; 7, 6-phosphogluconate dehydrogenase; 8, ribulose-5-phosphate 3- epimerase; 9, transketolase; 10, transaldolase; 11, 6-phosphofructokinase; 12, fructose-bisphosphate aldolase; 13, triosephosphate isomerase; 14, lactate dehydrogenase; 15, phosphomannose isomerase; 16, phosphoglucose isomerase; 17, phosphoglucomutase; 18, galactose-1-phosphate uridyl transferase; 19, glucosyltransferase; 20, ribulokinase; 21, xylulokinase; 22, phosphoketolase; 23, acetate kinase; 24, phosphotransacetylase; 25, aldehyde dehydrogenase; 26,

46

b. ENSILING PROCESS

As described earlier, ensiling is an anaerobic process designed to preserve crops to be used as feed for ruminants. The characteristics of silage (type of organic matter, bacteria present, but not water content) and the phenomena taking place (rapid acidification by lactic fermentation and secondary fermentation of lactate in spoiled silage) are similar to what could happen during the biological acidification of pig slurry. Consequently, exploring the metabolic processes occurring during ensiling should help understand and optimize the biological acidification of swine manure.

Initially (Figure 18), the remaining intact plant cells and aerobic microorganisms consume the oxygen left by oxidizing easily available carbohydrates, such as glucose and fructose. Once oxygen is depleted, facultative heterolactic LAB naturally present on the crops (such as Lactobacillus plantarum) are the quickest to develop on water soluble carbohydrates compared to the other bacteria or even other LAB, thanks to their capability to use hexose and pentose for lactic fermentation (Beck, 1972; Dunière et al., 2013; Grazia and Suzzi, 1984; Holzer et al., 2003). Once pH has dropped acid tolerant and acetate tolerant LAB can colonize the silage, usually strict heterofermenters such as Lactobacillus Brevis, L. buchneri and Leuconostoc (McDonald et al., 1990; McDonald, 1982). pH is stabilized around 4 and no hydrolysis of organic matter occurs, thanks to the inhibiting effect of pH, undissociated lactic acid, and synergetic antimicrobial effect of lactic and acetic acid produced by heterofermenters such as L. buchneri (Adams and Hall, 1988; Holzer et al., 2003).

When the toxic effect of lactic acid fermentation is not enough (too much residual oxygen initially, too high water content, not enough epiphytic LAB to start with), several bacteria can grow and use lactate, leading to the degradation of silage protein content and production of toxic compounds for animal and human health. pH re-increases due to the conversion of lactate into other organic acids with an higher pKa (Dunière et al., 2013). This secondary fermentation in which lactate is converted to other VFAs when the source of carbohydrate has been exhausted is not unique to silage. It happens in human/animal gut and plays an important role in digestion. The metabolic routes by which lactate is consumed are explored in the next sections.

Figure 18: Shifts in bacterial population during the different steps of silage processing (Dunière et al., 2013)

47

c. BUTYRIC ACID PRODUCTION FROM LACTATE

Several Clostridia (Clostridium butyricum, C. tyrobutyricum, C. paraputrificum) originating from soil contaminating silage are able to consume lactate to produce butyrate, CO2 and hydrogen gas while recovering energy (Equation 10) (McDonald, 1982; McDonald et al., 1973; Oude Elferink et al., 1999). Other microorganisms such as Megasphaera elsdenii, found in the rumen and swine feces (Hino and Kuroda, 1993; Tsukahara et al., 2002) perform the same reaction (Prabhu et al., 2012). Bacteria isolated from human feces, Eubacterium halii and Anaerostipes caccae were able to produce butyrate from lactate and acetate (Duncan et al., 2004) (Equation 11 and Figure 19).

Equation 10. Butyrate formation from lactate

Equation 11. Butyrate formation from lactate and acetate

Figure 19: Conversion of lactate and acetate to butyrate hydrogen and CO2 (Duncan, 2004)

d. OTHER ORGANIC ACIDS PRODUCED FROM LACTATE

Lactate can be converted to propionate via the acrylate pathway (Equation 12 and Figure 20) by Clostridium propionicum (Cardon and Barker, 1946; Leaver et al., 1955) and Megasphaera elsdenii (Counotte et al., 1981; Ladd and Walker, 1965). This last bacterium can also convert lactate back to pyruvate to produce acetate, CO2 and H2 (Grause et al., 2012; Prabhu et al., 2012) (Equation 13).

Valerate and hydrogen can be produced from lactate via the condensation of acetate and propionate themselves derived from lactate fermentation. Megasphaera elsdenii is capable of this reaction (Counotte et al., 1981; Ladd, 1959), which was also noticed in human faecal microflora enriched in

48 lactate (Bourriaud et al., 2005). Finally, Clostridium kluyveri can produce valerate from ethanol and propionate (Stadtman et al., 1949)

Equation 12. Consumption of lactate via the acrylate pathway

Equation 13. Conversion of lactate into acetate, CO2 and biohydrogen

Figure 20: Acrylate pathway for lactate fermentation to propionate by Clostridium propionicum (http://microbiochem.weebly.com/propionate.html)

e. FERMENTATION OF SWINE MANURE

Fermentation of swine manure and other agricultural waste streams have been studied in the context of hydrogen production (Chatellard et al., 2017), but using swine manure mostly as a medium where an hydrogen producing inoculum is added (Wu et al., 2013; 2010; 2009; 2017), with additional dilution of the manure making its bacterial content very low (Zhu et al., 2009). Tenca et al. (2011) however used swine manure at 8 g-VS/L and added fruit and vegetable waste, as well as an inoculum for hydrogen production. The organic co-substrate represented between 15 and 55% of the total weight, an amount that would be suitable for biological acidification of pig slurry and P recovery. Operated in semi continuous mode with 1 to 3 days HRT, the pH obtained was 4.3-5.6, values perfectly suitable for P dissolution in pig slurry. The details of the organic acids produced were not given unfortunately.

49

f. LACTIC ACID FERMENTATION FOLLOWED BY METHANE PRODUCTION

Bo et al. (2007) found that fermentation of kitchen waste led to a conversion of 50%-COD to lactate. The acidified mixture was sent to anaerobic digestion and compared with glucose fed anaerobic digestion. The reactor receiving fermented kitchen waste had a lower methane production due to the accumulation of , a known product of secondary fermentation of lactate. These inhibitory effects of lactate on anaerobic digestion were not observed by Daumer (unpublished data).

This review of acidogenesis provides useful hindsight regarding biological acidification of pig slurry to dissolve inorganic solid phosphorus. The low pH obtained by Tenca et al. (2011), perfectly sufficient for high P dissolution, indicates that the high alkalinity of swine slurry can be overcome at co-substrate concentration similar to what is commonly used in anaerobic digestion processes. The capacity of LAB to outcompete other bacteria for easily accessible carbohydrates suggests the need to use sugary co-substrates for an optimal acidification process.

50

V. CONCLUSION

The double problematic associated with phosphorus has been assessed. Depletion of this fossil resource is threatening global food security and runoffs from agricultural fields are severely polluting freshwater ecosystems. The role of agriculture in the inefficient use of P fertilizer and animal manure has been documented as well as its consequences in term of eutrophication. A major cause of inefficient use of phosphorus resources is the specialization of agriculture toward intensive livestock breeding in certain regions and high yield crops in others. As a result the latter relies exclusively on imported chemical P fertilizer since secondary sources of P such as animal manure are too far away to be transported there. Conversely, the regions specialized in intensive livestock breeding do not possess sufficient crop fields to spread their manure, leading to over-application, P build-up in soils, leaching and runoffs, polluting local water ecosystems. P-recovery processes have been developed to produce struvite, a slow release fertilizer with similar performances compared to conventional fertilizers. These processes have been implemented in a few large municipal wastewater treatment plants but most of them only recover already dissolved-P, leading to limited P-recovery rates. Scientific research for P- recovery from animal manure similarly focuses on dissolved-P or involves costly, high technology and chemical intensive pre-treatments. In this PhD, an innovative process has been developed to recover P as struvite from swine manure, using acidogenesis as a biological process to acidify the manure and dissolve a large proportion of total phosphorus. Because local organic waste is hydrolyzed and organic acids are produced during acidogenesis, the process could be implemented in parallel of, or prior to, anaerobic digestion (Figure 21). The solid fraction obtained after digestion would have a dramatically decreased P content, allowing it to be used as organic amendment to local arable soils. Conversely, struvite could be sold as P fertilizer and be exported at limited transport costs to regions specialized in crop production, thanks to the high P content of struvite compared to raw manure.

Figure 21. Envisioned implementation of struvite recovery within an anaerobic digestion process: parallel biological acidification. Implementation of biological acidification prior to digestion is also possible

51

REFERENCES

Abdel-Rahman, M.A., Tashiro, Y., Sonomoto, K., 2013. Recent advances in lactic acid production by microbial fermentation processes. Biotechnology Advances, 31(6), 877-902. Abma, W.R., Driessen, W., Haarhuis, R., van Loosdrecht, M.C.M., 2010. Upgrading of sewage treatment plant by sustainable and cost-effective separate treatment of industrial wastewater. Water Sci. Technol., 61(7), 1715-1722. Achat, D.L., Daumer, M.-L., Sperandio, M., Santellani, A.-C., Morel, C., 2014a. Solubility and mobility of phosphorus recycled from dairy effluents and pig manures in incubated soils with different characteristics. Nutrient cycling in agroecosystems, 99(1-3), 1-15. Achat, D.L., Sperandio, M., Daumer, M.-L., Santellani, A.-C., Prud'Homme, L., Akhtar, M., Morel, C., 2014b. Plant-availability of phosphorus recycled from pig manures and dairy effluents as assessed by isotopic labeling techniques. Geoderma, 232, 24-33. Ackerman, J.N., Zvomuya, F., Cicek, N., Flaten, D., 2013. Evaluation of manure-derived struvite as a phosphorus source for canola. Canadian Journal of Plant Science, 93(3), 419-424. Adam, C., Peplinski, B., Michaelis, M., Kley, G., Simon, F.G., 2009. Thermochemical treatment of sewage sludge ashes for phosphorus recovery. Waste Management, 29(3), 1122-1128. Adams, M.R., Hall, C.J., 1988. Growth inhibition of food-borne pathogens by lactic and acetic acids and their mixtures. International Journal of Food Science & Technology, 23(3), 287-292. Adnan, A., Mavinic, D.S., Koch, F.A., 2003. Pilot-scale study of phosphorus recovery through struvite crystallization examining the process feasibility. Journal of environmental engineering and science, 2(5), 315-324. Alam, K.Y., Clark, D.P., 1989. Anaerobic fermentation balance of escherichia-coli as observed by invivo nuclear magnetic-resonance spectroscopy. Journal of Bacteriology, 171(11), 6213-6217. Anderson, D.M., Hoagland, P., Kaoru, Y., White, A.W. 2000. Estimated annual economic impacts from harmful algal blooms (HABs) in the United States. National Oceanic And Atmospheric Administration Norman Ok National Severe Storms Lab. Antakyali, D., Meyer, C., Preyl, V., Maier, W., Steinmetz, H., 2013. Large-scale application of nutrient recovery from digested sludge as struvite. Water Practice and Technology, 8(2). Antonini, S., Arias, M.A., Eichert, T., Clemens, J., 2012. Greenhouse evaluation and environmental impact assessment of different urine-derived struvite fertilizers as phosphorus sources for plants. Chemosphere, 89(10), 1202-1210.Alexandratos, N., Bruinsma, J. 2012. World agriculture towards 2030/2050: the 2012 revision. ESA Working paper Rome, FAO. Appels, L., Degrève, J., Van der Bruggen, B., Van Impe, J., Dewil, R., 2010. Influence of low temperature thermal pre-treatment on sludge solubilisation, heavy metal release and anaerobic digestion. Bioresource Technology, 101(15), 5743-5748. Ashley, K., Cordell, D., Mavinic, D., 2011. A brief history of phosphorus: From the philosopher’s stone to nutrient recovery and reuse. Chemosphere, 84(6), 737-746. Azuara, M., Kersten, S.R.A., Kootstra, A.M.J., 2013. Recycling phosphorus by fast pyrolysis of pig manure: Concentration and extraction of phosphorus combined with formation of value-added pyrolysis products. Biomass and Bioenergy, 49(Supplement C), 171-180. Barak, P., Stafford, A., 2006. Struvite: A recovered and recycled phosphorus fertilizer. Barnett, G.M., 1994. Phosphorus forms in animal manure. Bioresour. Technol., 49(2), 139-147. Baroutian, S., Gapes, D.J., Sarmah, A.K., Farid, M.M., Young, B.R., 2016. Formation and degradation of valuable intermediate products during wet oxidation of municipal sludge. Bioresource Technology, 205(Supplement C), 280-285.

52

Bateman, A., van der Horst, D., Boardman, D., Kansal, A., Carliell-Marquet, C., 2011. Closing the phosphorus loop in England: The spatio-temporal balance of phosphorus capture from manure versus crop demand for fertiliser. Resources, Conservation and Recycling, 55(12), 1146-1153. Beck, T., 1972. The quantitative and qualitative composition of the lactic acid population of silage. Landwirtsch. Forsch, 27, 55-63. Bennett, E.M., Carpenter, S.R., Caraco, N.F., 2001. Human Impact on Erodable Phosphorus and Eutrophication: A Global Perspective: Increasing accumulation of phosphorus in soil threatens rivers, lakes, and coastal oceans with eutrophication. AIBS Bulletin, 51(3), 227-234. Bernardi, M., Cretenot, D., Deleris, S., Descorme, C., Chauzy, J., Besson, M., 2010. Performances of soluble metallic salts in the catalytic wet air oxidation of sewage sludge. Catalysis Today, 157(1), 420- 424. Biswas, B.K., Inoue, K., Harada, H., Ohto, K., Kawakita, H., 2009. Leaching of phosphorus from incinerated sewage sludge ash by means of acid extraction followed by adsorption on orange waste gel. Journal of Environmental Sciences, 21(12), 1753-1760. Bo, Z., Wei-min, C.A.I., Pin-jing, H.E., 2007. Influence of lactic acid on the two-phase anaerobic digestion of kitchen wastes. Journal of Environmental Sciences, 19(2), 244-249. Bonk, F., Bastidas-Oyanedel, J.-R., Yousef, A.F., Schmidt, J.E., 2017. Exploring the selective lactic acid production from food waste in uncontrolled pH mixed culture fermentations using different reactor configurations. Bioresource Technology, 238(Supplement C), 416-424. Bourriaud, C., Robins, R.J., Martin, L., Kozlowski, F., Tenailleau, E., Cherbut, C., Michel, C., 2005. Lactate is mainly fermented to butyrate by human intestinal microfloras but inter-individual variation is evident. Journal of Applied Microbiology, 99(1), 201-212. Brett, S., 1997. Phosphorus removal and recovery technologies. Selper. Britton, A., 2009. P-Recovery in North America–Ostaras Pearl Process. Conference Baltic. pp. 28-30. Britton, A., Wirtel, S., Prasad, R., Turner, N., 2015. Industry Update for WASSTRIP® impacts on nutrient flows, digestion and dewatering of EBPR solids. Proceedings of the Water Environment Federation, 2015(3), 1-10. Capdevielle, A., Sykorova, E., Beline, F., Daumer, M.L., 2016. Effects of organic matter on crystallization of struvite in biologically treated swine wastewater. Environmental Technology, 37(7), 880-892. Cardon, B.P., Barker, H.A., 1946. Two New Amino-Acid-Fermenting Bacteria, Clostridium propionicum and Diplococcus glycinophilus. J. Bacteriol., 52(6), 629-634. Carpenter, S.R., 2005. Eutrophication of aquatic ecosystems: Bistability and soil phosphorus. Proceedings of the National Academy of Sciences of the United States of America, 102(29), 10002- 10005. Cerrillo, M., Palatsi, J., Comas, J., Vicens, J., Bonmatí, A., 2015. Struvite precipitation as a technology to be integrated in a manure anaerobic digestion treatment plant–removal efficiency, crystal characterization and agricultural assessment. Journal of Chemical Technology and Biotechnology, 90(6), 1135-1143. Chatellard, L., Marone, A., Carrère, H., Trably, E., 2017. Trends and Challenges in Biohydrogen Production from Agricultural Waste. in: Biohydrogen Production: Sustainability of Current Technology and Future Perspective, (Eds.) A. Singh, D. Rathore, Springer India. New Delhi, pp. 69- 95. Chen, Y., Cheng, J.J., Creamer, K.S., 2008. Inhibition of anaerobic digestion process: A review. Bioresource Technology, 99(10), 4044-4064.

53

Christensen, J.E., Dudley, E.G., Pederson, J.A., Steele, J.L., 1999. Peptidases and amino acid catabolism in lactic acid bacteria. Antonie Van Leeuwenhoek International Journal of General and Molecular Microbiology, 76(1-4), 217-246. Christensen, M.L., Hjorth, M., Keiding, K., 2009. Characterization of pig slurry with reference to flocculation and separation. Water Res., 43(3), 773-783. Chung, J., Lee, M., Ahn, J., Bae, W., Lee, Y.-W., Shim, H., 2009. Effects of operational conditions on sludge degradation and organic acids formation in low-critical wet air oxidation. Journal of Hazardous Materials, 162(1), 10-16. Cleveland, C.C., Houlton, B.Z., Smith, W.K., Marklein, A.R., Reed, S.C., Parton, W., Del Grosso, S.J., Running, S.W., 2013. Patterns of new versus recycled primary production in the terrestrial biosphere. Proceedings of the National Academy of Sciences, 110(31), 12733-12737. Cohen, Y., 2009. Phosphorus dissolution from ash of incinerated sewage sludge and animal carcasses using sulphuric acid. Environmental Technology, 30(11), 1215-1226. Colmar, A. 2017. Synthèse du projet TRANS-P. https://www6.rennes.inra.fr/umrsas/Media/theme-de- recherche/Synthese-8-pages-du-Projet-TRANS-P. Converti, A., Perego, P., 2002. Use of carbon and energy balances in the study of the anaerobic of Enterobacter aerogenes at variable starting glucose concentrations. Applied microbiology and biotechnology, 59(2-3), 303-309. Cooper, J., Lombardi, R., Boardman, D., Carliell-Marquet, C., 2011. The future distribution and production of global phosphate rock reserves. Resources, Conservation and Recycling, 57, 78-86. Cordell, D., Drangert, J.O., White, S., 2009. The story of phosphorus: Global food security and food for thought. Glob. Environ. Chang., 19(2), 292-305. Cordell, D., Neset, T.S.S., 2014. Phosphorus vulnerability: A qualitative framework for assessing the vulnerability of national and regional food systems to the multi-dimensional stressors of phosphorus scarcity. Global Environmental Change, 24(Supplement C), 108-122. Counotte, G., Prins, R., Janssen, R., 1981. Role of Megasphaera elsdenii in the fermentation of DL-[2- 13C] lactate in the rumen of dairy cattle. Applied and Environmental Microbiology, 42(4), 649-655. Csatho, P., Radimszky, L., 2009. Two worlds within EU27: Sharp contrasts in organic and mineral nitrogen–phosphorus use, nitrogen–phosphorus balances, and soil phosphorus status: Widening and deepening gap between Western and Central Europe. Communications in Soil Science and Plant Analysis, 40(1-6), 999-1019. Cullen, N., Baur, R., Schauer, P., 2013. Three years of operation of North America's first nutrient recovery facility. Water Science and Technology, 68(4), 763-768. Dai, L., Tan, F., Wu, B., He, M., Wang, W., Tang, X., Hu, Q., Zhang, M., 2015. Immobilization of phosphorus in cow manure during hydrothermal carbonization. Journal of Environmental Management, 157(Supplement C), 49-53. Dai, L., Yang, B., Li, H., Tan, F., Zhu, N., Zhu, Q., He, M., Ran, Y., Hu, G., 2017. A synergistic combination of nutrient reclamation from manure and resultant hydrochar upgradation by acid- supported hydrothermal carbonization. Bioresource Technology, 243(Supplement C), 860-866. Daumer, M.-L., Béline, F., Spérandio, M., Morel, C., 2008. Relevance of a perchloric acid extraction scheme to determine mineral and organic phosphorus in swine slurry. Bioresour. Technol., 99(5), 1319-1324. Daumer, M.L., Picard, S., Saint-Cast, P., Dabert, P., 2010. Technical and economical assessment of formic acid to recycle phosphorus from pig slurry by a combined acidification–precipitation process. J. Hazard. Mater., 180(1–3), 361-365.

54

Decrey, L., Udert, K.M., Tilley, E., Pecson, B.M., Kohn, T., 2011. Fate of the pathogen indicators phage ΦX174 and Ascaris suum eggs during the production of struvite fertilizer from source-separated urine. Water Research, 45(16), 4960-4972. Dersin, D., 1997. What life was like when Rome ruled the world: The Roman Empire 100 BC-AD 200. Time-Life Books. Desmidt, E., Ghyselbrecht, K., Zhang, Y., Pinoy, L., Van der Bruggen, B., Verstraete, W., Rabaey, K., Meesschaert, B., 2015. Global Phosphorus Scarcity and Full-Scale P-Recovery Techniques: A Review. Critical Reviews in Environ. Sci. Technol., 45(4), 336-384. Dodds, W.K., Bouska, W.W., Eitzmann, J.L., Pilger, T.J., Pitts, K.L., Riley, A.J., Schloesser, J.T., Thornbrugh, D.J., 2009. Eutrophication of U.S. Freshwaters: Analysis of Potential Economic Damages. Environmental Science & Technology, 43(1), 12-19. Donatello, S., Cheeseman, C.R., 2013. Recycling and recovery routes for incinerated sewage sludge ash (ISSA): A review. Waste Management, 33(11), 2328-2340. Donatello, S., Tong, D., Cheeseman, C.R., 2010. Production of technical grade phosphoric acid from incinerator sewage sludge ash (ISSA). Waste Management, 30(8), 1634-1642. Doyle, J.D., Parsons, S.A., 2002. Struvite formation, control and recovery. Water Research, 36(16), 3925-3940. Duncan, S.H., Louis, P., Flint, H.J., 2004. Lactate-utilizing bacteria, isolated from human feces, that produce butyrate as a major fermentation product. Applied and environmental microbiology, 70(10), 5810-5817. Dunière, L., Sindou, J., Chaucheyras-Durand, F., Chevallier, I., Thévenot-Sergentet, D., 2013. Silage processing and strategies to prevent persistence of undesirable microorganisms. Animal Feed Science and Technology, 182(1–4), 1-15. Dupas, R., Delmas, M., Dorioz, J.-M., Garnier, J., Moatar, F., Gascuel-Odoux, C., 2015. Assessing the impact of agricultural pressures on N and P loads and eutrophication risk. Ecological Indicators, 48(Supplement C), 396-407. Egle, L., Rechberger, H., Krampe, J., Zessner, M., 2016. Phosphorus recovery from municipal wastewater: An integrated comparative technological, environmental and economic assessment of P recovery technologies. Science of The Total Environment, 571(Supplement C), 522-542. Egle, L., Rechberger, H., Zessner, M., 2015. Overview and description of technologies for recovering phosphorus from municipal wastewater. Resources, Conservation and Recycling, 105(Part B), 325- 346. Ekpo, U., Ross, A.B., Camargo-Valero, M.A., Fletcher, L.A., 2016. Influence of pH on hydrothermal treatment of swine manure: Impact on extraction of nitrogen and phosphorus in process water. Bioresource Technology, 214(Supplement C), 637-644. El Rafie, S., Hawash, S., Shalaby, M., 2013. Evaluation of struvite precipitated from chemical fertilizer industrial effluents. Adv. Appl. Sci. Res, 4(1), 113-123. Elser, J., Bennett, E., 2011. Phosphorus cycle: a broken biogeochemical cycle. Nature, 478(7367), 29- 31. Elser, J.J., 2012. Phosphorus: a limiting nutrient for humanity? Current Opinion in Biotechnology, 23(6), 833-838. Emsley, J., 2000. 13th element. John Wiley & Sons, Inc. Falkowski, P., Scholes, R.J., Boyle, E., Canadell, J., Canfield, D., Elser, J., Gruber, N., Hibbard, K., Högberg, P., Linder, S., Mackenzie, F.T., Moore III, B., Pedersen, T., Rosenthal, Y., Seitzinger, S., Smetacek, V., Steffen, W., 2000. The Global Carbon Cycle: A Test of Our Knowledge of Earth as a System. Science, 290(5490), 291-296.

55

FAO. 2013. Statistical yearbook World food and agriculture. Food and Agriculture Organization of the United Nations, Rome, 2013. FAO. 2017. World fertilizer trends and outlook to 2020, summary report. Food and Agriculture Organization of the United Nations, Rome. Fattah, K.P., 2010. Development of control strategies for the operation of a struvite crystallization process, University of British Columbia. Fernández, M., Zúñiga, M., 2006. Amino Acid Catabolic Pathways of Lactic Acid Bacteria. Critical Reviews in Microbiology, 32(3), 155-183. Forrest, A., Fattah, K., Mavinic, D., Koch, F., 2008. Optimizing struvite production for phosphate recovery in WWTP. Journal of Environmental Engineering, 134(5), 395-402. Foy, R., Withers, P., 2002. The contribution of agricultural phosphorus to eutrophication. Nawozy i Nawożenie, 4(3 [12]). Franz, M., 2008. Phosphate fertilizer from sewage sludge ash (SSA). Waste Management, 28(10), 1809-1818. Furness, D.T., Hoggett, L.A., Judd, S.J., 2000. Thermochemical Treatment of Sewage Sludge. Water and Environment Journal, 14(1), 57-65. Ganrot, Z., Dave, G., Nilsson, E., Li, B., 2007. Plant availability of nutrients recovered as solids from human urine tested in climate chamber on Triticum aestivum L. Bioresource Technology, 98(16), 3122-3129. Garvie, E.I., 1980. Bacterial lactate dehydrogenases. Microbiological Reviews, 44(1), 106-139. Gell, K., de Ruijter, F., Kuntke, P., De Graaff, M., Smit, A., 2011. Safety and effectiveness of struvite from black water and urine as a phosphorus fertilizer. Journal of Agricultural Science, 3(3), 67. Ghosh, G., Mohan, K., Sarkar, A., 1996. Characterization of soil-fertilizer P reaction products and their evaluation as sources of P for gram (Cicer arietinum L.). Nutrient Cycling in Agroecosystems, 46(1), 71-79. Gilbert, N., 2009. Environment: the disappearing nutrient. Nature News, 461(7265), 716-718. GisSol. 2011. L'état des sols en France. Groupement d'intérêt scientifique sur les sols, 188p. Gleick, P.H., 1998. Water in crisis: paths to sustainable water use. Ecological applications, 8(3), 571- 579. Goto, M., Nada, T., Ogata, A., Kodama, A., Hirose, T., 1998. Supercritical water oxidation for the destruction of municipal excess sludge and alcohol distillery wastewater of molasses. The Journal of Supercritical Fluids, 13(1), 277-282. Graja, S., Chauzy, J., Fernandes, P., Patria, L., Cretenot, D., 2005. Reduction of sludge production from WWTP using thermal pretreatment and enhanced anaerobic methanisation. Water Science and Technology, 52(1-2), 267-273. Grause, G., Igarashi, M., Kameda, T., Yoshioka, T., 2012. Lactic acid as a substrate for fermentative hydrogen production. International Journal of Hydrogen Energy, 37(22), 16967-16973. Grazia, L., Suzzi, G., 1984. A survey of lactic acid bacteria in Italian silage. Journal of Applied Bacteriology, 56(3), 373-379. Grizzetti, B., Bouraoui, F., Aloe, A., 2012. Changes of nitrogen and phosphorus loads to European seas. Global Change , 18(2), 769-782. Gull, M., Hafeez, F.Y., Saleem, M., Malik, K.A., 2004. Phosphorus uptake and growth promotion of chickpea by co-inoculation of mineral phosphate solubilising bacteria and a mixed rhizobial culture. Australian Journal of Experimental Agriculture, 44(6), 623-628.

56

Hallenbeck, P.C., 2009. Fermentative hydrogen production: Principles, progress, and prognosis. International Journal of Hydrogen Energy, 34(17), 7379-7389. Hallenbeck, P.C., 2005. Fundamentals of the fermentative production of hydrogen. Water Science and Technology, 52(1-2), 21-29. Hao, X., Wang, C., van Loosdrecht, M.C., Hu, Y., 2013. Looking beyond struvite for P-recovery, ACS Publications. Heckenmüller, M., Narita, D., Klepper, G. 2014. Global availability of phosphorus and its implications for global food supply: an economic overview. Kiel Working Paper. Heckrath, G., Brookes, P.C., Poulton, P.R., Goulding, K.W.T., 1995. Phosphorus Leaching from Soils Containing Different Phosphorus Concentrations in the Broadbalk Experiment. Journal of Environmental Quality, 24(5), 904-910. Heffer, P., Prud’homme, M., 2016. Fertilizer Outlook 2016–2020. International Fertilizer Association (IFA), Paris, France. Heifer, P., Prud'homme, M., 2014. Fertilizer outlook 2014-2018. Sydney: The 82nd IFA Annual Conference. Heilmann, S.M., Molde, J.S., Timler, J.G., Wood, B.M., Mikula, A.L., Vozhdayev, G.V., Colosky, E.C., Spokas, K.A., Valentas, K.J., 2014. Phosphorus Reclamation through Hydrothermal Carbonization of Animal Manures. Environmental Science & Technology, 48(17), 10323-10329. Hii, K., Baroutian, S., Parthasarathy, R., Gapes, D.J., Eshtiaghi, N., 2014. A review of wet air oxidation and Thermal Hydrolysis technologies in sludge treatment. Bioresource Technology, 155(Supplement C), 289-299. Hino, T., Kuroda, S., 1993. Presence of lactate dehydrogenase and lactate racemase in Megasphaera elsdenii grown on glucose or lactate. Applied and environmental microbiology, 59(1), 255-259. Hjorth, M., Fernandez, M., Jayaram, S., Sørensen, J.A., Adamsen, A.P.S., 2015. Bio-acidification of manure–By supplying manure with 2-3% or cellulose. RAMIRAN 2015–16th International Conference. TuTech Innovation GmbH, Hamburg. pp. 96-96. Heinzmann, B., Engel, G., 2006. Induced Magnesium Ammonia Phosphate Precipitation to Prevent Incrustations and Measures for Phosphorus Recovery. Water Practice and Technology, 1(3). Holzapfel, W., Wood, B.J., 2012. The genera of lactic acid bacteria. Springer Science & Business Media. Holzer, M., Mayrhuber, E., Danner, H., Braun, R., 2003. The role of Lactobacillus buchneri in forage preservation. Trends in biotechnology, 21(6), 282-287. Hotaling, J., 2006. Struvite recovery from digested dairy manure and regional manure anaerobic digestion study New York State Energy Research and Development Authority. Hutchinson, G., 1950. Survey of contemporary knowledge of biogeochemistry 3. The biogeochemistry of vertebrate excretion: Bulletin of the American Museum of Natural History, 96, 553. Ignatowicz, K., 2017. The impact of sewage sludge treatment on the content of selected heavy metals and their fractions. Environmental Research, 156(Supplement C), 19-22. Jelic, A., Gros, M., Ginebreda, A., Cespedes-Sánchez, R., Ventura, F., Petrovic, M., Barcelo, D., 2011. Occurrence, partition and removal of pharmaceuticals in sewage water and sludge during wastewater treatment. Water Research, 45(3), 1165-1176. Jin, Y., Hu, Z., Wen, Z., 2009. Enhancing anaerobic digestibility and phosphorus recovery of dairy manure through microwave-based thermochemical pretreatment. Water Research, 43(14), 3493-3502. Jungermann, K., Thauer, R.K., Leimenstoll, G., Decker, K., 1973. Function of reduced pyridine nucleotide-ferredoxin in saccharolytic Clostridia. Biochimica et biophysica acta, 305(2), 268-80.

57

Kahiluoto, H., Kuisma, M., Ketoja, E., Salo, T., Heikkinen, J., 2015. Phosphorus in Manure and Sewage Sludge More Recyclable than in Soluble Inorganic Fertilizer. Environmental Science & Technology, 49(4), 2115-2122. Karakashev, D., Schmidt, J.E., Angelidaki, I., 2008. Innovative process scheme for removal of organic matter, phosphorus and nitrogen from pig manure. Water Research, 42(15), 4083-4090. Kataki, S., West, H., Clarke, M., Baruah, D.C., 2016. Phosphorus recovery as struvite from farm, municipal and industrial waste: Feedstock suitability, methods and pre-treatments. Waste Management, 49(Supplement C), 437-454. Kataki, S., West, H., Clarke, M., Baruah, D.C., 2016. Phosphorus recovery as struvite: Recent concerns for use of seed, alternative Mg source, nitrogen conservation and fertilizer potential. Resources, Conservation and Recycling, 107(Supplement C), 142-156. Kellogg, R.L., Lander, C.H., Moffitt, D.C., Gollehon, N., 2000. Manure Nutrients Relative to The Capacity Of Cropland And Pastureland To Assimilate Nutrients: Spatial and Temporal Trends for the United States. Proceedings of the Water Environment Federation, 2000(16), 18-157. Kepp, U., Machenbach, I., Weisz, N., Solheim, O.E., 2000. Enhanced stabilisation of sewage sludge through thermal hydrolysis - three years of experience with full scale plant. Water Science and Technology, 42(9), 89-96. Kidd, P.S., Domínguez-Rodríguez, M.J., Díez, J., Monterroso, C., 2007. Bioavailability and plant accumulation of heavy metals and phosphorus in agricultural soils amended by long-term application of sewage sludge. Chemosphere, 66(8), 1458-1467. Kirchmann, H., Nyamangara, J., Cohen, Y., 2005. Recycling municipal wastes in the future: from organic to inorganic forms? Soil Use and Management, 21, 152-159. Koppelaar, R.H.E.M., Weikard, H.P., 2013. Assessing phosphate rock depletion and phosphorus recycling options. Global Environmental Change, 23(6), 1454-1466. Krogstad, T., Sogn, T.A., Asdal, Å., Sæbø, A., 2005. Influence of chemically and biologically stabilized sewage sludge on plant-available phosphorous in soil. Ecological Engineering, 25(1), 51-60. Ladd, J., 1959. The fermentation of lactic acid by a gram-negative coccus. Biochemical Journal, 71(1), 16. Ladd, J.N., Walker, D.J., 1965. FERMENTATION OF LACTIC ACID BY THE RUMEN , PEPTOSTREPTOCOCCUS ELSDENII. Annals of the New York Academy of Sciences, 119(3), 1038-1047. Lake, D., Kirk, P., Lester, J., 1984. Fractionation, characterization, and speciation of heavy metals in sewage sludge and sludge-amended soils: a review. Journal of environmental quality, 13(2), 175-183. Lambers, H., Raven, J.A., Shaver, G.R., Smith, S.E., 2008. Plant nutrient-acquisition strategies change with soil age. Trends in ecology & evolution, 23(2), 95-103. Lawton, L.A., Codd, G., 1991. Cyanobacterial (blue‐green algal) toxins and their significance in UK and European waters. Water and Environment Journal, 5(4), 460-465. Lay, J.J., 2000. Modeling and optimization of anaerobic digested sludge converting starch to hydrogen. Biotechnology and Bioengineering, 68(3), 269-278. Le Corre, K.S., Valsami-Jones, E., Hobbs, P., Parsons, S.A., 2009. Phosphorus Recovery from Wastewater by Struvite Crystallization: A Review. Critical Reviews in Environmental Science and Technology, 39(6), 433-477. Leaver, F.W., Wood, H.G., Stjernholm, R., 1955. The fermentation of three carbon substrates by Clostridium propionicum and Propionibacterium. Journal of bacteriology, 70(5), 521-30. Levlin, E., 2007. Phosphorus recovery with acid and base from inorganic sewage sludge residues. Water Practice and Technology, 2(1).

58

Levlin, E., Löwén, M., Stark, K., 2004. Phosphorus recovery from sludge incineration ash and Supercritical Water Oxidation residues with use of acids and bases. Proceedings of a Polish-Swedish seminar. pp. 25-10.28. Li, X.Z., Zhao, Q.L., 2003. Recovery of ammonium-nitrogen from landfill leachate as a multi-nutrient fertilizer. Ecological Engineering, 20(2), 171-181. Liberti, L., Petruzzelli, D., De Florio, L., 2001. REM NUT Ion Exchange Plus Struvite Precipitation Process. Environmental Technology, 22(11), 1313-1324. Liu, H., Wang, J., Liu, X., Fu, B., Chen, J., Yu, H.-Q., 2012. Acidogenic fermentation of proteinaceous sewage sludge: Effect of pH. Water Research, 46(3), 799-807. Liu, Y., Villalba, G., Ayres, R.U., Schroder, H., 2008. Global Phosphorus Flows and Environmental Impacts from a Consumption Perspective. Journal of Industrial Ecology, 12(2), 229-247. Liu, Y., Kwag, J.-H., Kim, J.-H., Ra, C., 2011. Recovery of nitrogen and phosphorus by struvite crystallization from swine wastewater. Desalination, 277(1), 364-369. Liu, Y., Rahman, M., Kwag, J.-H., Kim, J.-H., Ra, C.-S., 2011. Eco-friendly production of maize using struvite recovered from swine wastewater as a sustainable fertilizer source. Asian-Australasian Journal of Animal Sciences, 24(12), 1699-1705. MacDonald, G.K., Bennett, E.M., Potter, P.A., Ramankutty, N., 2011. Agronomic phosphorus imbalances across the world's croplands. Proceedings of the National Academy of Sciences, 108(7), 3086-3091. Macy, J.M., Ljungdahl, L.G., Gottschalk, G., 1978. Pathway of succinate and propionate formation in Bacteroides fragilis. Journal of Bacteriology, 134(1), 84-91. Maitland, P.S., 1993. Restoration of aquatic ecosystems: Science, technology, and public policy: National Academy of Sciences, National Academy Press, Washington, DC. 1992. 552 pp. ISBN 0- 309-04534-7. Elsevier. Mårald, E., 2002. Everything circulates: agricultural chemistry and recycling theories in the second half of the nineteenth century. Environment and History, 65-84. Mårald, E., 1998. I mötet mellan jordbruk och kemi: agrikulturkemins framväxt på Lantbruksakademiens experimentalfält 1850-1907. Univ Umeå. Martí, N., Barat, R., Seco, A., Pastor, L., Bouzas, A., 2017. Sludge management modeling to enhance P-recovery as struvite in wastewater treatment plants. Journal of Environmental Management, 196(Supplement C), 340-346. Marti, N., Bouzas, A., Seco, A., Ferrer, J., 2008. Struvite precipitation assessment in anaerobic digestion processes. Chemical Engineering Journal, 141(1), 67-74. McDonald, L.C., Fleming, H.P., Hassan, H.M., 1990. Acid Tolerance Of Leuconostoc-Mesenteroides And Lactobacillus-Plantarum. Applied and Environmental Microbiology, 56(7), 2120-2124. McDonald, P., 1982. Silage fermentation. Trends Biochem. Sci., 7(5), 164-166. McDonald, P., Henderson, A.R., Ralton, I., 1973. Energy changes during ensilage. Journal of the Science of Food and Agriculture, 24(7), 827-834. McNeill, J.R., Winiwarter, V., 2004. Breaking the Sod: Humankind, History, and Soil. Science, 304(5677), 1627-1629. Melia, P.M., Cundy, A.B., Sohi, S.P., Hooda, P.S., Busquets, R., 2017. Trends in the recovery of phosphorus in bioavailable forms from wastewater. Chemosphere, 186(Supplement C), 381-395. Moerman, W., Carballa, M., Vandekerckhove, A., Derycke, D., Verstraete, W., 2009. Phosphate removal in agro-industry: Pilot- and full-scale operational considerations of struvite crystallization. Water Research, 43(7), 1887-1892.

59

Moletta, R., 1993. La digestion anaérobie: du plus petit au plus grand. Biofutur(119), 16-25. Montag, D., Pinnekamp, J., 2009. The PASH process for P recovery and overview of the German Funding Programme’Recycling management of plant nutrients, especially phosphorus, BALTIC 21 Phosphorus Recycling and Good Agricultural Management Practice. Joint Conference by Julius Kühn- Institut (JKI) and BAM Bundesanstalt für Materialforschung und–prüfung. Morgan-Sagastume, F., Pratt, S., Karlsson, A., Cirne, D., Lant, P., Werker, A., 2011. Production of volatile fatty acids by fermentation of waste activated sludge pre-treated in full-scale thermal hydrolysis plants. Bioresource Technology, 102(3), 3089-3097. Müller, J., Günther, L., Dockhorn, T., Dichtl, N., Phan, L., Urban, I., Weichgrebe, D., Rosenwinkel, K., Bayerle, N., 2007. Nutrient recycling from sewage sludge using the Seaborne process. Conference, Proceedings on Moving Forward Wastewater Biosolids Sustainability: Technical, Managerial, and Public Synergy. GMSC. pp. 629-633. Munch, E.V., Barr, K., 2001. Controlled struvite crystallisation for removing phosphorus from anaerobic digester sidestreams. Water Research, 35(1), 151-159. Nagase, M., Matsuo, T., 1982. Interactions between amino-acid-degrading bacteria and methanogenic bacteria in anaerobic digestion. Biotechnology and Bioengineering, 24(10), 2227-2239. Némery, J., Garnier, J., Morel, C., 2005. Phosphorus budget in the Marne Watershed (France): urban vs. diffuse sources, dissolved vs. particulate forms. Biogeochemistry, 72(1), 35-66. Ngoze, S., Riha, S., Lehmann, J., Verchot, L., Kinyangi, J., Mbugua, D., Pell, A., 2008. Nutrient constraints to tropical agroecosystem productivity in long-term degrading soils. Global Change Biology, 14(12), 2810-2822. Niewersch, C., 2013. Nanofiltration for Phosphorus Recycling from Sewage Sludge. Mainz. Nixon, S., 2004. The Artificial Nile The Aswan High Dam blocked and diverted nutrients and destroyed a Mediterranean fishery, but human activities may have revived it. American Scientist, 92(2), 158-165. Norli, M., 2006. The cambi thermal hydrolysis process (THP)-dramatically improving sludge digestion and dewatering-11 years experience of full scale project. Gdansk Conference, Gdansk, Poland. Ober, J.A. 2017. Mineral commodity summaries 2017. US Geological Survey. Ott, C., Rechberger, H., 2012. The European phosphorus balance. Resources, Conservation and Recycling, 60, 159-172. Oude Elferink, S.J., Driehuis, F., Gottschal, J.C., Spoelstra, S.F., 1999. Silage fermentation processes and their manipulation FAO electronic conference on tropical silage, Rome, Italy. Pan, S.-H., Lo, K.V., Liao, P.H., Schreier, H., 2006. Microwave pretreatment for enhancement of phosphorus release from dairy manure. Journal of Environmental Science and Health Part B, 41(4), 451-458. Papoutsakis, E.T., Meyer, C.L., 1985. Fermentation equations for propionic-acid bacteria and production of assorted oxychemicals from various . Biotechnology and Bioengineering, 27(1), 67-80. Pastor, L., Marti, N., Bouzas, A., Seco, A., 2008. Sewage sludge management for phosphorus recovery as struvite in EBPR wastewater treatment plants. Bioresource Technology, 99(11), 4817- 4824. Peplinski, B., Adam, C., Michaelis, M., Kley, G., Emmerling, F., Simon, F.-G., 2009. Reaction sequences in the thermochemical treatment of sewage sludge ashes revealed by X-ray powder diffraction—a contribution to the European project SUSAN. Z. Kristallogr. Suppl, 30, 459-464.

60

Perera, P.A., Han, Z.-Y., Chen, Y.-X., Wu, W.-X., 2007. Recovery of nitrogen and phosphorous as struvite from swine waste biogas digester effluent. Biomedical and environmental sciences, 20(5), 343-350. Perera, P.W.A., Wu, W.-X., Chen, Y.-X., Han, Z.-Y., 2009. Struvite Recovery from Swine Waste Biogas Digester Effluent through a Stainless Steel Device under Constant pH Conditions. Biomedical and Environmental Sciences, 22(3), 201-209. Pérez, R.C., Steingrobe, B., Romer, W., Claassen, N., 2009. Plant availability of Pfertilizers recycled from sewage sludge and meat-and-bone meal in field and pot experiments. International Conference on Nutrient Recovery From Wastewater Streams Vancouver, 2009. IWA publishing. pp. 215. Petzet, S., Cornel, P., 2011. Towards a complete recycling of phosphorus in wastewater treatment – options in Germany. Water Science and Technology, 64(1), 29-35. Pickworth, B., Adams, J., Panter, K., Solheim, O.E., 2006. Maximising biogas in anaerobic digestion by using engine waste heat for thermal hydrolysis pre-treatment of sludge. Water Science and Technology, 54(5), 101-108. Plaza, C., Sanz, R., Clemente, C., Fernández, J.M., González, R., Polo, A., Colmenarejo, M.F., 2007. Greenhouse evaluation of struvite and sludges from municipal wastewater treatment works as phosphorus sources for plants. Journal of agricultural and food chemistry, 55(20), 8206-8212. Potter, P., Ramankutty, N., Bennett, E.M., Donner, S.D., 2010. Characterizing the Spatial Patterns of Global Fertilizer Application and Manure Production. Earth Interactions, 14(2), 1-22. Prabhu, R., Altman, E., Eiteman, M.A., 2012. Lactate and acrylate metabolism by Megasphaera elsdenii under batch and steady-state conditions. Applied and environmental microbiology, 78(24), 8564-8570. Probst, M., Walde, J., Pümpel, T., Wagner, A.O., Insam, H., 2015. A closed loop for municipal organic solid waste by lactic acid fermentation. Bioresource Technology, 175(Supplement C), 142- 151. Prud'homme, M., 2010. challenges and opportunities for the fertilizer industry: fertilizers and agricultural production, International Fertilizer Industry Association. Qureshi, A., Lo, K.V., Liao, P.H., 2008. Microwave treatment and struvite recovery potential of dairy manure. Journal of Environmental Science and Health, Part B, 43(4), 350-357. Rahman, M.M., Liu, Y., Kwag, J.-H., Ra, C., 2011. Recovery of struvite from animal wastewater and its nutrient leaching loss in soil. J. Hazard. Mater., 186(2), 2026-2030. Ray, D.K., Mueller, N.D., West, P.C., Foley, J.A., 2013. Yield Trends Are Insufficient to Double Global Crop Production by 2050. PLOS ONE, 8(6), e66428. Redwood, M.D., Mikheenko, I.P., Sargent, F., Macaskie, L.E., 2008. Dissecting the roles of Escherichia coli in biohydrogen production. Fems Microbiology Letters, 278(1), 48-55. Reichardt, N., Duncan, S.H., Young, P., Belenguer, A., Leitch, C.M., Scott, K.P., Flint, H.J., Louis, P., 2014. Phylogenetic distribution of three pathways for propionate production within the human gut microbiota. The ISME journal, 8(6), 1323-1335. Reijnders, L., 2014. Phosphorus resources, their depletion and conservation, a review. Resources, Conservation and Recycling, 93(Supplement C), 32-49. Remy, D.-I.C., Ruhland, I.A., 2006. Ecological assessment of alternative sanitation concepts with Life Cycle Assessment. Technical University Berlin, Berlin, Germany, 55. Remy, M., Kruit, J., Hendrickx, T., Haarhuis, R., van Loosdrecht, M., 2013. Phospaq: Full scale experience with phosphorus recovery via controlled struvite precipitation. Proceedings of the Water Environment Federation, 2013(4), 311-317.

61

Ringeval, B., Nowak, B., Nesme, T., Delmas, M., Pellerin, S., 2014. Contribution of anthropogenic phosphorus to agricultural soil fertility and food production. Global Biogeochemical Cycles, 28(7), 743-756. Rossmann, R., Sawers, G., Bock, A., 1991. Mechanism of regulation of the formate-hydrogenlyase pathway by oxygen, nitrate, and ph - definition of the formate regulon. Molecular Microbiology, 5(11), 2807-2814. Russelle, M.P., Entz, M.H., Franzluebbers, A.J., 2007. Reconsidering Integrated Crop–Livestock Systems in North America. Agronomy Journal, 99(2), 325-334. Ryu, H.-D., Lee, S.-I., 2010. Application of struvite precipitation as a pretreatment in treating swine wastewater. Process , 45(4), 563-572. Sahlström, L., Aspan, A., Bagge, E., Tham, M.-L.D., Albihn, A., 2004. Bacterial pathogen incidences in sludge from Swedish sewage treatment plants. Water Research, 38(8), 1989-1994. Salminen, S., Von Wright, A., 2004. Lactic acid bacteria: microbiological and functional aspects. CRC Press. Sattari, S.Z., Bouwman, A.F., Giller, K.E., van Ittersum, M.K., 2012. Residual soil phosphorus as the missing piece in the global phosphorus crisis puzzle. Proceedings of the National Academy of Sciences, 109(16), 6348-6353. Sattari, S., Van Ittersum, M., Bouwman, A., Smit, A., Janssen, B., 2014. Crop yield response to soil fertility and N, P, K inputs in different environments: testing and improving the QUEFTS model. Field Crops Research, 157, 35-46. Schaum, C., Cornel, P., Jardin, N., 2007. Phosphorus recovery from sewage sludge ash–a wet chemical approach. Proceeding of the IWA Conference on Biosolids, Moving Forward Wastewater Biosolids Sustainability: Technical, Managerial, and Public Synergy. pp. 24-27. Schaum, C., Cornel, P., Jardin, N., 2005. Possibilities for Phosphorus Recovery from Sewage Sludge Ash, 9: 12 August 2005. Management of Residues Emanating from Water and Wastewater Treatment, Johannesburg, South Africa. Schievano, A., Tenca, A., Scaglia, B., Merlino, G., Rizzi, A., Daffonchio, D., Oberti, R., Adani, F., 2012. Two-stage vs single-stage thermophilic anaerobic digestion: comparison of energy production and biodegradation efficiencies. Environmental Science and Technology, 46(15), 8502-8510. Schindler, D.W., Hecky, R.E., Findlay, D.L., Stainton, M.P., Parker, B.R., Paterson, M.J., Beaty, K.G., Lyng, M., Kasian, S.E.M., 2008. Eutrophication of lakes cannot be controlled by reducing nitrogen input: Results of a 37-year whole-ecosystem experiment. Proceedings of the National Academy of Sciences, 105(32), 11254-11258. Schipanski, M.E., Bennett, E.M., 2012. The Influence of Agricultural Trade and Livestock Production on the Global Phosphorus Cycle. Ecosystems, 15(2), 256-268. Scholz, R.W., Roy, A.H., Hellums, D.T., 2014. Sustainable phosphorus management: a transdisciplinary challenge. in: Sustainable Phosphorus Management, Springer, pp. 1-128. Scholz, R.W., Ulrich, A.E., Eilittä, M., Roy, A., 2013. Sustainable use of phosphorus: A finite resource. Science of The Total Environment, 461(Supplement C), 799-803. Schoumans, O.F., Bouraoui, F., Kabbe, C., Oenema, O., van Dijk, K.C., 2015. Phosphorus management in Europe in a changing world. AMBIO, 44(2), 180-192. Schuiling, R.D., Andrade, A., 1999. Recovery of Struvite from Calf Manure. Environ. Technol., 20(7), 765-768. Schulte, E.E., Kelling, K.A., 1985. Soil Calcium to Magnesium Ratios--should You be Concerned? University of Wisconsin--Extension.

62

Senthilkumar, K., Nesme, T., Mollier, A., Pellerin, S., 2012. Conceptual design and quantification of phosphorus flows and balances at the country scale: The case of France. Global Biogeochemical Cycles, 26(2), n/a-n/a. Sharpley, A., Moyer, B., 2000. Phosphorus forms in manure and compost and their release during simulated rainfall. Journal of Environmental Quality, 29(5), 1462-1469. Sheldrick, W., Keith Syers, J., Lingard, J., 2003. Contribution of livestock excreta to nutrient balances. Nutrient Cycling in Agroecosystems, 66(2), 119-131. Shen, D., Yin, J., Yu, X., Wang, M., Long, Y., Shentu, J., Chen, T., 2017. Acidogenic fermentation characteristics of different types of protein-rich substrates in food waste to produce volatile fatty acids. Bioresource Technology, 227, 125-132. Shen, Y., Ogejo, J., Bowers, K., 2011. Abating the effects of calcium on struvite precipitation in liquid dairy manure. Transactions of the ASABE, 54(1), 325-336. Shinh, A., 2012. The outlook for industrial & food phosphates. Phosphates. CRU Group, El Jadida. Smil, V., 2000. phosphorus in the environment: natural flows and human interferences. Annual Review of Energy and the Environment, 25(1), 53-88. Smil, V., 2002. Phosphorus: global transfers. Encyclopedia of Global Environmental Change. John Wiley & Sons, Chichester, 536-542. Smith, V.H., Schindler, D.W., 2009. Eutrophication science: where do we go from here? Trends in Ecology & Evolution, 24(4), 201-207. Snell-Castro, R., Godon, J.-J., Delgenès, J.-P., Dabert, P., 2005. Characterisation of the microbial diversity in a pig manure storage pit using small subunit rDNA sequence analysis. FEMS Microbiology Ecology, 52(2), 229-242. Song, Y.-H., Qiu, G.-L., Yuan, P., Cui, X.-Y., Peng, J.-F., Zeng, P., Duan, L., Xiang, L.-C., Qian, F., 2011. Nutrients removal and recovery from anaerobically digested swine wastewater by struvite crystallization without chemical additions. Journal of Hazardous Materials, 190(1), 140-149. Soto, M., Méndez, R., Lema, J.M., 1993. Methanogenic and non-methanogenic activity tests. Theoretical basis and experimental set up. Water Research, 27(8), 1361-1376. Stadtman, E., Stadtman, T.C., Barker, H., 1949. Tracer experiments on the mechanism of synthesis of valeric and caproic acids by Clostridium kluyveri. J Biol Chem, 178, 677-682. Stamatis, N., Hela, D., Konstantinou, I., 2010. Occurrence and removal of fungicides in municipal sewage treatment plant. Journal of Hazardous Materials, 175(1), 829-835. Stancik, L.M., Stancik, D.M., Schmidt, B., Barnhart, D.M., Yoncheva, Y.N., Slonczewski, J.L., 2002. pH-dependent expression of periplasmic proteins and amino acid catabolism in Escherichia coli. Journal of bacteriology, 184(15), 4246-4258. Sterner, R.W., Elser, J.J., 2002. Ecological stoichiometry: the biology of elements from molecules to the biosphere. Princeton University Press. Stevens, G., Gladbach, T., Motavalli, P., Dunn, D., 2005. Soil calcium: magnesium ratios and lime recommendations for cotton. Journal of cotton science. Sturm, G., Weigand, H., Marb, C., Weiß, W., Huwe, B., 2010. Electrokinetic phosphorus recovery from packed beds of sewage sludge ash: yield and energy demand. Journal of Applied Electrochemistry, 40(6), 1069-1078. Suárez-Iglesias, O., Urrea, J.L., Oulego, P., Collado, S., Díaz, M., 2017. Valuable compounds from sewage sludge by thermal hydrolysis and wet oxidation. A review. Science of The Total Environment, 584-585(Supplement C), 921-934. Suzuki, K., Tanaka, Y., Kuroda, K., Hanajima, D., Fukumoto, Y., 2005. Recovery of phosphorous from swine wastewater through crystallization. Bioresource Technology, 96(14), 1544-1550.

63

Suzuki, K., Tanaka, Y., Kuroda, K., Hanajima, D., Fukumoto, Y., Yasuda, T., Waki, M., 2007. Removal and recovery of phosphorous from swine wastewater by demonstration crystallization reactor and struvite accumulation device. Bioresource Technology, 98(8), 1573-1578. Svanström, M., Fröling, M., Modell, M., Peters, W.A., Tester, J., 2004. Environmental assessment of supercritical water oxidation of sewage sludge. Resources, Conservation and Recycling, 41(4), 321- 338. Sverdrup, H.U., Ragnarsdottir, K.V., 2011. Challenging the planetary boundaries II: Assessing the sustainable global population and phosphate supply, using a systems dynamics assessment model. Applied Geochemistry, 26, S307-S310. Szögi, A.A., Vanotti, M.B., Hunt, P.G., 2015. Phosphorus recovery from pig manure solids prior to land application. Journal of Environmental Management, 157(0), 1-7. Taddeo, R., Kolppo, K., Lepistö, R., 2016. Sustainable nutrients recovery and recycling by optimizing the chemical addition sequence for struvite precipitation from raw swine slurries. Journal of Environmental Management, 180(Supplement C), 52-58. Takahashi, M., Kato, S., Shima, H., Sarai, E., Ichioka, T., Hatyakawa, S., Miyajiri, H., 2001. Technology for recovering phosphorus from incinerated wastewater treatment sludge. Chemosphere, 44(1), 23-29. Takiguchi, N., Kuroda, A., Kato, J., Nukanobu, K., Ohtake, H., 2003. Pilot plant tests on the novel process for phosphorus recovery from municipal wastewater. Journal of chemical engineering of Japan, 36(10), 1143-1146. Tang, J., Wang, X., Hu, Y., Zhang, Y., Li, Y., 2016. Lactic acid fermentation from food waste with indigenous microbiota: Effects of pH, temperature and high OLR. Waste Management, 52(Supplement C), 278-285. Tang, J., Wang, X.C., Hu, Y., Ngo, H.H., Li, Y., 2017. Dynamic membrane-assisted fermentation of food wastes for enhancing lactic acid production. Bioresource Technology, 234(Supplement C), 40- 47. Taylor, C.R., Moss, D.L., 2013. The fertilizer oligopoly: The case for global antitrust enforcement. The American Antitrust Institute, USA. Tenca, A., Schievano, A., Perazzolo, F., Adani, F., Oberti, R., 2011. Biohydrogen from thermophilic co-fermentation of swine manure with fruit and vegetable waste: Maximizing stable production without pH control. Bioresource Technology, 102(18), 8582-8588. Thauer, R.K., Jungermann, K., Decker, K., 1977. Energy conservation in chemotrophic anaerobic bacteria. Bacteriological reviews, 41(1), 100. Thompson, L.B., 2013. Field evaluation of the availability for corn and soybean of phosphorus recovered as struvite from corn fiber processing for bioenergy, Iowa State University. Tilman, D., 1998. The greening of the green revolution. Nature, 396(6708), 211-212. Tilman, D., Balzer, C., Hill, J., Befort, B.L., 2011. Global food demand and the sustainable intensification of agriculture. Proceedings of the National Academy of Sciences, 108(50), 20260- 20264. Toor, G.S., Hunger, S., Peak, J.D., Sims, J.T., Sparks, D.L., 2006. Advances in the Characterization of Phosphorus in Organic Wastes: Environmental and Agronomic Applications. Adv. Agron., 89, 1-72. Tsukahara, T., Koyama, H., Okada, M., Ushida, K., 2002. Stimulation of butyrate production by gluconic acid in batch culture of pig cecal digesta and identification of butyrate-producing bacteria. Journal of Nutrition, 132(8), 2229-2234. Ueno, Y., Fujii, M., 2001. Three Years Experience of Operating and Selling Recovered Struvite from Full-Scale Plant. Environmental Technology, 22(11), 1373-1381.

64

Uludag-Demirer, S., Demirer, G.N., Chen, S., 2005. Ammonia removal from anaerobically digested dairy manure by struvite precipitation. Process Biochemistry, 40(12), 3667-3674. Van Drecht, G., Bouwman, A.F., Harrison, J., Knoop, J.M., 2009. Global nitrogen and phosphate in urban wastewater for the period 1970 to 2050. Global Biogeochemical Cycles, 23(4), n/a-n/a. Van Vuuren, D.P., Bouwman, A.F., Beusen, A.H.W., 2010. Phosphorus demand for the 1970–2100 period: A scenario analysis of resource depletion. Global Environmental Change, 20(3), 428-439. Vanotti, M.B., Szogi, A.A., Hunt, P.G., Millner, P.D., Humenik, F.J., 2007. Development of environmentally superior treatment system to replace anaerobic swine lagoons in the USA. Bioresour. Technol., 98(17), 3184-3194. Vitousek, P.M., Naylor, R., Crews, T., David, M., Drinkwater, L., Holland, E., Johnes, P., Katzenberger, J., Martinelli, L., Matson, P., 2009. Nutrient imbalances in agricultural development. Science, 324(5934), 1519-1520.Vogel, T., Nelles, M., Eichler-Löbermann, B., 2015. Phosphorus application with recycled products from municipal waste water to different crop species. Ecological Engineering, 83(Supplement C), 466-475. Wang, Y., Tashiro, Y., Sonomoto, K., 2015. Fermentative production of lactic acid from renewable materials: Recent achievements, prospects, and limits. Journal of bioscience and bioengineering, 119(1), 10-18. Weil, B., 2006. The rivers come: Colonial flood control and knowledge systems in the Indus basin, 1840s-1930s. Environment and History, 3-29. Wild, D., Kisliakova, A., Siegrist, H., 1997. Prediction of recycle phosphorus loads from anaerobic digestion. Water Research, 31(9), 2300-2308. Withers, P.J.A., van Dijk, K.C., Neset, T.-S.S., Nesme, T., Oenema, O., Rubæk, G.H., Schoumans, O.F., Smit, B., Pellerin, S., 2015. Stewardship to tackle global phosphorus inefficiency: The case of Europe. AMBIO, 44(2), 193-206. Wollmann, I., Möller, K., 2015. Assessment of Alternative Phosphorus Fertilizers for Organic Farming: Sewage Precipitation Products. Wu, X., Zhu, J., Dong, C., Miller, C., Li, Y., Wang, L., Yao, W., 2009. Continuous biohydrogen production from liquid swine manure supplemented with glucose using an anaerobic sequencing batch reactor. International Journal of Hydrogen Energy, 34(16), 6636-6645. Wu, X., Yao, W., Zhu, J., 2010. Effect of pH on continuous biohydrogen production from liquid swine manure with glucose supplement using an anaerobic sequencing batch reactor. International Journal of Hydrogen Energy, 35(13), 6592-6599. Wu, X., Lin, H., Zhu, J., 2013. Optimization of continuous hydrogen production from co-fermenting molasses with liquid swine manure in an anaerobic sequencing batch reactor. Bioresource Technology, 136(Supplement C), 351-359. Wu, X., Zhu, J., Lin, H., 2017. In-depth observations of fermentative hydrogen production from liquid swine manure using an anaerobic sequencing batch reactor. Journal of Integrative Agriculture, 16(6), 1276-1285. Xing, Y., Li, Z., Fan, Y., Hou, H., 2010. Biohydrogen production from dairy manures with acidification pretreatment by anaerobic fermentation. Environmental Science and Pollution Research, 17(2), 392-399. Xu, H., He, P., Gu, W., Wang, G., Shao, L., 2012. Recovery of phosphorus as struvite from sewage sludge ash. Journal of Environmental Sciences, 24(8), 1533-1538. Xue, T., Huang, X., 2007. Releasing characteristics of phosphorus and other substances during thermal treatment of excess sludge. Journal of Environmental Sciences, 19(10), 1153-1158.

65

Ye, Z.-L., Chen, S.-H., Lu, M., Shi, J.-W., Lin, L.-F., Wang, S.-M., 2011. Recovering phosphorus as struvite from the digested swine wastewater with bittern as a magnesium source. Water Sci. Technol., 64(2), 334-340. Yin, J., Yu, X., Wang, K., Shen, D., 2016. Acidogenic fermentation of the main substrates of food waste to produce volatile fatty acids. Int. J. Hydrog. Energy, 41(46), 21713-21720. Yu, Y., Lei, Z., Yuan, T., Jiang, Y., Chen, N., Feng, C., Shimizu, K., Zhang, Z., 2017. Simultaneous phosphorus and nitrogen recovery from anaerobically digested sludge using a hybrid system coupling hydrothermal pretreatment with MAP precipitation. Bioresource Technology, 243(Supplement C), 634-640. Zeng, L., Li, X., 2006. Nutrient removal from anaerobically digested cattle manure by struvite precipitation. Journal of Environmental Engineering and Science, 5(4), 285-294. Zhang, B., Zhang, L.L., Zhang, S.C., Shi, H.Z., Cai, W.M., 2005. The influence of pH on hydrolysis and acidogenesis of kitchen wastes in two-phase anaerobic digestion. Environmental Technology, 26(3), 329-339. Zhang, D.-m., Chen, Y.-x., Jilani, G., Wu, W.-x., Liu, W.-l., Han, Z.-y., 2012. Optimization of struvite crystallization protocol for pretreating the swine wastewater and its impact on subsequent anaerobic biodegradation of pollutants. Bioresource Technology, 116(Supplement C), 386-395. Zhang, H., Lo, V.K., Thompson, J.R., Koch, F.A., Liao, P.H., Lobanov, S., Mavinic, D.S., Atwater, J.W., 2015. Recovery of phosphorus from dairy manure: a pilot-scale study. Environmental Technology, 36(11), 1398-1404. Zhang, X.C., Norton, L.D., 2002. Effect of exchangeable Mg on saturated hydraulic conductivity, disaggregation and clay dispersion of disturbed soils. Journal of Hydrology, 260(1), 194-205. Zhu, J., Li, Y., Wu, X., Miller, C., Chen, P., Ruan, R., 2009. Swine manure fermentation for hydrogen production. Bioresource Technology, 100(22), 5472-5477. Zhu, Y., Yang, S.T., 2004. Effect of pH on metabolic pathway shift in fermentation of xylose by Clostridium tyrobutyricum. Journal of Biotechnology, 110(2), 143-157.

66

OBJECTIVES OF THESIS

• Determine the effect of acidogenesis on pH, production of organic acids and dissolution of P, + Mg, NH4 and Ca when sucrose is used as a model organic co-substrate for biological acidification of pig slurry.

• Verify that lactic acid bacteria are responsible for lactate production during biological acidification of raw and digested swine manure using sucrose as co-substrate.

• Determine the effects of various organic wastes on acidification, the bacterial communities involved and the metabolic pathways they used.

• Evaluate the possibility to predict the lowest pH reached during acidogenesis based on co- substrate type and characteristics of swine slurry (initial pH, buffer capacity).

• Investigate the possible adaptation of acidogenic biomass to acidic conditions using semi- continuous feeding.

+ • Produce struvite from dissolved P, Mg, NH4 in the supernatant after acidification and centrifugation.

• Determine the purity of struvite in term of P content and organic, calcium and heavy metal contamination.

• Evaluate the mass balance for phosphorus throughout the process to calculate the associated P recovery.

67

CHAPTER 1 DISSOLUTION OF PARTICULATE PHOSPHORUS IN PIG SLURRY THROUGH BIOLOGICAL ACIDIFICATION

Piveteau S., Picard S., Dabert P., Daumer M-L. (2017). Dissolution of particulate phosphorus in pig slurry through biological acidification: a critical step for maximum phosphorus recovery as struvite. Water Research. DOI: 10.1016/j.watres.2017.08.017

68

I. SUMMARY OF CHAPTER 1

L’étude bibliographique a montré que pour optimiser la valorisation du phosphore il est nécessaire d’obtenir un produit à l’efficacité agronomique au moins comparable à celle des engrais phosphatés les plus efficaces. La struvite répond à ces conditions. Or, la production de struvite se fait à partir de P dissous. Le phosphore des lisiers étant majoritairement sous une forme minérale particulaire mélangée à la matière organique, une étape de dissolution suivie d’une étape permettant de séparer le liquide enrichi en P de la matière organique est nécessaire pour valoriser une fraction significative du P des lisiers de porcs. L’acidification chimique testée lors de travaux précédents s’est révélée efficace mais coûteuse sur le plan économique. De plus, elle impacte fortement le bilan environnemental du procédé. Des procédés d’acidification biologique comme l’ensilage pour la conservation des fourrages permettent de faire produire aux bactéries endogènes de l’acide in-situ. Ce type de procédé a également été utilisé pour acidifier le lisier avec l’objectif de réduire les émissions d’ammoniac. Dans ce cas l’apport d’un co-substrat facilement dégradable est nécessaire. La production d’acide in situ en quantité suffisante pour obtenir une baisse significative du pH à partir de déchets organiques utilisés habituellement comme co-substrat de la digestion anaérobie est une alternative qui permettrait de s’affranchir du coût de l’acide et de l’impact environnemental lié à sa production. L’étude bibliographique a également montré que les co-substrats sucrés sont les plus efficaces car ils permettent de produire de l’acide lactique qui est l‘un des plus forts des acides organiques.

Les objectifs des travaux décrits dans ce premier article sont de tester la faisabilité de l’acidification biologique et d’évaluer la quantité de sucre nécessaire en fonction des caractéristiques des lisiers (pH initial, pouvoir tampon …). L’effet de l’acidification biologique sur la dissolution des autres composés de la struvite (ammonium et magnésium) et du calcium qui peut interférer avec la cristallisation de la struvite a aussi été étudié. Pour cette première approche du saccharose (sucre blanc du commerce) a été utilisé. Les résultats obtenus avec le sucre permettront d’évaluer le potentiel de dissolution du P par voie biologique sur différents lisiers.

Quatre lisiers différents ont été utilisés. Leurs caractéristiques ont été analysées. Le pouvoir tampon a été estimé à partir de la quantité d’acide sulfurique 1mol/L nécessaire pour atteindre un pH de 4 qui correspond au pH le plus faible observé par acidification biologique sur du lisier (Table 4). Les essais ont été réalisés dans des conditions définies lors de travaux précédents sur les boues. Cinq ou sept concentrations de sucre comprises entre 0 et 60 g.L-1 ont été testées. Des prélèvements ont été réalisés toutes les 12 heures en début d’incubation puis toutes les 24 heures jusqu’à 96 heures. Les essais réalisés avec le lisier n°1 et sept concentrations de sucre différentes montrent que la baisse de pH obtenue est directement liée à la concentration en sucre ajoutée. Très peu de variations sont observées pour les faibles concentrations (0 et 10 g.L-1). Pour les valeurs intermédiaires (20 et 30 g.L-1) le pH minium, 6.2 et 5.5 est atteint en 24 h mais la baisse est suivie d’une remontée rapide 12 h après. Pour

69 les fortes concentrations le pH minimum est atteint entre 36 et 48 h puis reste stable. Un pH de 3.99 a été atteint avec la concentration la plus forte (Figure 22). Le dosage des acides organiques a montré que la production d’acide lactique est le principal responsable de la baisse du pH. Le bilan carbone montre que la presque totalité du sucre a été convertie en acide lactique et en acide gras volatils. Le faible pH atteint avec les concentrations en sucre les plus fortes permettrait d’inhiber les métabolismes de dégradation de l’acide lactique et de maintenir le pH stable. Les concentrations en phosphore, calcium et magnésium dissous sont liées à l’évolution du pH. Jusqu’à 95% du P total initial du lisier ont été dissous avec les trois concentrations en sucre les plus élevées. Une re-précipitation est observée lorsque le pH remonte pour les concentrations intermédiaires de sucre. Le même phénomène est observé avec le magnésium dont la dissolution atteint 100% avec les trois concentrations en sucre les plus fortes. Cependant la re-précipitation observée pour les valeurs moyennes est plus faible que pour le P. La fraction de calcium dissous est inférieure. Les deux plus fortes concentrations en sucre permettent de dissoudre 75% tandis que la concentration de 40 g.L-1 permet de dissoudre seulement 60% du calcium. Les courbes de dissolution de l’ammonium montrent que de l’azote est consommé pour la croissance de la biomasse acidogène dans les essais avec sucre. Une minéralisation des formes organiques sous l’effet du pH serait observée ensuite. La concentration en ammonium dissous dépend de l’équilibre qui s’établit entre ces deux mécanismes (Figure 24).

Les résultats obtenus avec les autres lisiers ont des tendances similaires. Ils ont permis de montrer que la dissolution du P et du magnésium en fonction du pH n’est pas linéaire. Elle intervient principalement pour des pH compris entre 5.5 et 6.5. En revanche la dissolution du calcium en fonction du pH est quasi-linéaire entre pH 7 et pH 4.5. Cette évolution différente permet de définir une zone de pH comprise entre 5.5 et 6 pour laquelle la dissolution du Ca comparée au P et au Mg est moindre ce qui favoriserait la cristallisation du P sous forme de struvite lors des étapes ultérieures. Le fait d’opérer la séparation avant cristallisation à ce pH réduirait également la quantité de réactif nécessaire à la remontée du pH lors de la cristallisation. En revanche la faible stabilité du pH dans cette zone demande une gestion précise du procédé d’acidification.

Finalement la régression linéaire mise en œuvre sur les données obtenues avec les quatre lisiers a permis d’établir une relation bien ajustée entre le pH minimal, la quantité de sucre, le pH initial du lisier et son pouvoir tampon pour une zone de pH entre 4 et 8. Cette relation pourra être utilisée pour calculer la quantité de co-substrat à ajouter. Ces travaux ont montré le lien étroit entre la production d’acide lactique et le pH. Or l’acide lactique est le produit majoritaire de dégradation des sucres simples très facilement dégradables. Les travaux suivants ont pour objectif de s’assurer que l’acidification peut être obtenue avec les co-substrats riches en glucides complexes disponibles sur les unités de méthanisation et de tester la faisabilité d’un recyclage en parallèle de la méthanisation pour exploiter la présence d’une flore déjà accoutumée à la dégradation anaérobie de co-substrats complexes comme cela a été envisagé à la fin de l’étude bibliographique.

70

Dissolution of particulate phosphorus in pig slurry through biological acidification: a critical step for maximum phosphorus recovery as struvite

Piveteau Simona,b,*, Picard Sylviea,b, Dabert Patricka,b, Daumer Marie-Linea,b a Irstea, UR OPAALE, 17 Avenue de Cucillé-CS 64427, F-35044 Rennes, France b Univ de Bretagne Loire, France.

*Corresponding author. Simon Piveteau. UR OPAALE, Irstea. E-mail address: [email protected]

71

II. ABSTRACT

Recycling phosphorus as struvite from pig slurry requires an acidification step to dissolve the inorganic solids containing most of the phosphorus. This study focused on the biological acidification of several pig slurries using sucrose as a model organic co-substrate. Lactic acid fermentation occurred systematically, dissolving 60 to 90% of TP (total phosphorus) and T-Mg (total magnesium) at pH 6 or lower. Optimal pH range for maximum P dissolution aimed at struvite recovery was 5.5-6. A simple model was developed correlating pH, sucrose and buffer capacity to optimize P dissolution and future recovery using real organic waste.

72

III. INTRODUCTION

The anthropogenic use of phosphorus presents a double problematic: on one hand it is contributing to water bodies pollution due to a combination of fertilization excess and leakage/runoff (MacDonald et al., 2011), and on the other hand it is a fossil resource critical for agriculture and worldwide food production (Cordell et al., 2009). As a result, recovering phosphorus from pig slurry as a mineral fertilizer could be a useful tool to mitigate water pollution from pig breeding and participate in a more sustainable food production (Ashley et al., 2011). Brittany (France) possesses only 6% of the national arable land but represents 60% of French the pig production (Comité régional porcin de Bretagne, 2013). Stringent regulations now prevent the unlimited spreading of the slurry into the fields and various handling strategies are currently used in the region (Landrain et al., 2013). Among them is anaerobic digestion (Levasseur and Lemaire, 2006), in which various organic wastes are co-digested with the slurry in order to produce energy and heat. Exportation of manure out of Brittany is another strategy to reduce the environmental impact on soil (Martinez et al., 2009). Part of the digested and undigested manure is separated using screw press or centrifuge decanter to collect a liquid phase rich in nitrogen that is treated via biological processes, while most of the phosphorus remains in the solid phase under a mineral form (Christensen et al., 2009; Daumer et al., 2005). A large part of this solid phase has to be exported out of Brittany (usually as organic fertilizer) but this product has trouble finding its market. It cannot compete with compost because of its high P-content when organic matter is the customer’s driving choice, and cannot compete either with mineral fertilizers because of the transport cost of organic matter when P is the component demanded.

Phosphorus recovery as struvite (NH4MgPO4, 6 H2O), a slow-release fertilizer, would make the export of P more cost effective, meet the market demand for a concentrated, P-based mineral fertilizer and allow the organic matter in the digestate/compost to be maintained locally thanks to its low P content. Phosphorus recovery as struvite has been studied in a large variety of N and P concentrated waste streams: WWTP centrate liquor from anaerobic digestion (Battistoni et al., 2001; Jaffer et al., 2002), industrial wastewater (Abma et al., 2010), municipal landfill leachate (Di Iaconi et al., 2010), dairy wastewater (Zhang et al., 2010) as well as swine lagoon effluent (Nelson et al., 2003), digested swine wastewater (Song et al., 2011; Vanotti et al., 2017; Ye et al., 2011) and raw swine manure (Burns et al., 2001; Çelen et al., 2007). Several full scale reactors have been built for struvite crystallization, such as the OSTARA, WASSTRIP (Cullen et al., 2013) and PHOSPAQ processes (Remy et al., 2013). However these technologies are only used so far to treat municipal and industrial wastewater. In the case of animal manure, precipitation of phosphorus as struvite is still investigated at laboratory and pilot scales (Rahman et al., 2014), focusing mainly on reducing the dissolved phosphate concentration to meet the discharge standard for the liquid effluent. However, if the goal is to recover phosphorus, dissolving the inorganic solids that contain most of the phosphorus is necessary. It has been demonstrated that up to 70% of TP could be dissolved at pH 5.5 (Christensen et al., 2009). P recovery

73 as struvite can therefore be achieved through acidic dissolution, solid-liquid separation and pH increase in the liquid phase using magnesium oxide (MgO) or magnesium hydroxide (Mg(OH)2) leading to phosphorus precipitation (Capdevielle et al., 2016). The use of chemicals for acidic dissolution was demonstrated as too expensive and having a negative relative impact on the environment (Daumer et al., 2010). For the process to be efficient and sustainable, changing this acidification step is required. A biological process known as acidogenesis enables the natural acidification of a mixture under anaerobic conditions. During acidogenesis, organic matter is hydrolyzed and converted by fermenting bacteria into VFAs, lactic acid and alcohols. This acidifying reaction process is a crucial step in silage production (Dunière et al., 2013), a technic used to preserve the nutritive value of forage for cattle feeding (Buxton et al., 2003). Acidogenesis also occurs during anaerobic digestion of organic waste. The syntrophic cooperation of acidogenic, acetogenic and methanogenic organisms enables the conversion of organic waste into methane and carbon dioxide. Because acidogens have different optimal conditions compared to acetogens and methanogens (Liu et al., 2006), and especially a higher maximum growth rate, an overloading of a digester with easily fermentable substrates often lead to acidosis and inhibition of methanogens (McCarty and McKinney, 1961). As a result, the process can be physically separated into two consecutive reactors one acidic and one methanogenic, each of them functioning under their respective optimum (Cohen et al., 1979). This two stages process has been applied to crop residues (Kalia et al., 1992; Parawira et al., 2008), agro-industrial waste (Dareioti and Kornaros, 2014), and food waste (Uçkun Kiran et al., 2014). Biological treatment of swine slurry in a two stage anaerobic process has recently been studied (Ren et al., 2014; Schievano et al., 2012), often with the purpose of biohydrogen and methane production (Choi et al., 2015; Wu et al., 2010). It has been shown that a two-stage anaerobic digestion of fruit and vegetable waste mixed with swine manure at a ratio of 70/30 (w/w) lead to a pH of 4.2 in the acidogenic stage (Tenca et al., 2011), a pH at which phosphorus is known to be mostly dissolved in pig slurry (Sharpley and Moyer, 2000). Therefore, biological acidification of swine slurry with organic co-substrates could be proposed as the first stage in a bioprocess aimed at nutrient recovery (N & P) and green energy production (methane). Acidogenesis would dissolve P from pig slurry at lower cost than chemical acidification while providing VFAs and potentially increasing the biodegradability of the waste for the subsequent anaerobic digestion. This batch test study focuses on the biological acidification of pig slurry using various concentrations of sucrose as a model organic co-substrate, adapting the methods developed by Braak et al. (2016) and Guilayn et al. (2017) as a first step for phosphorus recycling. Biological acidification of pig slurry using glucose, cellulose or lactic acid has been tested to minimize ammonia emission by lowering the pH (Hjorth et al., 2015). The results in terms of pH decrease and organic acids produced were very similar to this study. The objective here was to determine the relationship between on one hand initial amount of sucrose and slurry’s characteristics (initial pH, buffer capacity) and on the other hand pH change, fermentation products and dissolution of phosphorus, magnesium, ammonia (the three

74 molecules forming struvite) and calcium (which lead to the undesired precipitation of calcium phosphate). This endeavor could then translate into a targeted acidification step using real organic waste that would maximize P-dissolution, favor struvite formation over calcium phosphate and minimize the amount of reactant necessary to re-increase the pH.

IV. MATERIALS AND METHODS

1. PIG SLURRY

The raw swine slurry (“slurry 1”) was collected at a commercial pig fattening farm in Melesse (Brittany, France). Its composition can be found in Table 4. It was used to study the metabolism of acidogenesis in details. Three additional pig slurries (Table 4) were also tested in order to correlate the level of sucrose added with the initial and lowest pH reached, as well as the buffer capacity of the slurry. The buffer capacity was measured in a beaker containing 100 mL of raw pig slurry with a magnetic stirrer, a pH probe and a graduated burette containing sulfuric acid (1 mol/L). The buffer capacity was calculated has the concentration of sulfuric acid necessary to reach pH 4, the lowest pH obtained during the biological acidification of the slurries (pH 3.99 was reached in slurry 2 with 60 g/L of sucrose).

Table 4. Composition of the pig slurries (concentrations expressed per liter of raw manure)

TSS VSS TS VS TP TN T-Ca T-Mg pH Buffer PO -P NH -N Ca2+ Mg2+ 4 4 g/L mEq/L g/L slurry 1 75,8 61.0 83.4 65.1 1.68 4.60 2.65 1.10 7.15 443 0.005 2.03 0.21 0.07 slurry 2 15.3 10.4 22.0 13.8 0.38 2.346 1.22 0.29 7.70 99 0.006 1.62 0.11 0.03 slurry 3 64.3 42.9 77.5 49.2 1.37 5.742 3.59 1.34 7.92 283 0.018 3.42 0.07 0.03 slurry 4 41.9 33.8 58.0 41.4 0.53 4.774 1.59 0.50 6.70 380 0.055 2.51 0.59 0.33

2. BIOLOGICAL ACIDIFICATION OF PIG SLURRY IN BATCH TESTS USING SUCROSE AS CO- SUBSTRATE

One liter bottles were inoculated with 640 g of raw pig slurry 1 and six different sucrose concentrations (10, 20, 30, 40, 50 and 60 g/kg-slurry, in reactors called R10, R20, R30, R40, R50, R60 respectively) in triplicate, with an additional bottle inoculated with slurry only and used as a control. The bottles were sealed once inoculated. Temperature was set at 38°C on heating plates and magnetic stirrers provided continuous mixing at 300 rpm. Samples were collected simply by removing the lid and pouring the liquid from the bottles. Contrary to previous batch tests with wastewater sludge, purging the gas in the head-space of the bottles at the beginning and after sample collection had no effect on the system when applied to pig slurry, therefore the bottles were never flushed with inert gas. Sample collection occurred after 12, 24, 36, 48, 72 and 96 hours. The same protocol was applied to slurry 2, 3 and 4 using only 5 different sucrose concentrations for each, from 0 to 60 g/L, in order to investigate the variations between slurries.

75

3. ANALYSIS pH was measured using a WTW probe right after each sample collection. Total solids (TS), volatile solids (VS), total suspended solids (TSS), and volatile suspended solids (VSS) were measured in triplicates on the raw slurry with standard methods (APHA, 1998). An acidic dissolution of the ashes was realized to measure TP, T-Ca and T-Mg. 200mg of ashes were added to 0.5g of K2SO8 and 5mL of a mix of H2SO4 / HNO3 (75:25) in triplicate. The samples were autoclaved at 110°C during one hour at 1 bar. The concentrations in TP, T-Ca and T-Mg (expressed in gram per liter of raw manure in Table 4) were measured using automated colorimetric methods on a spectrophotometer (Gallery, Thermoscientific). For TP measurement, two reactants (containing sulfuric acid, potassium antimonyl tartrate ammonium molybdate and ascorbic acids, Thermoscientific products) were added automatically to the sample. Phosphates react with ammonium molybdate under acid conditions to form a 12-molybdo-phosphoric acid complex. The complex is reduced by the ascorbic acid to form a blue compound. The absorbance was measured by spectrophotometry at the wavelength of 880 nm. For T-Ca measurement, a reactant containing Arsenazo III and an imidazole buffer was added to the sample. Calcium ions form a highly colored complex at neutral pH. The absorbance of the complex was measured at 660nm. For T-Mg measurement, a reactant containing Xylidyl I blue, EGTA, K2CO3 and a Tris buffer was added to the sample. At alkaline pH magnesium ions react with Xylidyl I blue to form a red-colored complex. The absorbance was measured at 520 nm. . Supernatants were obtained after 20 min of centrifugation (4°C, 20,000 g) and filtration on a 0.45 µm polypropylene membrane. Phosphate concentration was measured with automated colorimetric methods on the Gallery. Cationic composition of the supernatant was measured by ionic chromatography Metrohm 940 Professional Vario IC with a Metrosep C4 -250/4,0 column. Sucrose, glucose, fructose and VFA composition of the supernatant was measured by HPLC (Ultimate 3000, Dionex) with a Hiplex-H column (Agilent).

4. STATISTICS

In order to evaluate the effect of sucrose concentration on pH and ionic concentrations, an ANOVA was performed at each sampling time (12, 24, 36, 48, 72 and 96 hours). The conditions of application for ANOVA (normal distribution of the residuals and equal variance of the residuals for each modality) were verified with a Shapiro-Wilk test and a Bartlett test. The different modalities (i.e. level of initial sucrose concentration) were then compared two by two with a Tukey’s HSD test.

76

V. RESULTS AND DISCUSSION

1. EFFECT OF SUCROSE CONCENTRATION ON PH AND FERMENTATION PRODUCT IN SLURRY 1 a. PH CHANGE

Figure 22 shows the effect of sucrose addition on pH in slurry 1. Considering that sucrose would likely be fermented mostly into VFAs and other organic acids, the pH drop was logically absent in R0, moderate at R10, R20, R30 and severe in R40, R50, R60 (Figure 22). Indeed, the pH initially dropped in all the reactors except in the control where it remained stable throughout the experiment. The pH drop was significantly more pronounced and longer at higher initial sucrose concentration. After 24 to 36 h, pH re-increased in the reactors with an initial sucrose concentration of 10, 20 and 30 g/L whereas at higher initial sucrose content (40, 50, 60 g/L), pH dropped below 5 at 36 hours and remained stable or kept decreasing afterwards. The pH re-increase in R10, R20, R30 could have resulted from either (consumption of VFAs to form methane and CO2) or secondary fermentation (conversion of VFAs initially produced into other, less acidic compounds). Since no significant amount of methane could be detected (measurement at 24 and 48 h, data not shown), a secondary fermentation seemed the most likely and was later confirmed (see below).

Figure 22. Evolution of pH across time for each initial sucrose concentration (standard deviation displayed)

b. EVOLUTION OF LACTIC ACID, VFAS AND SUCROSE ACROSS TIME

As shown in Figure 23A and 23B (R30 and R50 representing moderate vs. high initial sucrose content), a lactic acid fermentation occurred in all the reactors with lactate as the main product as long

77 as sucrose was still present. The preponderance of lactic acid in this first phase can be seen through both its proportion among the acids produced and its strong linear correlation with pH (r² = 0.97).

Once sucrose had been removed, lactate disappeared in the next 12 to 24 h in reactors R10, R20, R30 while acetate and propionate concentrations increased and butyrate and valerate appeared. The replacement of lactate by less acidic VFAs (acetate, propionate, butyrate and valerate have a pKa of 4.75, 4.87, 4.82 and 4.86 respectively versus a pKa of 3.89 for lactate) could explain the pH re- increase. On the contrary, lactate concentration remained stable in R40, R50 and R60 once sucrose had been consumed, without any increase in other VFAs concentration or any apparition of butyrate and valerate, while pH stayed low. Carbon recovery, calculated as the ratio between the net production of organic acids (VFAs + lactic acid) and the removed sucrose, can be found in Table 5 At 12 hours, the carbon recovery was relatively poor (67-89% and high standard deviations) possibly due to variations in the lag phase between triplicates and the transitory storing or initial catabolic steps of glucose and fructose (glycolysis) in the bacterial cells. At 36 h, the C-recovery was around 95% in all the reactors, with most of the sucrose converted to lactate. The C-recovery in R40 to R60 remained similar until the end since the metabolic activity was severely reduced due to the low pH. In R10 to R30, the C- recovery decreased down to 80-85% at 48 hours due to the conversion of lactate to acetate, propionate, butyrate and valerate, metabolic processes involving CO2 production, which was not measured in this study. The C-recovery in those reactors increased slightly at 72 and 96 hours to 87-93%, likely due to the fermentation of organic matter naturally present in the slurry.

Table 5. Carbon recovery in R10 to R60 at each sampling time. Mean of three incubations and standard deviation are displayed. Carbon recovery at maximum lactic concentration shown in bold

time (hours) R10 R20 R30 R40 R50 R60 12 82 ± 15 84 ± 41 67 ± 40 89 ± 59 78 ± 60 67± 9 24 95 ± 11 83 ± 11 91 ± 8 90 ± 4 103 ± 1 99 ± 4 36 94 ± 5 96 ± 1 97 ± 5 93 ± 11 93 ± 3 96 ± 2 48 85 ± 18 83 ± 9 81 ± 7 97 ± 5 94 ± 9 103 ± 4 72 89 ± 16 89 ± 5 87 ± 3 97 ± 9 97 ± 6 97 ± 2 96 93 ± 10 92 ± 12 88 ± 12 100 ± 6 97 ± 9 101 ± 3

The differing behaviors between low and high initial sucrose concentration in terms of pH change and whether or not lactate was converted to other VFAs could be explained by the physiology of the bacterial community during the batch tests and the scientific literature on ensiling process. Pig slurry is known to contain relatively high amounts of LAB and Clostridia (Snell-Castro et al., 2005), two bacteria critical in the silage process. During ensiling, in presence of a sufficient amount of carbohydrates (sucrose in this case), amino acids and vitamins (both present in large amounts in pig slurry), the LAB can outcompete the other microorganisms and dominate the fermentation thanks to their higher growth rate (McDonald, 1982). If the pH reached at the end of the lactic acid fermentation is around 5 or lower, the activity of other anaerobic organisms, such as Enterobacteria and Clostridia

78 is completely inhibited (Pahlow et al., 2003), preventing the conversion of lactate to other fermentation products and further degradation of the organic matter (Kempton and Clemente, 1959). Therefore the fermentation process in R40, R50 and R60 could reasonably be compared to a successful ensiling (yet at a much quicker pace and with a lower dry matter content), since lactate was the main product throughout the test and pH remained below 5 after 36 hours.

In a failed ensiling process, the initial carbohydrate content is too low and the pH reached after the lactic acid fermentation is too high to prevent a secondary fermentation of lactate and protein degradation into biogenic amines, ammonia and organic acids (Rooke and Hatfield, 2003). That seems to be what took place in R10, R20 and R30. After 36 hours, pH was still at 5.5 or higher (Figure 22), sucrose could not be detected anymore and the only substrates available were lactate and organic matter from the slurry. Lactic acid fermenting Clostridia are known to have the ability to convert lactate to acetate, propionate (Johns, 1952; Kuchta and Abeles, 1985) and butyrate (McDonald, 1982; Woolford and Pahlow, 1997). Some Clostridia are highly proteolytic and can use amino acids in their catabolism, producing acetate, propionate (Wood, 2012), butyrate (Elsden and Hilton, 1979) and valerate (Dehority et al., 1958). Pig slurry contains also large amounts of Selonomonas Ruminantium and to a lesser extent Megasphaera elsdenii (Tsukahara et al., 2002), two bacterial species able to grow on lactate. Megasphaera elsdenii can use lactate to form acetate, propionate, butyrate and valerate from lactate in absence of glucose (Rogosa, 1971). Several strains of Selonomonas

Ruminantium produce acetate, propionate and CO2 from lactate (Wallace, 1978). Therefore, the disappearance of lactate and higher production of acetate, propionate, butyrate and valerate were likely the result of increased activity from lactic acid fermenters.

Coming back to the goal of building a process allowing biological acidification of pig slurry as a first step of phosphorus recycling, these results indicate that the most suitable co-substrates for pig slurry acidification would contain high amounts of soluble carbohydrates to favor LAB, and low quantities of proteins, whose degradation produces amines and NH3 with an alkaline effect on pH (Rooke and Hatfield, 2003). Fruits and vegetable waste (apples, carrots, potatoes) as well as corn waste could be the most promising co-substrates and will be investigated in the near future.

79

Figure 23. Net production of VFAs and lactate, sucrose removed and pH across time. Standard deviation shown for the sucrose removed and the sum of VFAs and lactate. A: Low initial sucrose concentration (30 g/L). B: High initial sucrose concentration (50 g/L)

2. PHOSPHORUS, MAGNESIUM, CALCIUM AND AMMONIA DISSOLUTION PROCESSES IN SLURRY 1

Dynamics of phosphorus, magnesium, calcium and ammonia dissolution in slurry 1 are shown in Figure 24A, B, C and D respectively. The patterns of dissolution of phosphorus, magnesium and calcium were similar. Initially these compounds were mostly under solid form (>94%). At low initial sucrose concentration (R0 and R10), no additional dissolution occurred. A moderate dissolution followed by partial or total re-precipitation took place at moderate sucrose concentrations (R20 and R30), and at high sucrose concentration the dissolution was almost total (R40 to R60).

As seen in Figure 24A, less than 80 mg-P/L (<5% TP) was dissolved at any time in R0 and R10, an amount significantly lower than the other reactors at 24 and 36 hours. In R20 and R30, respectively 400 and 1300 mg-P/L were dissolved at 24 and 36 hours, but re-precipitated afterwards, down to levels statistically similar to R0 and R10. In R40, R50 and R60, dissolved P was always statistically similar among them, and became significantly higher than the other reactors from 36h to the end, with up to 1600 mg-P/L (95% TP).

The same pattern occurred for Mg2+ (Figure 24B). In R0 and R10, less than 160 mg/L (or 15% T-Mg) were dissolved throughout the experiment. In R20, 400 mg/L or 36% of T-mg were dissolved at 24 and 36 hours but the precipitation that followed was only partial, with approximately 220 mg/L or

80

20% remaining dissolved until the end of the test, a number significantly higher than in R0 and R10. The same but amplified phenomenon occurred in R30, with around 1100 mg/L (or close to 100% T- Mg) at 24 and 36 hours, followed by a partial re-precipitation with still 600 mg/L or 50% T-Mg. In R40, R50 and R60, dissolved magnesium concentrations were always statistically similar and became significantly higher than the other reactors from 48h to the end, with up to 1100 mg/L (100% of T- Mg).

Similarly in the case of calcium (Figure 24C), no more than 200 mg/L or 6% of T-Ca were dissolved in R0 and R10 during the whole duration of the test. Respectively 500 and 1000 mg/L (or 14 and 28% of T-Ca) were dissolved in R20 and R30 at 24 and 26 hours but only approximately 300 and 400 mg/L remained afterwards, concentrations still significantly higher than in R0 and R10. R40, R50 and R60 had statistically similar calcium concentrations until 36 hours but R40 then stagnated around 2200 mg/L or 60% of T-Ca while R50 and R60 had a further increase in Ca concentration with around 2700mg/L or 75% of T-Ca at 48 hours before remaining stable until the end of the test.

Interestingly, the evolution of P, Ca and Mg in all the reactors followed closely the changes in pH (Figure 25A). The dissolved P concentration did not increase linearly with the decreasing pH. A clear and relatively narrow threshold appeared between 5.9 and 6.3, through which dissolved P went from less than 10% to 80% of TP. Below pH 5.9, only 10 additional percent of TP could be gained. Mg2+ followed a similar pattern, with less than 20% of T-Mg above pH 6.5 and more than 95% below pH 5.9. Calcium dissolution was more gradual; with less than 20% of T-Ca above pH 6.2, 40% at pH 5.5 and more than 95% below pH 4.5.

The change in ammonium concentration did not follow the same pattern (Figure 24D) and was + apparently less linked to pH (Figure 25B). In Figure 24D, NH4 initially dropped in all the reactors except R0. The drop was more pronounced in the case of high initial sucrose concentrations. After this + initial drop, NH4 concentration in R10 to R60 re-increased and three statistically different groups + appeared after 48h. R40, R50, and R60 had a stable NH4 concentration around 2800 mg/L or 60%-TN until the end of the test. R30 had around 2500 mg/L or 55%-TN. R10 and R20 had still a slightly + increasing NH4 concentration from 2100 mg/L at 48h to 2300 mg/L or 50%-TN at the end. In R0 ammonium concentration increased linearly for 48 hours and stabilized around 2500 mg/L, similarly + to R30. The increase in NH4 in absence of sucrose and despite the high pH in R0 contrasted with the absence of any dissolution of P, Ca and Mg in the same reactor. The overall increase in ammonium concentration over time in all the reactors with and without pH decrease could be explained by the enzymatic degradation of organic nitrogenous compounds (e.g. amino acids). Initial sucrose concentration has an impact on two phenomena with opposite effects on ammonia concentration: dissolution of inorganic solids through acidification on one hand and bacterial growth on the other hand. The low pH reached at high sucrose concentration enabled a higher ammonia release through the dissolution of inorganic N-containing compounds (e.g. struvite initially present in the raw slurry).

81

However, initial sucrose concentration is also proportional to biomass growth and associated N- + assimilation. The assimilation of NH4 for bacterial growth could explain for example the drop at 12h in all the reactors except R0. The balance between these contradicting effects could explain why after 96 hours R0 and R30 have a similar ammonium concentration, higher than R10 and R20 but lower than R40, R50 and R60.

Figure 24. A: Dissolved phosphorus across time for each sucrose concentration. B: Dissolved magnesium. C: Dissolved calcium. D: Ammonium

82

Figure 25. A: Dissolved phosphorus expressed as percentage of the total amount, plotted against pH. B: Dissolved magnesium. C: Dissolved calcium. D: Ammonium

3. DISSOLUTION OF P, MG, CA, N IN THE OTHER SLURRIES

+ The dissolution processes in slurry 1 were closely linked to pH except for NH4 . The same was true for slurries 2, 3 and 4 and the results for all the slurries can be found in Figure 26A, B, C and D for P, Mg, Ca and N respectively.

In the case of phosphorus, the inflexion point noticed for slurry 1 can be seen again around a pH of 6- 6.2. However, the maximum proportion of dissolved P varied significantly in one slurry, with only 58 % of TP in slurry 2 compared to 95, 89, and 78% in slurry 1, 3 and 4 respectively. No analysis was performed to determine the inorganic compounds that contained phosphorus in the slurries. However, based on the limited calcium dissolution in slurry 2 (60% of T-Ca maximum), it could be hypothesized that this slurry contained an important amount of calcium phosphate under a low solubility form (e.g. hydroxyapatite, Ca10(PO4)6(OH)2). Maximum P dissolution was obtained when the minimum pH was reached in slurry 1 and 3 but not in slurry 2 and 4 where maximum P dissolution occurred at a moderate pH around 5 while dissolved P decreased at lower pH. A possible explanation could come from a lower TP content in those slurries compared to slurries 1 and 3. This low TP content means that at high sucrose concentration and consequently at low pH, the P incorporated into the biomass would not be negligible and impact negatively the ratio dissolved-P/TP. Considering a biomass chemical - formula as C5H7NO2P0.074 (Droste, 1997), and a yield on sucrose of 0.14 (biomass COD.sucrose COD 1) (Gujer and Zehnder, 1983), the consumption of 60g/L of sucrose corresponds to the incorporation of 126 mg-P/L, i.e. 33 and 23% of TP in slurries 2 and 4, but only 7 and 9% in slurries 1 and 3. Another

83 phenomenon potentially disconnecting dissolved P from pH is the hydrolysis of solid organic P and the mineralization of dissolved organic P. Organic phosphorus can represent up to 22% of TP (Daumer et al., 2008). Since hydrolysis and mineralization are partially time-dependent it could explain why in the case of slurry 4 the maximum dissolved P concentration was reached despite a pH re-increase (from 68% of TP at pH 4.9 to 78% at pH 5.2)

Calcium dissolution was not necessarily as linear as in slurry 1. Slurries 2 and 4 had Ca dissolution as high as 50% of T-Ca at a pH of 6.5, while the 50% value was reached only at pH 5.5 in slurries 1 and 3. The mineral form of calcium should be investigated to explain this difference. Magnesium dissolution varied somewhat among the slurries, with the 50% threshold reached between pH 6 and 7. The maximum dissolution was reached around pH 6, and varied between 70 and 100%. Ammonia was not linked with pH in any slurry, for the same reasons described above.

Figure 26. A: Dissolved P expressed as a percentage of TP slurry 1, 2, 3 and 4. B: Dissolved Mg C: Dissolved Ca. D: Ammonium

4. POTENTIAL FOR STRUVITE CRYSTALLIZATION

In order to precipitate the dissolved phosphorus into struvite, pH needs to be increased to 7 or above. This could be done through and methanogenesis with the conversion of lactate and VFAs into acetate, H2 and subsequently methane and carbon dioxide. However, phosphorus solids would be mixed with anaerobic sludge and inorganic solids, making the segregation of struvite crystals difficult. Therefore, the acidified slurry should be centrifuged to separate the dissolved P from the suspended

84 solids. Addition of MgO or Mg(OH)2 to the liquid phase can provide the pH increase necessary and shift the precipitation process toward struvite rather than calcium phosphate thanks to their magnesium content. The optimal parameters to get a maximum P-recovery and favor struvite versus calcium phosphate are a pH close to 7, a molar N:P ratio above 3:1, a Mg:P ratio above 1:1, a Mg:Ca ratio around 2.25:1 or higher, and an addition of MgO as low as possible to favor the formation of large crystals (Capdevielle et al., 2013). In the pH range where most of the P was dissolved (pH 4 to 6), the N:P ratio never went below 3.9 in any of the four slurries, indicating that the concentration of ammonium and phosphorus in acidified pig slurry are very suitable for struvite crystallization. The Mg:P ratio was between 1 and 2.7 in all the slurries. Since MgO would be added to increase the pH, this ratio would increase even further, leaving large amounts of Mg in excess. Impact of this excess on subsequent anaerobic digestion should be investigated. The Mg:Ca ratio however was always below 2.25 in all the slurries, with a value of 2 at pH 6 and a decrease to 0.5 at pH 4. Magnesium dissolution stopped at pH 6 while calcium concentration continued to increase until pH 4, which explains why the Mg:Ca ratio decreased with the decreasing pH. The acidic buffer created by the production of VFAs during the acidifying step could require a large amount of MgO or Mg(OH)2 to bring the pH up. This would reduce the crystals size and represent an additional cost. As a result, an optimal acidification would lead to a pH low enough for maximum P dissolution but also as close as possible to 7 to minimize Ca dissolution and the need for MgO. As seen in Figure 26A, the gain in dissolved P below a pH of 5.5 was minimal, null, or negative. On the other hand, above pH 6, dissolved P was always lower than 50%. As a result, targeting a pH between 5.5 and 6 should provide a relatively high dissolved P, allow for a minimum addition of MgO while dissolved magnesium is already high and calcium not yet completely dissolved. These recommendations only apply to biological acidification of pig slurry using sucrose. When organic waste is used, hydrolysis and mineralization takes place, releasing into the liquid phase part of the nitrogen, phosphorus, magnesium and calcium contained in the waste (Mehta and Batstone, 2013), changing the ionic equilibriums. As a result, the effect of organic waste on acidification and ionic equilibrium in pig slurry should be investigated.

5. CORRELATING LOWEST PH WITH INITIAL SUCROSE CONCENTRATION AND BUFFER CAPACITY

Considering that the ideal pH range to maximize P recovery and minimize MgO is 5.5-6, it would be practical to be able to predict the amount of co-substrate necessary to reach it in any given pig slurry. Indeed, an excess of co-substrate would guarantee a maximum dissolution of P, but require an important amount of MgO to precipitate it. Conversely an insufficient addition of co-substrate would lead to a low P dissolution. Applying a linear regression on the data obtained from the batch tests on four pig slurries, it was possible to correlate the initial sucrose concentration, initial pH, buffer capacity (as the explanatory variables) with the lowest pH reached (as the variable to be explained). Residuals’ normality was verified with a Shapiro-Wilk test. Other characteristics of the pig slurries

85 were also tested (such as total solids, inorganic solids) but none had a statistically significant effect. The distribution of the observed pH plotted against their predicted value can be seen in Figure 27. The relatively good fitting of the model was possibly due to the dominance of lactic acid during the initial acidifying phase, i.e. the lowest pH was always reached at maximum lactic acid concentration, which corresponded to 70-100% of the VFAs produced at the moment. The equation was rewritten to predict the amount of sucrose necessary to reach a given pH (1). This model cannot be used as such for real co-substrates. As a result, waste commonly used in Brittany in centralized biogas plants will be tested in batch tests in order to rate their acidifying capacity versus sucrose. The equivalency to be found should allow the substitution of sucrose in the equation (Equation 14) with its equivalent in co- substrate, expressed as volatile solids.

Equation 14. Calculation of initial sucrose concentration necessary to reach a targeted pH

Sucrose0 (g/L) = -119.96 -18.95 pHlow+ 30.29*pH0 + 0.09*Buffer (mEq/L)

Statistical coefficients based on the original equation obtained with the linear regression: R² = 96.2% Adjusted R²= 95.8% Estimated standard deviation = 0,276 Mean absolute error = 0,221

Figure 27. Measured pH plotted against model-predicted pH

86

VI. CONCLUSIONS

This work demonstrated that most of the phosphorus (60-90% of TP) present in pig slurry can be dissolved using a simple biological process known as acidogenesis. A limited acidification down to a pH of 5.5-6 using sucrose allowed maximum P dissolution while providing optimal conditions for P recovery as struvite. The model correlating the amount of co-substrate with buffer capacity, initial and lowest pH can be used to adjust the amount of co-substrate to reach the desired pH range. It should be improved and expanded using real organic waste. The full conversion of sucrose into lactic acid and VFA indicates that organic waste with high soluble carbohydrates content should be used as co- substrate for biological acidification. The synergy/antagonism between P recovery and renewable energy production in the subsequent methanisation stage needs to be investigated.

VII. ACKNOWLEDGEMENT

IRSTEA would like to thank Naïs Le Quenellec for her dedication and skills in HPLC analysis.

VIII. FUNDING SOURCE

S. Piveteau is the beneficiary of a PhD scholarship funded equally between Irstea and the Brittany Region. This PhD is realized within the framework of the Valodim project financed by BPIFRANCE PSPC call 2013.

87

IX. REFERENCES

Abma, W.R., Driessen, W., Haarhuis, R., van Loosdrecht, M.C.M., 2010. Upgrading of sewage treatment plant by sustainable and cost-effective separate treatment of industrial wastewater. Water Science and Technology, 61(7), 1715-1722.

APHA. 1998. Standard methods for the examination of water and wastewater. American Public Health Association, Washington, DC.

Ashley, K., Cordell, D., Mavinic, D., 2011. A brief history of phosphorus: From the philosopher’s stone to nutrient recovery and reuse. Chemosphere, 84(6), 737-746.

Battistoni, P., De Angelis, A., Pavan, P., Prisciandaro, M., Cecchi, F., 2001. Phosphorus removal from a real anaerobic supernatant by struvite crystallization. Water Research, 35(9), 2167-2178.

Braak, E., Auby, S., Piveteau, S., Guilayn, F., Daumer, M.-L., 2016. Phosphorus recycling potential assessment by a biological test applied to wastewater sludge. Environmental Technology, 37(11), 1398-1407.

Burns, R.T., Moody, L.B., Walker, F.R., Raman, D.R., 2001. Laboratory and in-situ reductions of soluble phosphorus in swine waste slurries. Environmental Technology, 22(11), 1273-1278.

Buxton, D.R., Muck, R.E., Harrison, J.H., 2003. Silage science and technology. American Society of Agronomy, Crop Science Society of America, Soil Science Society of America, Madison, WI.

Capdevielle, A., Sykorova, E., Beline, F., Daumer, M.L., 2016. Effects of organic matter on crystallization of struvite in biologically treated swine wastewater. Environmental Technology, 37(7), 880-892.

Capdevielle, A., Sýkorová, E., Biscans, B., Béline, F., Daumer, M.-L., 2013. Optimization of struvite precipitation in synthetic biologically treated swine wastewater—Determination of the optimal process parameters. Journal of Hazardous Materials, 244–245(0), 357-369.

Carpenter, J., Pritchard, D., Whaley, R., 1969. Observations of eutrophication and nutrient cycles in some coastal plain estuaries. in: Eutrophication: causes, consequences, correctives., National Academy of Sciences. Washington, DC, pp. 210-221.

Çelen, I., Buchanan, J.R., Burns, R.T., Bruce Robinson, R., Raj Raman, D., 2007. Using a chemical equilibrium model to predict amendments required to precipitate phosphorus as struvite in liquid swine manure. Water Research, 41(8), 1689-1696.

Choi, I.S., Lee, Y.G., Khanal, S.K., Park, B.J., Bae, H.-J., 2015. A low-energy, cost-effective approach to fruit and citrus peel waste processing for bioethanol production. Applied Energy, 140(0), 65-74.

Christensen, M.L., Hjorth, M., Keiding, K., 2009. Characterization of pig slurry with reference to flocculation and separation. Water Research, 43(3), 773-783.

Cohen, A., Zoetemeyer, R.J., van Deursen, A., van Andel, J.G., 1979. Anaerobic digestion of glucose with separated acid production and methane formation. Water Research, 13(7), 571-580.

Comité régional porcin de Bretagne. 2013. The pig industry in Brittany, key figures 2013.

Cordell, D., Drangert, J.O., White, S., 2009. The story of phosphorus: Global food security and food for thought. Global Environmental Change, 19(2), 292-305.

88

Cullen, N., Baur, R., Schauer, P., 2013. Three years of operation of North America's first nutrient recovery facility. Water Science and Technology, 68(4), 763-768.

Dareioti, M.A., Kornaros, M., 2014. Effect of hydraulic retention time (HRT) on the anaerobic co- digestion of agro-industrial wastes in a two-stage CSTR system. Bioresource Technology, 167, 407- 415.

Daumer, M.-L., Béline, F., Spérandio, M., Morel, C., 2008. Relevance of a perchloric acid extraction scheme to determine mineral and organic phosphorus in swine slurry. Bioresource Technology, 99(5), 1319-1324.

Daumer, M., Béline, F., Guiziou, F., 2005. Traitement biologique des lisiers de porcs en boues activées. Guide technique à l’usage des concepteurs, exploitants et organismes de contrôle des stations. Cemagref Editions, Jouve, Paris.

Daumer, M.L., Picard, S., Saint-Cast, P., Dabert, P., 2010. Technical and economical assessment of formic acid to recycle phosphorus from pig slurry by a combined acidification–precipitation process. Journal of Hazardous Materials, 180(1–3), 361-365.

Dehority, B.A., Johnson, R.R., Bentley, O.G., Moxon, A., 1958. Studies on the metabolism of valine, proline, leucine and isoleucine by rumen microorganisms in vitro. Archives of Biochemistry and Biophysics, 78(1), 15-27.

Di Iaconi, C., Pagano, M., Ramadori, R., Lopez, A., 2010. Nitrogen recovery from a stabilized municipal landfill leachate. Bioresource Technology, 101(6), 1732-1736.

Droste, R.L., 1997. Theory and practice of water and wastewater treatment. John Wiley & Sons Incorporated.

Dunière, L., Sindou, J., Chaucheyras-Durand, F., Chevallier, I., Thévenot-Sergentet, D., 2013. Silage processing and strategies to prevent persistence of undesirable microorganisms. Animal Feed Science and Technology, 182(1–4), 1-15.

Elsden, S.R., Hilton, M.G., 1979. Amino acid utilization patterns in clostridial taxonomy. Archives of Microbiology, 123(2), 137-141.

Guilayn, F., Braak, E., Piveteau, S., Daumer, M.-L., 2017. Sequencing biological acidification of waste-activated sludge aiming to optimize phosphorus dissolution and recovery. Environmental Technology, 38(11), 1399-1407.

Gujer, W., Zehnder, A.J.B., 1983. Conversion Processes in Anaerobic Digestion. Water Science and Technology, 15(8-9), 127-167.

Hjorth, M., Fernandez, M., Jayaram, S., Sørensen, J.A., Adamsen, A.P.S., 2015. Bio-acidification of manure–By supplying manure with 2-3% sugar or cellulose. RAMIRAN 2015–16th International Conference. TuTech Innovation GmbH, Hamburg. pp. 96-96.

Jaffer, Y., Clark, T.A., Pearce, P., Parsons, S.A., 2002. Potential phosphorus recovery by struvite formation. Water Research, 36(7), 1834-1842.

Johns, A.T., 1952. The Mechanism of Propionic Acid Formation by Clostridium propionicum. Journal of General Microbiology, 6(1-2), 123-127.

Kalia, V., Kumar, A., Jain, S., Joshi, A., 1992. Biomethanation of plant materials. Bioresource technology, 41(3), 209-212.

89

Kempton, A.G., Clemente, C.L.S., 1959. Chemistry and Microbiology of Forage-Crop Silage. Applied Microbiology, 7(6), 362-367.

Kuchta, R.D., Abeles, R.H., 1985. Lactate reduction in Clostridium propionicum. Purification and properties of lactyl-CoA dehydratase. Journal of Biological Chemistry, 260(24), 13181-13189.

Landrain, B., Congy, E., Loussouarn, A., Dezat, E., Defrance, P., Calvar, C., Lacocquerie, M., Guillaume, A., Decoopman, B., Losq, G., 2013. Le phosphore: comprendre, maîtriser, valoriser, Chambres d'agriculture de Bretagne.

Levasseur, P., Lemaire, N., 2006. Etat des lieux du traitement des lisiers de porc en France. Techni Porc, 29(1), 29-31.

Liu, D., Liu, D., Zeng, R.J., Angelidaki, I., 2006. Hydrogen and methane production from household solid waste in the two-stage fermentation process. Water Research, 40(11), 2230-2236.

MacDonald, G.K., Bennett, E.M., Potter, P.A., Ramankutty, N., 2011. Agronomic phosphorus imbalances across the world's croplands. Proceedings of the National Academy of Sciences, 108(7), 3086-3091.

Martinez, J., Dabert, P., Barrington, S., Burton, C., 2009. Livestock waste treatment systems for environmental quality, food safety, and sustainability. Bioresource Technology, 100(22), 5527-5536.

McCarty, P.L., McKinney, R.E., 1961. Volatile acid toxicity in anaerobic digestion. Journal Water Pollution Control Federation, 223-232.

McDonald, P., 1982. Silage fermentation. Trends in Biochemical Sciences, 7(5), 164-166.

Mehta, C.M., Batstone, D.J., 2013. Nutrient solubilization and its availability following anaerobic digestion. Water Science and Technology, 67(4), 756-763.

Nelson, N.O., Mikkelsen, R.L., Hesterberg, D.L., 2003. Struvite precipitation in anaerobic swine lagoon liquid: effect of pH and Mg:P ratio and determination of rate constant. Bioresource Technology, 89(3), 229-236.

Parawira, W., Read, J.S., Mattiasson, B., Björnsson, L., 2008. Energy production from agricultural residues: High methane yields in pilot-scale two-stage anaerobic digestion. Biomass and Bioenergy, 32(1), 44-50.

Rahman, M.M., Salleh, M.A.M., Rashid, U., Ahsan, A., Hossain, M.M., Ra, C.S., 2014. Production of slow release crystal fertilizer from wastewaters through struvite crystallization – A review. Arabian Journal of Chemistry, 7(1), 139-155.

Remy, M., Kruit, J., Hendrickx, T., Haarhuis, R., van Loosdrecht, M., 2013. Phospaq: Full scale experience with phosphorus recovery via controlled struvite precipitation. Proceedings of the Water Environment Federation, 2013(4), 311-317.

Ren, J., Yuan, X., Li, J., Ma, X., Zhao, Y., Zhu, W., Wang, X., Cui, Z., 2014. Performance and microbial community dynamics in a two-phase anaerobic co-digestion system using cassava dregs and pig manure. Bioresource Technology, 155, 342-351.

Rogosa, M., 1971. Transfer of Peptostreptococcus elsdenii Gutierrez et al. to a New Genus, Megasphaera [M. elsdenii (Gutierrez et al.) comb. nov.]. International Journal of Systematic and Evolutionary Microbiology, 21(2), 187-189.

90

Rooke, J.A., Hatfield, R.D., 2003. Biochemistry of ensiling. in: Silage Science and Technology, (Eds.) D.R. Buxton, R.E. Muck, J.H. Harrison, Vol. 42, pp. 95-139.

Schievano, A., Tenca, A., Scaglia, B., Merlino, G., Rizzi, A., Daffonchio, D., Oberti, R., Adani, F., 2012. Two-stage vs single-stage thermophilic anaerobic digestion: comparison of energy production and biodegradation efficiencies. Environmental Science and Technology, 46(15), 8502-8510.

Sharpley, A., Moyer, B., 2000. Phosphorus forms in manure and compost and their release during simulated rainfall. Journal of Environmental Quality, 29(5), 1462-1469.

Snell-Castro, R., Godon, J.-J., Delgenès, J.-P., Dabert, P., 2005. Characterisation of the microbial diversity in a pig manure storage pit using small subunit rDNA sequence analysis. FEMS Microbiology Ecology, 52(2), 229-242.

Song, Y.-H., Qiu, G.-L., Yuan, P., Cui, X.-Y., Peng, J.-F., Zeng, P., Duan, L., Xiang, L.-C., Qian, F., 2011. Nutrients removal and recovery from anaerobically digested swine wastewater by struvite crystallization without chemical additions. Journal of Hazardous Materials, 190(1), 140-149.

Tenca, A., Schievano, A., Perazzolo, F., Adani, F., Oberti, R., 2011. Biohydrogen from thermophilic co-fermentation of swine manure with fruit and vegetable waste: Maximizing stable production without pH control. Bioresource Technology, 102(18), 8582-8588.

Tsukahara, T., Koyama, H., Okada, M., Ushida, K., 2002. Stimulation of butyrate production by gluconic acid in batch culture of pig cecal digesta and identification of butyrate-producing bacteria. Journal of Nutrition, 132(8), 2229-2234.

Uçkun Kiran, E., Trzcinski, A.P., Ng, W.J., Liu, Y., 2014. Bioconversion of food waste to energy: A review. Fuel, 134(0), 389-399.

Vanotti, M.B., Dube, P.J., Szogi, A.A., García-González, M.C., 2017. Recovery of ammonia and phosphate minerals from swine wastewater using gas-permeable membranes. Water Research, 112, 137-146.

Wallace, R.J., 1978. Control of Lactate Production by Selenomonas ruminantium: Homotropic Activation of Lactate Dehydrogenase by Pyruvate. Journal of General Microbiology, 107(1), 45-52.

Wood, B.J., 2012. Microbiology of fermented foods. Springer Science & Business Media.

Woolford, M.K., Pahlow, G., 1997. The silage fermentation. in: Microbiology of Fermented Foods, (Ed.) B.J.B. Wood, Springer US. Boston, MA, pp. 73-102.

Wu, X., Yao, W., Zhu, J., 2010. Effect of pH on continuous biohydrogen production from liquid swine manure with glucose supplement using an anaerobic sequencing batch reactor. International Journal of Hydrogen Energy, 35(13), 6592-6599.

Ye, Z.-L., Chen, S.-H., Lu, M., Shi, J.-W., Lin, L.-F., Wang, S.-M., 2011. Recovering phosphorus as struvite from the digested swine wastewater with bittern as a magnesium source. Water Science and Technology, 64(2), 334-340.

Zhang, T., Bowers, K.E., Harrison, J.H., Chen, S., 2010. Releasing phosphorus from calcium for struvite fertilizer production from anaerobically digested dairy effluent. Water Environment Research, 82(1), 34-42.

91

X. APPENDIX

1. APPENDIX A

Net production of VFAs and lactate, sucrose removed and pH across time in R10. Standard deviation shown for the sucrose removed and the sum of VFAs and lactate.

2. APPENDIX B

Net production of VFAs and lactate, sucrose removed and pH across time in R20.

92

3. APPENDIX C

Net production of VFAs and lactate, sucrose removed and pH across time in R40.

4. APPENDIX D

Net production of VFAs and lactate, sucrose removed and pH across time in R60.

93

CHAPTER 2 BIOLOGICAL ACIDIFICATION OF SWINE SLURRY: EFFECT OF VARIOUS ORGANIC CO-SUBSTRATES ON PH, ORGANIC ACIDS PRODUCTION AND BACTERIAL POPULATIONS

Piveteau Simon, Le Roux Sophie, Dabert Patrick and Daumer Marie-Line. (2017).

94

I. SUMMARY OF CHAPTER 2

Les travaux précédents réalisés avec du saccharose ont montré que la production d’acide lactique, métabolite presque exclusif issu de la dégradation des sucres simples était déterminante pour acidifier le lisier biologiquement. L’étude bibliographique laisse à penser que cette voie n’est pas toujours majoritaire lors de la dégradation de substrats complexes comme les déchets utilisés en méthanisation. Le choix des substrats et la flore initialement présente peuvent également déterminer les voies métaboliques actives lors du procédé. Les travaux présentés dans ce chapitre ont pour objectif de vérifier que l’acidification peut être obtenue avec des sucres complexes et de tester l’influence de la nature de la flore présente au début du procédé en comparant les résultats obtenus avec du sucre ajouté au lisier brut et à du digestat de lisier.

Un lisier brut et deux digestats issus du même digesteur mais prélevés à deux périodes différentes ont été utilisés (Table 6). Une première série d’expérimentation a été conduite avec le lisier brut et le digestat 1 auxquels ont été ajoutés 5 concentrations de sucre différentes de 10 à 50 g.L-1. Trois co- substrats complexes (pommes de terre, déchets de biscuits et déchets de farine de blé noir) ont également été testés à une concentration de 50 g de MV.L-1 avec le digestat 1.

Dans la deuxième série d’expériences 8 nouveaux co-substrats ont été testés à la même concentration avec le digestat 2, les déchets de biscuits et de farine de blé noir stockés à 4 °C ainsi que des carottes, des haricots verts, des petits pois de la macédoine de la paille de blé et des pommes. Un dernier essai avec une concentration moindre de carottes (30 g.L-1) a également été effectué. La composition en glucides, lipides et protéines et en sucres sont présentées dans la Table 7.

Avec le digestat, le pH décroît presque linéairement jusqu’à des concentrations de sucre de 30 g.L-1. La baisse est moins importante à quantité de sucre équivalente avec le lisier brut et 50 g.L-1 sont nécessaires pour obtenir la valeur minimale et ce malgré un pH initial plus faible (Figure 28). La présence d’AGV dans le lisier avant le test peut expliquer à la fois le pH plus faible et le pouvoir tampon plus élevé. Le lactate est l’acide organique majoritaire pour les fortes concentrations en sucre dans les deux matrices utilisées. Pour des concentrations inférieures, le lactate a été produit pendant les premières 48 heures mais consommé ensuite lors de fermentations secondaires qui ont pu avoir lieu grâce à un pH insuffisamment acide pour les inhiber.

Un pH inférieur à 5 est obtenu avec les pommes, les carottes à 50g.L-1, les biscuits et la farine de blé noir stockés. Cette baisse de pH peut être attribuée à la production d’acide lactique (Figure 29). Aucune baisse de pH ni production d’AGV n’est observée pour la paille. Pour les autres co-substrats, le pH atteint des valeurs comprises entre 5 et 5.7. Les valeurs atteintes sont corrélées à la teneur en glucides. La conversion de la matière organique en acides organiques est comprise entre 25 et 56% inférieures aux valeurs obtenues avec le sucre (44-78%).

95

Les communautés microbiennes impliquées dans l’acidification ont été caractérisées par séquençage haut débit des gènes des ARNr 16S. L’analyse des indices de diversité suggère une large diversité initiale avec quelques espèces dominantes. La diversité se réduit pour les essais pour lesquels les pH les plus bas sont atteints. Le faible pH et la concentration en lactate exerceraient une pression de sélection ne permettant de conserver que les micro-organismes tolérants aux conditions acides.

L’analyse statistique de la structure des communautés microbiennes montre une différence importante entre le lisier brut et le digestat mais cette structure évolue pendant l’acidification et se rapproche d’autant plus à la fin des essais que la concentration en sucre est importante. Dans tous les échantillons les phyla majoritaires sont les Firmicutes et Bacteroidetes qui sont les phyla classiquement retrouvés dans les réacteurs anaérobies acidogènes. Dans tous les réacteurs l’acidification est associée à une baisse de l’abondance relative des Bacteroidetes et une augmentation des Firmicutes. Lactobacillus domine pour les concentrations en sucre les plus élevées tandis que l’adaptation des Clostridiales aux plus faibles concentrations serait due à leur capacité à croître sur le lactate. Dans les essais digestats et déchets organiques, Lactobacillus domine systématiquement lorsque le contenu en sucre facilement dégradable est important comme dans les carottes, les pommes ou les biscuits et sarrasin dégradés. Les Clostridiales dominent dans les échantillons moins dégradés indiquant une meilleure capacité à l’hydrolyse. Streptococcus est présent de façon significative dans la plupart des échantillons indiquant une capacité à agir et se développer sur des substrats plus divers ou en conditions limitantes. Il est capable de modifier son métabolisme passant de la fermentation homolactique à une fermentation produisant divers acides mais il est peu tolérant au milieu acide ce qui explique qu’il ne soit pas dominant dans les échantillons avec les plus fortes concentrations initiales en sucre dans lesquels le pH est inférieur à 5.

Ces essais montrent que la fermentation lactique avec des déchets organiques est possible et que la production d’acide lactique dépendra de la disponibilité de sucres facilement dégradables, qui favorisent le développement rapide des Lactobacillus, avec un effet inhibiteur empêchant les fermentations secondaires.

Afin de prédire la quantité de substrats à ajouter d’après l’équation décrite dans le chapitre précédent pour obtenir le pH optimal en vue du recyclage du P sous forme de struvite, un équivalent sucre sera calculé pour quelques-uns des co-substrats testés. Un test sur un réacteur semi-continu permettra de voir s’il est possible d’adapter la flore pour augmenter l’efficacité de la dégradation des glucides complexes et de la fermentation lactique.

96

Biological acidification of swine slurry: effect of various organic co-substrates on pH, organic acids production and bacterial populations

Piveteau Simona,b,*, Le Roux Sophiea,b, Dabert Patricka,b, Daumer Marie-Linea,b a Irstea, UR OPAALE, 17 Avenue de Cucillé-CS 64427, F-35044 Rennes, France b Univ de Bretagne Loire, France.

*Corresponding author. Simon Piveteau. UR OPAALE, Irstea. E-mail address: [email protected]

97

II. ABSTRACT

Swine manure contains large amounts of phosphorus, ammonium and magnesium, making this waste stream suitable for P-recovery as struvite, a slow release P-fertilizer. Phosphorus has to be dissolved through acidification to become available for struvite precipitation. Biological acidification of swine slurry under anaerobic conditions (i.e. acidogenesis) using sucrose as organic co-substrate led to a dissolution of as much as 90% of Total-P thanks to a high production of lactic acid. In this study, several organic waste and sucrose were tested as organic co-substrate for the biological acidification of raw and digested swine slurry in batch tests. Lactic acid fermentation occurred when the co-substrate had a high content in easily accessible carbohydrates and when a sufficient amount of waste was added to the slurry (50g-VS/L). A microbiological analysis indicated that acidogenesis led a significant reduction in bacterial diversity. The dominating clusters at the end of acidification were Clostridiales and Streptococcus when moderate acidification occurred and Lactobacillus when pH dropped below 5. The successful acidification of digested and raw pig manure when using easily biodegradable, carbohydrate rich organic waste is promising for developing an inexpensive, low-tech recovery process for struvite recovery.

98

III. INTRODUCTION

The green revolution during the 20th century enabled massive increase in crop yield and food production, thus sustaining a large population growth during the same period (Tilman, 1998). This achievement relied heavily on generous application of chemical fertilizer providing N, P and K to the crops. Because P is the main limiting nutrient in freshwater ecosystems (Mainstone and Parr, 2002; Sattari et al., 2014), a small increase in concentration due to runoff and leaching from agricultural fields can lead major ecological perturbations known as eutrophication (Schindler et al., 2008; Sharpley et al., 2005). While the treatment of domestic wastewater vastly improved in the last 50 years, diffuse pollution by agriculture becomes more prevalent (Dupas et al., 2015; Némery et al., 2005). More recently, the realization that the source of P for fertilizer manufacture was limited, more contaminated in cadmium and other heavy metals(Smil, 2000) and concentrated in very few countries, led to a major shift in evaluating the question of phosphorus. Indeed, P chemical fertilizer is the main industry using phosphate rock (80%), and tits use efficiency is very limited, since only 20% of extracted P is actually consumed as food (Cordell et al., 2009). A major cause of inefficient use of P is the localized excess of livestock manure (Senthilkumar et al., 2012). In highly specialized agricultural areas, livestock farmers struggle to find an output for their manure (Schipanski and Bennett, 2012). The amount of manure produced is exceeding by far the needs of local crops, and transportation of manure to long distance is both economically unprofitable and damaging to the environment. Actually, phosphorus could be recovered from animal manure as struvite, a highly concentrated and slow released P fertilizer that could be exported and sold to agricultural regions specialized in crop production, reducing their need for chemical P fertilizer (Daumer et al., 2010). Pig slurry contains a significant amount of P, 200-1500mg/L (Karakashev et al., 2008; Suzuki et al., 2005), but mostly as inorganic solids (Christensen et al., 2009). Several processes for P recovery from animal manure have been tested on lab scale (Cerrillo et al., 2015; Perera et al., 2007; Perera et al., 2009) and full scale (Schuiling and Andrade, 1999; Vanotti et al., 2007), however most of them focus on already dissolved phosphorus. As a result, the recovery rate cannot excess 10 to 30% (Egle et al., 2015). Chemicals such as hydrochloric acid (Zhang et al., 2012), formate (Daumer et al., 2010) and citrate (Szögi et al., 2015) have proven efficient to dissolve P. However, these chemicals represent an important cost and make the process economically uncompetitive (Egle et al., 2016; Molinos-Senante et al., 2011). They are also dangerous when used for farmers and the environment. Biological acidification of pig slurry, i.e. the natural production of organic acids by endogenous acidogenic bacteria under anaerobic conditions, has shown promising results when sucrose was used as organic co-substrate. Up to 90% of TP was dissolved at pH 4-5.5 with 30 to 60g-sucrose per liter of raw slurry (Piveteau et al., 2017). The main organic acid produced was lactate, when pH dropped low and fast enough. Lactic acid fermentation presents several advantages regarding pig slurry acidification for P recovery: lactate has a low pKa compared to other organic acid, meaning that a limited amount of co-substrate could be sufficient to

99 reach the desired pH, and the proportion of undissociated acid after acidification is lowered compared to usual VFA. As a result the amount of alkaline compound needed thereafter in the process to increase the pH and favor struvite precipitation would be reduced. In this paper, several organic waste as well as sucrose were tested as organic co-substrate for biological acidification of raw swine slurry and digested swine slurry. pH, amount and type of organic acids produced were investigated and correlated to the waste composition. Microbial community analysis was performed in order to identify the bacterial populations responsible for organic acid production and understand how to favor lactic acid bacteria during biological acidification of pig slurry.

IV. MATERIALS AND METHODS

1. CHARACTERIZATION OF PIG SLURRY AND DIGESTED PIG SLURRY

Raw swine slurry was collected from a pig fattening farm in Brittany (France). Digested swine slurry (digestate) was withdrawn at two different dates from a 87 L laboratory digester treating liquid swine slurry supplemented with horse feed (Table 6).

Table 6. Characteristics of pig slurry and digested slurries (standard deviation displayed between brackets)

TS VS TSS VSS NH4-N pH g/L

digestate 1 19.9 (0.0) 10.1 (0.1) 10.4 (0.1) 6.5 (0.0) 1.3 (0.0) 7.86

digestate 2 22.1 (0.3) 10.2 (0.1) 10.9 (0.2 7.1 (0.1) 1.4 (0.0) 7.99 raw slurry 58.0 (2.2) 41.4 (2.0) 41.9 (3.9) 33.9 (3.0) 2.5 (3.0) 6.73

2. BATCH EXPERIMENTS

Batch acidification experiments using sucrose as organic co-substrate were conducted on raw swine slurry and digestate 1 using 1 L bottles inoculated with 400 mL of raw slurry or digestate 1 and five different sucrose concentrations (10, 20, 30, 40, 50 g/ L-slurry). The batch bottle for raw slurry and 10g-sucrose/L was mishandled and broke, the partial results obtained were not included. Batch acidification test were also conducted on digestate 1 with the addition of potato, grounded biscuit and buckwheat waste at 50g-MV per liter of digestate. The bottles were sealed once inoculated, and placed in a thermostatic shaker, at 120 rpm and 38°C. Purging oxygen from the head space was not done since it had been shown to have no impact on acidification (Piveteau et al., 2017). The batch test ran for four days with daily sample collection for pH and organic acid analysis.

A second set of batch tests were conducted using digestate 2 and different organic waste (Table 7) at 50g-MV per liter of digestate. In addition of carrot, bean, pea, macédoine (diced mixed vegetables), wheat straw and apple, the same buckwheat flour and biscuit waste were tested again (hereby referred as “buckwheat flour 2” and “biscuit 2”) , after 3 months of storage at 4°C in order to evaluate the effect of aging and degradation of waste. An additional batch test was conducted using a lower

100 concentration of carrot, 30 g-MV/L instead of 50. All batch tests were done in duplicate and gave very similar results (not shown).

A sample of liquid was taken at the end of the batch test, centrifuged at 20,000g for 20 minutes and the pellet was frozen immediately at -20°C for microbiological analysis.

Table 7. Composition of the organic waste used during biological acidification of pig slurry. Standard deviation of triplicate measures for TS and VS displayed between brackets.

TS VS Carbohydrates* Sugars* Proteins* Lipids*

g/L % dry weight

potato 192.3 (0.2) 175.1 (0.2) 76 2 10 1

apple 145.4 (0.3) 139.8 (0.1) 75 75 2 1

macédoine 161.8 (2.7) 114.8 (3.0) 50 18 14 6

wheat straw 893.4 (4.3) 833.4 (3.3) 87 2 3 7

carrot 83.9 (3.0) 79.1 (2.8) 66 49 8 3

bean 109.5 (4.6) 101.8 (4.4) 51 23 14 2

pea 205.1 (1.9) 198.7 (2.0) 40 16 25 3

buckwheat flour 852.1 (2.2) 762.2 (1.7) 71 2.6 13 3

biscuits 857.3 (2.1) 769.5 (1.4) 59 32 7 2

*average values from literature http://www.aprifel.com/profils-des-fruits-et-legumes.php, https://ndb.nal.usda.gov/ndb/foods/show/6482

3. PHYSIC AND CHEMICAL ANALYSES pH was measured using a WTW probe right after each sample collection. TS, VS, TSS and VSS were measured in triplicates on the raw slurry with standard methods (APHA, 1998). Supernatants were obtained after 20 min of centrifugation (4°C, 20,000 g) and filtration on a 0.45 mm polypropylene membrane. Sucrose, glucose, fructose, lactic acid and VFA composition of the supernatant was measured by HPLC (Ultimate 3000, Dionex) with a Hiplex-H column (Agilent).

4. MICROBIOLOGICAL ANALYSIS

Total DNA was extracted from about 200 mg of pellet. Extraction was performed using the Macherey Nagel Genomic DNA for Soil kit according to the manufacturer’s instructions. The extracted DNA was eluted in 100 µL of sterile water and stored at -20°C until further analyses. Concentration and purity of extracted DNA were checked by spectrophotometry (ND-1000, NanoDrop Tech.).

Bacterial and archaeal diversity were studied by amplifying the hypervariable region V4–V5 of the 16S rRNA genes using primers 515F (5’-GTGYCAGCMGCCGCGGTA-3’) (Wang et al., 2007) and 928R (5’-CCCCGYCAATTCMTTTRAGT-3’) (Wang and Qian, 2009) as described in detail in Poirier et al. (2016a); Poirier et al. (2016b). The amplified DNA fragments were then sequenced using high throughput sequencing Ion Torrent™ technology and methods (Life Technologies, USA) at the NGS facility of the BIOMIC Team of IRSTEA (Antony, France). Quality filtered sequences were

101 exported as FastQ files (47751 to 382063 reads for each sample). These reads were then quality checked by two amplicon read processing pipelines: QIIME 1.8.0 (Caporaso et al., 2010) and USEARCH v5.2.136 (Edgar, 2013). Sequences with low quality score (<20) and chimeric sequences were removed while the remaining sequences were clustered into Operational Taxonomic Units (OTUs) at 97% sequence similarity using Usearch quality filter reference set (http://www.drive5.com/usearch/) (Edgar, 2010). The longest sequence of each OTU was chosen as representative and used for taxonomic identification using the Ribosomal Database Project (RDP II). Phylogenetic alignment of sequences was done with the PyNAST program with a minimum length of 150 bp and a minimum percent identity of 75.0. Alpha diversity was used to describe the microbial richness, diversity and evenness of each sample using Shannon, Simpson and equitability metrics respectively.

In each sample, OTUs whose number of reads was greater than 1% of the total number of reads were considered as predominant OTUs. Predominant OTUs who had a low affiliation score were further identified by phylogenetic analysis using the ARB software (Ludwig et al., 2004) and SSURef_Nr99 123.1 SILVA 03032016 database. OTUs were aligned and edited against the 59847 sequences of the database using the ARB fast aligner software and fitted in the global phylogenetic tree using the ARB parsimony interactive tool with E. coli and termini filters.

5. STATISTICAL ANALYSIS

The microbial community structure of samples was compared by statistical methods using the open source software R (V3.2.3) (Venables and Smith, 2016) including the vegan package (v2.3-2) (Oksanen et al., 2007). Non-metric multidimensional scaling (nMDS) based on Bray-Curtis distances was conducted using the metaMDS function of vegan. Correlations between pH, lactate concentration and microbial community structures were searched by fitting data of each factor to the nMDS ordination of the microbial community using the envfit function of vegan. The 20 tests run, with 999 permutations each, gave an excellent representativeness value of stress at 0.366994 (Table 8).

Table 8. Statistical results for nMDS analysis

Parameters nMDS 1 nMDS 2 R² Pr (>r) pH -0.55605 0.83115 0.9126 0.006 ** (signification 0.001) Lactacte 0.96560 -0.26004 0.7754 0.010 ** (signification 0.001)

V. RESULTS AND DISCUSSION

1. EVOLUTION OF PH AND ORGANIC ACIDS CONCENTRATION IN BATCH TESTS WITH SUCROSE

Final pH logically decreased with increasing sucrose concentration, almost linearly until 30 g/L but not afterward (Figure 28). pH did not decrease below 4 despite higher initial sucrose concentration. Those results are in line with those of Piveteau et al. (2017). The effect of sucrose on pH was stronger

102 in the digestate, indicating a lower buffer capacity. This lower buffer capacity could be due to the little suspended solid concentration compared to the pig slurry, and the absence of VFA initially while the slurry had 10 g-VFA/L (which could itself explain the low initial pH in the raw slurry, 6.73 vs 7.86)

Lactate was the main organic acid produced after 96 hours in batch tests with 30 g-sucrose/L and higher, while very little to none was measured at 10 and 20 g/L. Homolactic fermentation of sucrose in pig slurry was observed before by Hjorth et al. (2015) and (Piveteau et al., 2017). Homolactic fermentation of glucose also occurred in digested cattle manure used for biohydrogen production (Chatellard et al., 2016). Lactate was produced at 10 and 20 g-sucrose/L but only during the first 48 hours before being converted to other organic acids, leading to a pH re-increase. This secondary fermentation is a common phenomenon in spoiled silage (Oude Elferink et al., 1999), and also occurred in lactic acid fermenters when pH was above 5, leading to the conversion of lactate to acetate, propionate and/or butyrate (Kim et al., 2003; Komemoto et al., 2009; Li et al., 2014; Wu et al., 2015).

The carbon recovery was only partial in the two batch tests series with 44 to 78% for digestate and 67- 97% for raw slurry (Table 9). All sucrose was removed in all the batch reactors except for the one with digestate and 50 g-sucrose/L, where 5.6 g/L remained. The slower pH decrease in the raw slurry and the relatively low initial pH could explain why sucrose was entirely consumed in the batch with raw slurry and not with digestate. Based on the stoichiometric equation for acetate, succinate and butyrate production, a theoretical CO2 production was calculated. The carbon recovery improved for the raw pig slurry with 90-110%, yet not for the digestate. Because the HPLC analysis was not conducted at the same time, lactate concentration in digestate batch test could have been underestimated by the machine due to a calibration issue.

A B

n.a.*

Figure 28. Organic acids produced and pH after 4 days. A: batch tests with digestate 1. B: batch tests with raw slurry. *:not applicable

103

Table 9. Carbon recovery: C-mol organic acids produced / C-mol sucrose consumed (%)

Carbon recovery (%) Initial sucrose concentration digestate raw slurry 10g/L 44 na* 20g/L 75 70 30g/L 78 67 40g/L 72 96 50g/L 67 97 * not applicable

2. EVOLUTION OF PH AND ORGANIC ACID PRODUCTION IN BATCH TESTS WITH ORGANIC WASTE AS CO-SUBSTRATES

No production of organic acid and nor pH change occurred in the batch with wheat straw, indicating the inability of the biomass to degrade it. Similar low degradation of wheat straw was noticed by Chatellard et al. (2016) during co fermentation of digested cattle manure and wheat straw for biohydrogen production. In the other batch tests, pH decreased below 5 only with apple, carrots (at normal 50 g/L VS concentration) biscuit 2 and flour 2, corresponding to the production of lactate, an organic acid with a lower pKa compared to VFA (Figure 29). In the other batch tests, acetic acid was the dominant VFA followed by butyrate and propionate. The conversion of organic matter to organic acids was between 25 and 56% for macédoine and buckwheat flour 2 respectively (Table 10). A positive correlation was found between acidification and the amount of carbohydrates added to each batch test (Figure 30). Interestingly, while the repetition of the test for macédoine gave almost identical result (not shown), the same co-substrates at the same initial concentration in two different batch tests gave very different results in the case of biscuit and buckwheat flour. The two co-substrates had certainly started to degrade when used in the second batch experiment, with more accessible carbohydrates this time, leading to lactic acid fermentation rather than mixed acid. This indicates that predicting the acidifying capacity of a given organic waste is very uncertain depending on the age of the waste and how it has been stored.

In the case of carrots, the two concentrations tested also gave very different results with a lower acidification and the absence of lactic acid in the batch with 30 g-VS/L. Lactate had been produced during the first 48 hours but disappeared afterward, a secondary fermentation similar to that of acidification with sucrose.

104

350 6 300 valerate 250 5 butyrate 200 propionate 150 acetate 4 final pH 100 formate

organicacids (mmol/L) 50 lactate 0 3 succinate malate pH

Figure 29. Final pH and organic acids produced after 4 days

Table 10. Conversion ratio of organic matter to organic acids

Co-substrate organic acids/VS (% g/g) potatoe 41 apple 45 macédoine 25 carrot 50g-VS/L 32 carrot 30g-VS/L 39 bean 32 pea 31 biscuit 1 28 biscuit 2 47 flour 1 31 flour 2 56

Figure 30. Correlation between carbohydrate content and final pH

105

3. MICROBIAL DIVERSITY DETECTION USING 16S RDNA HIGH THROUGHPUT SEQUENCING

The microbial community of all samples was analyzed by 16S rRNA gene high throughput sequencing giving between 47 751 and 382 063 reads for each sample that grouped within a mean of 861 +/- 166 OTUs per sample (minimum number at 583 and maximum at 1135) at 97% similarity. The estimated coverage of the samples diversity was above 98% showing a good representativeness of the data (not shown).

The diversity indexes calculated for the raw slurry and digestate before acidification were above 7.0 and 0.97 for Shannon and Simpson indexes respectively (Figure 31). Their equitability index was 0.7. These values are high but in the range of usual values for these types of microbial communities (Chen et al., 2016; Leite et al., 2016; Lu et al., 2014). They suggest a high microbial diversity with a few dominant species. As shown in Figure 31, the addition of co-substrate induced a drop of the three diversity indexes when accompanied by a large pH decrease. pH values far from neutrality are known to reduce bacterial diversity (Guo et al., 2010). Thus, the co-substrates with easily accessible carbohydrates present in large amounts (sucrose, apple, carrot 50 g-VS/L, buckwheat, flour 2) had a high lactate production, leading to a strong pH drop and decreasing bacterial diversity. This was not the case for co-substrates with fewer carbohydrates or less accessible ones (wheat straw, macédoine, bean, pea). The pH decrease was lower with those co-substrates and bacterial diversity remained stable. It appears likely that low pH and high lactate concentration acted as a selection pressure on bacterial populations, with only acid tolerant microorganisms growing in these conditions.

Figure 31. Comparison of the microbial community diversity indexes (Shannon, Simpson and equitability) and pH of the different samples

Statistical comparison of the samples microbial community structures between themselves and with their corresponding values of pH and lactate was performed using the nMDS method (R vegan package v2.3-2). The nMDS representation showed several correlations (Figure 32). First, the raw slurry samples clustered all away from the digestate samples in the lower right square of the Figure 32, suggesting different global communities between these two types of matrices. Second, 9 out of the 11

106 samples that possessed a pH below 5 clustered in the higher right square of the Figure 32 and faraway from similar samples with higher pH, suggesting that samples with strong acidification undertook similar changes in their microbial communities. This hypothesis is comforted by the clear evolution of the slurry and digestate communities under increasing sucrose addition. While the communities belonged to diametrically opposed location initially, they gradually joined the “lactic acid area”.

Finally, the distribution of the samples shows clear gradients of distribution along 2 axes influenced by the pH and lactate concentration.

Figure 32. nMDS statistical analysis of samples microbial communities

4. IDENTIFICATION OF THE DOMINANT SPECIES

Identification of the OTUs and screening of their representativeness within the community showed that unclassified OTUs represented less than 1% of the sequences of each sample except for one sample where it reached 4% (biscuit 1). Also, the archaea were undetected in the manure samples and represented usually less than 1% of the sequences of each digestate sample except for the sample wheat straw where a methanogenic acetotrophic Methanosaeta reached about 1.7% of the sequences. Thus the majority of the community was made of Bacteria that grouped within 628 OTUs identified up to the family or genus rank.

Figure 33A represents the microbial community structure, at the phylum level, of the 153 OTUs that contained at least 1% of the total sequences collected and that were assumed to represent the dominant

107 species of the communities. In all samples, the dominant phyla were Firmicutes and Bacteroidetes, commonly found in anaerobic/acidogenic reactors (Chen et al., 2016; Fitamo et al., 2017; Leite et al., 2016; Regueiro et al., 2016; Shen et al., 2017). In raw slurry these phyla represented 57% and 23% of the total sequences respectively and were dominated by members of Clostridiaceae (10% of sequences) and Prevotella (7% of sequences) respectively, both microbial groups widely present in swine slurry (Snell-Castro et al., 2005). In the initial digestate, Bacteroidetes was the main phylum with about 37% of the total sequences. It contained several genera with 2 representing 7 and 10% of the sequences that were related to uncultured species retrieved from anaerobic digesters (Accession numbers: AB494402 and CU922250 respectively). The Firmicutes represented 30% of the initial digestate sequences with Clostridium as the most represented genus and sequences close to sequences retrieved from anaerobic digesters (Accession numbers: MF769126, KY224082, GQ135833).

The comparison of all the samples highlights the decisive role of a few dominant species during acidification (Figure 33B). In all reactors, acidification was associated with an increase in the proportion of Firmicutes and a reduction of Bacteroidetes. In the case of slurry with sucrose, Clostridiales increased massively at 20 and 30 g-sucrose/L but were dominated by Lactobacillus at higher sucrose concentrations. The dominant species of Lactobacillus were L. amylophylus and L. amylovorus, both strict homofermentative lactic acid bacteria (LAB) (Mercier et al., 1992; Weinberg et al., 1998), as reflected by the high proportion of lactate in the batch test with high initial sucrose concentration. These LAB species were discovered in the fermentation broth of swine wastewater and corn (Nakamura and Crowell, 1979) and in the fermentation broth of cattle waste and corn respectively (Nakamura, 1981). In the batch test with digestate and sucrose, Clostridiales remained stable at any sucrose concentration while both Lactobacillus and Streptococcus increased massively with sucrose concentration, Streptococcus at 10 and 20 g/L, Lactobacillus at 30 and 40 g/L. Thus, it appears that Clostridiales and Lactobacilliales were able to compete successively for sucrose compared to Bacteroidetes and other bacteria, but also competed with each other, Lactobacillus dominating at higher initial sucrose concentration and associated low pH. The success of Clostridiales at moderate sucrose concentration was likely due to their capacity to grow on lactate.

In the batch tests with digestate and organic waste, Lactobacillus systematically dominated when the co-substrate had a high content in easily accessible carbohydrates, e.g. apple, carrot (at high initial concentration, biscuit 2 and buckwheat flour 2). In contrast, Clostridiales performed better than Lactobacillus in biscuit 1 and buckwheat flour 1, indicating a better capacity for hydrolysis. Streptococcus was significantly present in most samples, indicating a capacity to function and grow with more diverse or more limiting substrate than Lactobacillus. Indeed, Lactobacillus was absent in the batch with low concentration of carrot, while Streptococcus represented 20% of the population. It has been documented that Streptococcus is able to grow on a large variety of carbohydrates (Holzapfel and Wood, 2012) and shift its metabolic pathway from homolactic fermentation to mixed acid under

108 limiting substrate concentration (Thomas et al., 1979). However, Streptococcus has a very low tolerance to acidic condition (Axelsson and Ahrné, 2000), which would explain its inability to dominate at high initial sucrose concentration where pH fell below 5.

Among the Bacteroidetes, Porphyromonadaceae were able to increase their share of the population from 6% initially to 30-40% in macédoine and biscuit 1, and to at least 10% in all the batch tests at pH above 5. This family of bacteria is strictly anaerobic and originates from the digestive tract of mammal, hence its presence in raw and digested swine slurry. Moreover, they have been observed in several food waste anaerobic digesters (Fitamo et al., 2017; He et al., 2017; Jang et al., 2016; Leite et al., 2016; Chen et al., 2016). They are able to grow on various carbohydrates and produce mainly butyric acid (Sakamoto et al., 2009). These bacteria were found in significant amounts during the fermentation of swine slurry at 35°C with initial pH adjustment to 10 (Lin et al., 2015).

This microbiological analysis demonstrated that even though Clostridiaceae were a major cluster of bacterial population in all the batch tests, they did not systematically dominate as it is often the case during acidogenic fermentation (Feng et al., 2009; Ren et al., 2007). High initial sucrose concentration and easily degradable carbohydrate rich co-substrates enabled LAB and especially Lactobacillus to outcompete other bacteria for substrate and remain dominant thanks to its tolerance for acidic pH (Holzapfel and Wood, 2012) and possibly the production of inhibiting substances such as undissociated lactic acid and bacteriocins (Monlau et al., 2013; Noike et al., 2002).

109

A

B

Figure 33. A: community structures and corresponding value of pH the swine slurry and digestate samples at the phylum taxonomic rank. B: inclusion of dominant Orders (Porphyromonadaceae), Families (Clostridiales, Lactobacilliales) and Genera (Lactobacillus, Streptococcus)

VI. CONCLUSION

Strong acidification of raw swine slurry and digested swine slurry was obtained using sucrose and organic waste as co-substrate for biological acidification under anaerobic conditions, i.e. acidogenesis. Production of lactic acid occurred at high initial sucrose concentration (30 g/L and above) and with organic waste rich in easily biodegradable carbohydrates (apple, carrot). pH drop below 5 only occurred when lactic acid was the main product of acidogenesis, as a result of its lower pKa compared to VFAs. The use of recent and aged organic waste (buckwheat flour and biscuit) led to very different results, with lactic fermentation occurring only with the aged sample of waste, likely due to a beginning of hydrolysis during storage, resulting in more accessible carbohydrates and thus favoring LAB. Microbiological analysis indicated a decrease in diversity associated with lower pH. Firmicutes increased their dominance with increasing sugar concentration, favoring Clostridiales and

110

Streptococcus at medium concentration and Lactobacillus at high initial concentration. Despite quite different initial microbial community compositions between raw and digested slurry, the presence of increasing amounts of easily biodegradable carbohydrates systematically shifted the bacterial populations toward a homogenous, Lactic Acid Bacteria rich community, confirming other lactic acid fermentation studies using undefined mixed culture as inoculum (Dreschke et al., 2015; Liang et al., 2015; Probst et al., 2015). These batch tests and microbiological results indicate that lactic acid fermentation of raw or digested swine slurry with cheap/free organic waste is easily achievable if the waste is rich enough in easily accessible carbohydrates. The organic acid rich supernatant, after P recovery, could be used as feed for anaerobic digestion and energy production from methane.

111

VII. REFERENCES

APHA. 1998. Standard methods for the examination of water and wastewater. American Public Health Association, Washington, DC. Axelsson, L., Ahrné, S., 2000. Lactic acid bacteria. in: Applied microbial systematics, Springer, pp. 367-388. Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., Fierer, N., Peña, A.G., Goodrich, J.K., Gordon, J.I., 2010. QIIME allows analysis of high-throughput community sequencing data. Nature methods, 7(5), 335-336. Cerrillo, M., Palatsi, J., Comas, J., Vicens, J., Bonmatí, A., 2015. Struvite precipitation as a technology to be integrated in a manure anaerobic digestion treatment plant–removal efficiency, crystal characterization and agricultural assessment. Journal of Chemical Technology and Biotechnology, 90(6), 1135-1143. Chatellard, L., Trably, E., Carrère, H., 2016. The type of carbohydrates specifically selects microbial community structures and fermentation patterns. Bioresource Technology, 221(Supplement C), 541- 549. Chen, S., Cheng, H., Wyckoff, K.N., He, Q., 2016. Linkages of Firmicutes and Bacteroidetes populations to methanogenic process performance. Journal of Industrial Microbiology & Biotechnology, 43(6), 771-781. Christensen, M.L., Hjorth, M., Keiding, K., 2009. Characterization of pig slurry with reference to flocculation and separation. Water Res., 43(3), 773-783. Cordell, D., Drangert, J.O., White, S., 2009. The story of phosphorus: Global food security and food for thought. Glob. Environ. Chang., 19(2), 292-305. Daumer, M.L., Picard, S., Saint-Cast, P., Dabert, P., 2010. Technical and economical assessment of formic acid to recycle phosphorus from pig slurry by a combined acidification–precipitation process. J. Hazard. Mater., 180(1–3), 361-365. Dreschke, G., Probst, M., Walter, A., Pümpel, T., Walde, J., Insam, H., 2015. Lactic acid and methane: Improved exploitation of biowaste potential. Bioresource Technology, 176(Supplement C), 47-55. Dupas, R., Delmas, M., Dorioz, J.-M., Garnier, J., Moatar, F., Gascuel-Odoux, C., 2015. Assessing the impact of agricultural pressures on N and P loads and eutrophication risk. Ecological Indicators, 48(Supplement C), 396-407. Edgar, R.C., 2010. Search and clustering orders of magnitude faster than BLAST. Bioinformatics, 26(19), 2460-2461. Edgar, R.C., 2013. UPARSE: highly accurate OTU sequences from microbial amplicon reads. Nature methods, 10(10), 996-998. Egle, L., Rechberger, H., Krampe, J., Zessner, M., 2016. Phosphorus recovery from municipal wastewater: An integrated comparative technological, environmental and economic assessment of P recovery technologies. Science of The Total Environment, 571(Supplement C), 522-542. Egle, L., Rechberger, H., Zessner, M., 2015. Overview and description of technologies for recovering phosphorus from municipal wastewater. Resources, Conservation and Recycling, 105(Part B), 325- 346. Feng, L., Chen, Y., Zheng, X., 2009. Enhancement of Waste Activated Sludge Protein Conversion and Volatile Fatty Acids Accumulation during Waste Activated Sludge Anaerobic Fermentation by Carbohydrate Substrate Addition: The Effect of pH. Environmental Science & Technology, 43(12), 4373-4380.

112

Fitamo, T., Treu, L., Boldrin, A., Sartori, C., Angelidaki, I., Scheutz, C., 2017. Microbial population dynamics in urban organic waste anaerobic co-digestion with mixed sludge during a change in feedstock composition and different hydraulic retention times. Water Research, 118(Supplement C), 261-271. Guo, X.M., Trably, E., Latrille, E., Carrère, H., Steyer, J.-P., 2010. Hydrogen production from agricultural waste by : A review. International Journal of Hydrogen Energy, 35(19), 10660-10673. Hjorth, M., Fernandez, M., Jayaram, S., Sørensen, J.A., Adamsen, A.P.S., 2015. Bio-acidification of manure–By supplying manure with 2-3% sugar or cellulose. RAMIRAN 2015–16th International Conference. TuTech Innovation GmbH, Hamburg. pp. 96-96. Holzapfel, W., Wood, B.J., 2012. The genera of lactic acid bacteria. Springer Science & Business Media. Karakashev, D., Schmidt, J.E., Angelidaki, I., 2008. Innovative process scheme for removal of organic matter, phosphorus and nitrogen from pig manure. Water Research, 42(15), 4083-4090. Kim, M., Gomec, C.Y., Ahn, Y., Speece, R.E., 2003. Hydrolysis and acidogenesis of particulate organic material in mesophilic and thermophilic anaerobic digestion. Environmental Technology, 24(9), 1183-1190. Komemoto, K., Lim, Y.G., Nagao, N., Onoue, Y., Niwa, C., Toda, T., 2009. Effect of temperature on VFA’s and biogas production in anaerobic solubilization of food waste. Waste Management, 29(12), 2950-2955. Leite, A.F., Janke, L., Harms, H., Richnow, H.-H., Nikolausz, M., 2016. Lessons learned from the microbial ecology resulting from different inoculation strategies for biogas production from waste products of the bioethanol/sugar industry. Biotechnology for Biofuels, 9(1), 144. Li, X., Chen, Y., Zhao, S., Wang, D., Zheng, X., Luo, J., 2014. Lactic acid accumulation from sludge and food waste to improve the yield of propionic acid-enriched VFA. Biochemical Engineering Journal, 84, 28-35. Liang, S., Gliniewicz, K., Mendes-Soares, H., Settles, M.L., Forney, L.J., Coats, E.R., McDonald, A.G., 2015. Comparative analysis of microbial community of novel lactic acid fermentation inoculated with different undefined mixed cultures. Bioresource Technology, 179(Supplement C), 268-274. Lin, L., Wen, L., Chen, S., Yang, X., Liu, X., Wan, C., 2015. Effect of alkaline treatment pattern on anaerobic fermentation of swine manure. Process Biochemistry, 50(11), 1710-1717. Lu, X.-M., Lu, P.-Z., Zhang, H., 2014. Bacterial communities in manures of piglets and adult pigs bred with different feeds revealed by 16S rDNA 454 pyrosequencing. Applied Microbiology and Biotechnology, 98(6), 2657-2665. Ludwig, W., Strunk, O., Westram, R., Richter, L., Meier, H., Yadhukumar, Buchner, A., Lai, T., Steppi, S., Jobb, G., 2004. ARB: a software environment for sequence data. Nucleic acids research, 32(4), 1363-1371. Mainstone, C.P., Parr, W., 2002. Phosphorus in rivers — ecology and management. Science of The Total Environment, 282(Supplement C), 25-47. Mercier, P., Yerushalmi, L., Rouleau, D., Dochain, D., 1992. Kinetics of lactic acid fermentation on glucose and corn by Lactobacillus amylophilus. Journal of Chemical Technology & Biotechnology, 55(2), 111-121. Molinos-Senante, M., Hernández-Sancho, F., Sala-Garrido, R., Garrido-Baserba, M., 2011. Economic feasibility study for phosphorus recovery processes. Ambio, 40(4), 408-416. Monlau, F., Aemig, Q., Trably, E., Hamelin, J., Steyer, J.-P., Carrere, H., 2013. Specific inhibition of biohydrogen-producing Clostridium sp. after dilute-acid pretreatment of sunflower stalks. International Journal of Hydrogen Energy, 38(28), 12273-12282.

113

Nakamura, L., Crowell, C., 1979. Lactobacillus amylophilus, a new starchhydrolyzing species from swine waste-corn fermentation. Dev. Ind. Microbiol, 20, 531-540. Nakamura, L.K., 1981. Lactobacillus amylovorus, a New Starch-Hydrolyzing Species from Cattle Waste-Corn Fermentations. International Journal of Systematic and Evolutionary Microbiology, 31(1), 56-63. Némery, J., Garnier, J., Morel, C., 2005. Phosphorus budget in the Marne Watershed (France): urban vs. diffuse sources, dissolved vs. particulate forms. Biogeochemistry, 72(1), 35-66. Noike, T., Takabatake, H., Mizuno, O., Ohba, M., 2002. Inhibition of hydrogen fermentation of organic wastes by lactic acid bacteria. International Journal of Hydrogen Energy, 27(11–12), 1367- 1371. Oksanen, J., Kindt, R., Legendre, P., O’Hara, B., Stevens, M.H.H., Oksanen, M.J., Suggests, M., 2007. The vegan package. Community ecology package, 10, 631-637. Oude Elferink, S.J., Driehuis, F., Gottschal, J.C., Spoelstra, S.F., 1999. Silage fermentation processes and their manipulation FAO electronic conference on tropical silage, Rome, Italy. Perera, P.A., Han, Z.-Y., Chen, Y.-X., Wu, W.-X., 2007. Recovery of nitrogen and phosphorous as struvite from swine waste biogas digester effluent. Biomedical and environmental sciences, 20(5), 343-350. Perera, P.W.A., Wu, W.-X., Chen, Y.-X., Han, Z.-Y., 2009. Struvite Recovery from Swine Waste Biogas Digester Effluent through a Stainless Steel Device under Constant pH Conditions. Biomedical and Environmental Sciences, 22(3), 201-209. Piveteau, S., Picard, S., Dabert, P., Daumer, M.-L., 2017. Dissolution of particulate phosphorus in pig slurry through biological acidification: A critical step for maximum phosphorus recovery as struvite. Water Res., 124(Supplement C), 693-701. Poirier, S., Bize, A., Bureau, C., Bouchez, T., Chapleur, O., 2016a. Community shifts within anaerobic digestion microbiota facing phenol inhibition: Towards early warning microbial indicators? Water research, 100, 296-305. Poirier, S., Desmond-Le Quéméner, E., Madigou, C., Bouchez, T., Chapleur, O., 2016b. Anaerobic digestion of biowaste under extreme ammonia concentration: identification of key microbial phylotypes. Bioresource technology, 207, 92-101. Probst, M., Walde, J., Pümpel, T., Wagner, A.O., Insam, H., 2015. A closed loop for municipal organic solid waste by lactic acid fermentation. Bioresource Technology, 175(Supplement C), 142- 151. Regueiro, L., Carballa, M., Lema, J.M., 2016. Microbiome response to controlled shifts in ammonium and LCFA levels in co-digestion systems. Journal of Biotechnology, 220(Supplement C), 35-44. Ren, N.Q., Chua, H., Chan, S.Y., Tsang, Y.F., Wang, Y.J., Sin, N., 2007. Assessing optimal fermentation type for bio-hydrogen production in continuous-flow acidogenic reactors. Bioresource Technology, 98(9), 1774-1780. Sakamoto, M., Takagaki, A., Matsumoto, K., Kato, Y., Goto, K., Benno, Y., 2009. Butyricimonas synergistica gen. nov., sp. nov. and Butyricimonas virosa sp. nov., butyric acid-producing bacteria in the family ‘Porphyromonadaceae’ isolated from rat faeces. International Journal of Systematic and Evolutionary Microbiology, 59(7), 1748-1753. Sattari, S., Van Ittersum, M., Giller, K., Zhang, F., Bouwman, A., 2014. Key role of China and its agriculture in global sustainable phosphorus management. Environmental Research Letters, 9(5), 054003. Schindler, D.W., Hecky, R.E., Findlay, D.L., Stainton, M.P., Parker, B.R., Paterson, M.J., Beaty, K.G., Lyng, M., Kasian, S.E.M., 2008. Eutrophication of lakes cannot be controlled by reducing

114 nitrogen input: Results of a 37-year whole-ecosystem experiment. Proceedings of the National Academy of Sciences, 105(32), 11254-11258. Schipanski, M.E., Bennett, E.M., 2012. The Influence of Agricultural Trade and Livestock Production on the Global Phosphorus Cycle. Ecosystems, 15(2), 256-268. Schuiling, R.D., Andrade, A., 1999. Recovery of Struvite from Calf Manure. Environ. Technol., 20(7), 765-768. Senthilkumar, K., Nesme, T., Mollier, A., Pellerin, S., 2012. Conceptual design and quantification of phosphorus flows and balances at the country scale: The case of France. Global Biogeochemical Cycles, 26(2), n/a-n/a. Sharpley, a.n., Withers, p.j., Abdalla, c.w., Dodd, a.r., 2005. Strategies for the Sustainable Management of Phosphorus. Shen, D., Yin, J., Yu, X., Wang, M., Long, Y., Shentu, J., Chen, T., 2017. Acidogenic fermentation characteristics of different types of protein-rich substrates in food waste to produce volatile fatty acids. Bioresource Technology, 227, 125-132. Smil, V., 2000. phosphorus in the environment: natural flows and human interferences. Annual Review of Energy and the Environment, 25(1), 53-88. Snell-Castro, R., Godon, J.-J., Delgenès, J.-P., Dabert, P., 2005. Characterisation of the microbial diversity in a pig manure storage pit using small subunit rDNA sequence analysis. FEMS Microbiology Ecology, 52(2), 229-242. Suzuki, K., Tanaka, Y., Kuroda, K., Hanajima, D., Fukumoto, Y., 2005. Recovery of phosphorous from swine wastewater through crystallization. Bioresource Technology, 96(14), 1544-1550. Szögi, A.A., Vanotti, M.B., Hunt, P.G., 2015. Phosphorus recovery from pig manure solids prior to land application. Journal of Environmental Management, 157(0), 1-7. Thomas, T.D., Ellwood, D.C., Longyear, V.M.C., 1979. Change from homo-to heterolactic fermentation by Streptococcus lactis resulting from glucose limitation in anaerobic chemostat cultures. Journal of bacteriology, 138(1), 109-117. Tilman, D., 1998. The greening of the green revolution. Nature, 396(6708), 211-212. Vanotti, M.B., Szogi, A.A., Hunt, P.G., Millner, P.D., Humenik, F.J., 2007. Development of environmentally superior treatment system to replace anaerobic swine lagoons in the USA. Bioresour. Technol., 98(17), 3184-3194. Venables, W., Smith, D., 2016. An introduction to R. Notes on R: a programming environment for data analysis and graphics. version 3.2.4 ed, Vol. (Accessed 28March 2016). Wang, Q., Garrity, G.M., Tiedje, J.M., Cole, J.R., 2007. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Applied and environmental microbiology, 73(16), 5261-5267. Wang, Y., Qian, P.-Y., 2009. Conservative fragments in bacterial 16S rRNA genes and primer design for 16S ribosomal DNA amplicons in metagenomic studies. PloS one, 4(10), e7401. Weinberg, Z.G., Szakacs, G., Ashbell, G., Hen, Y., 1998. The effect of temperature and Lactobacillus amylovorus and Lact. plantarum, applied at ensiling, on wheat silage. Journal of Applied Microbiology, 84(3), 404-408. Wu, Y., Ma, H., Zheng, M., Wang, K., 2015. Lactic acid production from acidogenic fermentation of fruit and vegetable wastes. Bioresource Technology, 191(Supplement C), 53-58. Zhang, D.-m., Chen, Y.-x., Jilani, G., Wu, W.-x., Liu, W.-l., Han, Z.-y., 2012. Optimization of struvite crystallization protocol for pretreating the swine wastewater and its impact on subsequent anaerobic biodegradation of pollutants. Bioresource Technology, 116(Supplement C), 386-395.

115

CHAPTER 3 BIOLOGICAL ACIDIFICATION OF PIG SLURRY USING ORGANIC CO-SUBSTRATES: AN EFFICIENT PROCESS FOR PHOSPHORUS DISSOLUTION PRIOR TO STRUVITE CRYSTALLIZATION

Piveteau S., Dabert P., Daumer M-L. (2017). Submitted in Water Research

116

I. SUMMARY OF CHAPTER 3

Les travaux précédents ont permis d’une part de définir une relation entre les caractéristiques du lisier et la quantité de sucre à ajouter pour obtenir le pH désiré et d’autre part que l’acidification pouvait être obtenue avec des co-substrats complexes. Cependant la baisse de pH pour une même quantité de matière organique ajoutée est moindre.

L’objectif des travaux décrits dans ce troisième article est de définir un « équivalent sucre » pour quelques déchets issus de la production agricole régionale puis de vérifier la relation établie précédemment pour calculer la quantité de co-substrats à ajouter pour atteindre le pH désiré en fonction des caractéristiques du lisier. Enfin l’influence de l’acidification biologique sur la cristallisation du P sous forme de struvite et la qualité du solide final a été testée.

Quatre lisiers issus d’élevages industriels ont été caractérisés comme précédemment et 5 co-substrats, dont le sucre, y ont été ajoutés à la concentration de 50 g de matière organique par litre pour déterminer l’équivalent sucre. Les 4 substrats sont des légumes ou des fruits habituellement produits en Bretagne et dont les invendus où les déchets de transformation alimentent les méthaniseurs. Ce sont des co-substrats riches en sucre mais avec des teneurs ou une accessibilité différente. Il s’agit des carottes, pommes petits pois et haricots verts.

L’équivalent sucre est le rapport entre la quantité de sucre et la quantité de co-substrat permettant d’obtenir un pH équivalent. Des valeurs respectives de 0,74, 0,69, 0,42 et 0,42 ont été calculées pour les pommes, les carottes, les haricots et les petits pois. Les quantités pour atteindre des valeurs de pH de 4, 5 et 6 ont été calculées d’après l’équation citée plus haut et ont été ajoutées aux 4 lisiers. L’adéquation entre le pH obtenu et le pH calculé a été estimée par la moyenne des valeurs absolues des différences ou par la moyenne des différences positives ou négatives.

L’adéquation est satisfaisante pour le sucre comme pour la plupart des co-substrats sauf les petits pois. Dans ce dernier cas, les acides majoritairement produits ne sont pas le lactate mais l’acétate, le propionate, le butyrate et le valérate. Ces acides ayant un pKa supérieur à 4.5, il n’est pas possible d’atteindre un pH de 4. La suppression des valeurs obtenues pour les essais dont l’objectif était d’atteindre ce pH permet d’améliorer l’adéquation entre les valeurs de pH attendue et mesurée. La différence est alors au maximum de 0,45 point de pH. La valeur optimale de pH ayant été fixée à 5.5-6 le modèle peut être utilisé pour estimer la quantité de co-substrat à ajouter en fonction des caractéristiques du lisier. Le dosage des acides organiques a permis de montrer que l’équivalent sucre est d’autant plus important que le taux de conversion du carbone en acide lactique est élevé.

Pour des valeurs de pH proches de la valeur optimale de 5.5-6 une fermentation secondaire intervient systématiquement ce qui confirme l’hypothèse émise précédemment selon laquelle le procédé devra être piloté avec un suivi rapproché du pH pour réaliser la séparation lorsque le ratio Ca/P est optimal.

117

Comme lors des essais précédents, l’évolution de la concentration en P dissous lors de l’acidification avec les co-substrats est étroitement liée à l’évolution du pH pour 3 des quatre lisiers. Un comportement différent a cependant été observé pour un des lisiers qui avait une concentration initiale en P total très faible. Dans ce cas et surtout lorsque les quantités de co-substrat ajoutées sont importantes l’assimilation du P pour la croissance de la biomasse n’est plus négligeable et réduit la concentration en P dissous observée. Si le co-substrat est riche en P comme c’est le cas des petits pois, la concentration en P dissous sera liée à l’hydrolyse progressive des formes organiques contenant le P sous l’effet de l’acidification.

La dissolution du Ca est plus variable suivant les lisiers. Elle est comparable à celle observée précédemment avec 2 des 4 lisiers. Dans le lisier pauvre en P et en Ca, une part importante du Ca vient du co-substrat et comme pour le P sa dissolution est liée à l’hydrolyse du co-substrat. Dans le lisier le plus riche en Ca, on observe une augmentation brutale de la concentration dissoute lorsque le pH est proche de 5 ce qui laisserait supposer que la forme sous laquelle se trouve le Ca dans ce lisier est différente de celle des autres lisiers. La concentration en ammonium évolue peu avec l’acidification sauf dans les essais avec les petits pois dans lesquels l’évolution constante semble être liée à l’hydrolyse du co-substrat.

Afin d’évaluer l’intérêt de conserver un inoculum acidogène adapté au co-substrat des essais en semi- continu ont également été réalisés. Une alimentation tous les jours ou tous les quatre jours a été mise en œuvre. Les résultats étant identiques, seul l’essai avec une alimentation quotidienne sont présentés dans l’article. Après avoir rempli un réacteur avec 4 litres du lisier 4, un mélange petits pois carottes a été ajouté et le réacteur placé sous agitation pendant 4 jours. Ensuite, un quart du réacteur était remplacé quotidiennement par un mélange lisier petits pois et carottes dans les mêmes proportions (temps de séjour de 4 jours) pendant 12 jours. Le pH est descendu rapidement à 5.7 pendant les 4 premiers jours puis le pH minimum a baissé progressivement jusqu’à 5.4. La concentration en P dissous est passée de 150 à 680 mg.L-1 en quatre jours puis a augmenté progressivement pour atteindre 747 mg.L-1 soit 57% du P total du mélange au bout des 16 jours.

L’effluent recueilli pendant la durée du test a été centrifugé puis analysé avant d’être utilisé pour des essais de cristallisation. La centrifugation a permis de récupérer 83% du volume sous forme liquide soit 47% du P total initial. Sa composition est proche de celle observée lors des tests batchs avec un ratio N :P :Mg :Ca de 6.5 :1 :0.9 :0.4 soit une quantité d’ammonium favorable à la cristallisation de struvite, un léger déficit en magnésium et la présence de calcium en proportion assez importante. La cristallisation a été réalisée en bécher agité. Environ 3.5g d’hydroxyde de magnésium ont été nécessaire pour atteindre pH 8. Le mélange a été laissé sous agitation pendant 2 heures. Plus de 99% du P dissous a été précipité. Après cristallisation le solide a un ratio N :P :Mg :Ca de 1 :1.19 :1.32 :0.05 proche de celui de la struvite avec un léger excès de magnésium et de phosphore et une faible contamination par le calcium. La struvite étant l’un des seuls minéraux contenant de

118 l’ammonium susceptible de se former dans nos conditions la quantité de struvite a été calculée d’après la concentration en ammonium. Le solide produit est composé à 69% de struvite, le reste étant probablement d’autres composés minéraux phosphatés, de l’hydroxyde de magnésium non dissous et de la matière organique.

Le faible temps de contact et le mode d’agitation ont conduit à l’obtention de cristaux de petite taille et hétérogènes. Aussi, malgré une précipitation presque totale du P dissous, le solide issu de la filtration à 100 µm n’a permis de recueillir que 26% du P total du mélange initial. L’obtention d’un pH légèrement inférieur lors de l’acidification aurait probablement permis d’améliorer le taux de dissolution du P et la réalisation de la cristallisation dans un équipement industriel optimisé d’obtenir des cristaux plus gros. Ces deux étapes sont déterminantes pour améliorer le rendement global du procédé. Si des équipements industriels développés spécifiquement pour la cristallisation de la struvite sont disponibles sur le marché, une bonne connaissance des co-substrats et notamment de leur équivalent sucre est déterminante pour choisir les cocktails qui favoriseront la production d’acide lactique et limiteront les fermentations secondaires. Des études complémentaires sur l’influence des paramètres de l’étape d’acidification biologique (temps de séjour, température …) devraient également permettre d’orienter les métabolismes.

119

Biological acidification of pig slurry using organic co-substrates: An efficient process for phosphorus dissolution prior to struvite crystallization

Piveteau Simona,b,*, Dabert Patricka,b, Daumer Marie-Linea,b

a Irstea, UR OPAALE, 17 Avenue de Cucillé-CS 64427, F-35044 Rennes, France b Univ de Bretagne Loire, France.

*Corresponding author. Simon Piveteau. UR OPAALE, Irstea. E-mail address: [email protected]

120

II. ABSTRACT

Pig slurry is a suitable waste stream for phosphorus recovery as struvite but requires a dissolution step through acidification prior to crystallization. Biological acidification of pig slurry was studied in batch tests by adding several co-substrates (carrots, beans, peas, apples and sucrose). Dissolved phosphorus increased from 7-10% of total-P at initial pH (6.5-7.7) to 60-90% when low pH (4-5.2) was reached. A simple crystallization step using magnesium hydroxide to increase pH to 8 led to complete phosphate removal. The dried solid obtained contained 69% of struvite and less than 0.5% of calcium based on ionic concentrations obtained after acidic dissolution. The loss of small crystals through the sieve (100 µm pore size) limited the recovery to 58% of dissolved-P and 26% of Total-P, but simple optimization steps such as longer retention time and recirculation of the filtrate should enable the formation of larger crystals and substantially improve the P-recovery rate.

121

III. INTRODUCTION

Phosphorus, a necessary nutrient for life (DNA, RNA, ATP) is a fossil resource in limited supply. Population growth increases the need for higher crop yield and therefore P-based fertilizers. Despite varying estimations regarding the amount, quality and accessibility of P reserves, it is commonly accepted that a sustainable agricultural system assuring long term food security requires a more efficient use of fertilizer and the large scale implementation of P recovery processes for industrial and municipal wastewaters as well as livestock waste streams (Cordell et al., 2009). Agronomical progress to optimize fertilizer use and development of P recycling technologies would also severely limit soil leaching and associated eutrophication of water bodies. The agricultural specialization of large areas into crop farming on one hand and livestock breeding on the other hand has led to major imbalances in term of localized phosphorus need/excess. While crop-oriented regions rely heavily on fertilizer to compensate their P exports, regions focused on intensive livestock production struggle to find outputs for their manure. Recovering phosphorus from livestock farms under a concentrated and easily transportable form would limit the over application of manure in P-saturated soils and provide an alternative P-based fertilizer for other areas in need.

Many full-scale P-recovery processes have been successfully implemented for a wide range of domestic and industrial wastewaters, e.g. production of calcium phosphate from the supernatant of anaerobic EBPR sludge through the Crystalactor® process, (Giesen, 1999), production of a N-P-K fertilizer from sewage sludge ashes using the Ash Dec® process (Mattenberger et al., 2010), production of struvite (NH4MgPO4,6H2O) from reject water of a municipal digester and anaerobic effluent of a potato processing plant with PHOSPAQTM technology (Abma et al., 2010), from municipal EBPR sludge dewatering liquid using the WASSTRIP® system (Schauer and Laney, 2013) and from acidified digested sludge through the Seaborne process (Günther et al., 2008). However, very few P-recovery technologies have been implemented on full scale to treat livestock waste (Desmidt et al., 2015). A full scale demonstration unit for P-recovery as calcium phosphate from liquid swine slurry was tested during one year in Mount Olive, North Carolina using nitrification to reduce the buffering capacity and calcium hydroxide to bring the pH up into precipitation range and promote calcium phosphate precipitation (Vanotti et al., 2007). A full-scale plant producing K-struvite from the liquid phase of calf manure has been operating in Putten (Netherlands) since 1999, using magnesium oxide to increase the pH (Schuiling and Andrade, 1999). Several P-recovery technologies applied to swine wastewater are still tested on lab-scale. Because magnesium is usually the limiting ion for struvite precipitation, MgCl2 is added, while aeration and/or NaOH are used for pH adjustment (Burns et al., 2003; Liu et al., 2011; Rahman et al., 2011; Ye et al., 2011). Each livestock wastewater treatment described above only recovered precipitated phosphorus from the liquid phase, leaving P in the solid phase untouched. Because phosphorus in pig slurry is mostly present in inorganic solids

122

(Barnett, 1994; Daumer et al., 2008; Toor et al., 2006), precipitation of orthophosphate in the liquid phase enables a low concentration in the effluent that meets discharge standards, but also a very limited P-recovery as mineral fertilizer overall. For example, in the Mount Olive treatment unit described earlier, the solid phase containing 70% of total phosphorus (TP) was not directed to the calcium phosphate recovery process but sent instead to a centralized solid processing unit nearby for aerobic composting.

Therefore, acidification of pig slurry to dissolve inorganic phosphorus is a necessary step if high P- recovery as mineral fertilizer is the goal. It has been shown that 70% of TP could be dissolved when pH was decreased to 5.5 (Christensen et al., 2009). Chemical acidification of swine slurry has been investigated on lab-scale using organic acids such as acetate and formate allowing a dissolution 89% of TP at pH 4.9 and 95% at pH 4.75 respectively (Daumer et al., 2010). Following this dissolution experiment, phosphorus was precipitated as struvite using magnesium oxide (MgO), but the process was considered economically noncompetitive due to the cost of formic acid. A possible solution to avoid the cost of chemical acidification is to take advantage of the biological process known as acidogenesis. Under anaerobic conditions, organic co-substrates added to pig slurry are fermented by endogenous acidogenic bacteria into organic acids, which decreases pH. Besides, the organic acids produced during acidogenesis could be used for methane production in a subsequent digestion process. Biological acidification of swine slurry using glucose as co-substrate (30g/L) led to the production of lactic acid and a pH of 4.3 (Hjorth et al., 2015). Another experiment with sucrose as co-substrate showed that 60 to 90% of TP could be dissolved when pH reached 4-5.5, with lactic acid as the main fermentation product (Piveteau et al., 2017).

In Brittany, France, intensive pig breeding and over application of slurry in the fields has led to the contamination of fresh waters and coastal algal blooms. Due to the regulatory limits imposed on spreading of slurry during the 90’s, around 500 treatment units have been built in the region, mostly with nitrification-denitrification and activated sludge processes. Since the beginning of the new millennium, 50 anaerobic digesters have also been installed to produce energy by co-digesting pig slurry with organic waste from agricultural, industrial and municipal activities nearby. In both treatment pathways, the solid phase that contains most of the phosphorus is partially spread while complying with the stringent regulatory guidelines or composted, exported and sold as an organic fertilizer in other regions. Struvite, a P-based, slow-release mineral fertilizer would be more easily exported due to its higher P content and the organic, solid phase could be kept in Brittany. The hydrolysis taking place during biological acidification of organic waste and pig slurry could also enhance the energy production through higher methane yield.

In this study, biological acidification of pig slurry was investigated in anaerobic batch tests using apples, carrots, green beans, peas and sucrose as organic co-substrates. Those fruits and vegetables

123 were chosen because there are common products grown in Brittany and available locally as agricultural or industrial waste. They possess different levels of sugar content, apple having the highest content followed by carrot, green bean and pea. Sugars are the most effective organic compounds to produce organic acids and lower the pH (Yin et al., 2016). The model linking buffer capacity of the slurry, initial pH and sucrose concentration with lowest pH reached during acidification (Piveteau et al., 2017) was extended to those co-substrates and its reliability was tested in batch tests with four slurries and three different concentrations of co-substrates. pH, organic acids concentration and dissolution of P, Mg, N and Ca were studied throughout the batch tests. A semi continuous reactor fed with swine slurry, carrots and peas was set up to produce a stable acidified effluent (pH = 5.42). After centrifugation, magnesium hydroxide was added to the supernatant to bring the pH to 8 and precipitate phosphorus. The dried solid recovered was dissolved at acidic pH and ionic concentrations were measured to verify that struvite was effectively produced.

IV. MATERIALS AND METHODS

1. PIG SLURRIES

Four raw swine slurries were collected from four different commercial pig fattening farms in Brittany. The composition of the slurries can be found in Table 11. Their buffer capacity was measured in a beaker containing 100 mL of slurry with a magnetic stirrer, a pH probe and a graduated burette containing sulfuric acid (1 mol/L). The buffer capacity was expressed as milliequivalent per liter per pH point and calculated according to the equation (Equation 15) below:

Equation 15. Calculation of buffer capacity of the slurry

Buffer capacity (mEq/L) = A / 0.1 / (pH0 - 4)

A: amount of sulfuric acid in milliequivalent added to 0.1L of slurry to reach pH 4. pH0: initial pH of the slurry

Table 11. Composition of the pig slurries

3- + 2+ 2+ TS VS TP TN T-Ca T-Mg pH Buffer PO4 -P NH4 -N Ca Mg g/L mEq/L g/L slurry 1 28.45 17.16 0.23 4.06 0.44 0.38 7.63 71.6 0.019 3.237 0.191 0.306 slurry 2 50.50 36.79 0.99 3.50 1.93 0.65 7.05 87.4 0.073 2.259 0.217 0.055 slurry 3 15.58 9.47 0.48 2.50 1.35 0.48 7.74 45.5 0.032 1.702 0.176 0.071 slurry 4 68.69 50.81 1.56 4.25 3.23 0.90 6.53 110.7 0.150 1.836 0.264 0.030 2. ORGANIC CO-SUBSTRATES

Frozen carrots, beans and peas as well as fresh apples from the local supermarket were used as complex organic co-substrates for pig slurry biological acidification. Their composition can be found in Table 12.

124

Table 12. Composition of the organic co-substrates

TS VS TP TN T-Ca T-Mg g/L mg/L test 1-2 89.33 83.35 197 1532 390 83 carrots test 3 91.39 85.84 193 1329 206 86 test 4 102.38 94.54 322 1265 200 120

test 1-2 107.69 100.65 317 3477 656 184 beans test 3 105.61 97.90 382 3585 509 294 test 4 122.02 113.86 382 3547 690 269

test 1-2 211.29 205.59 860 8756 376 361 peas test 3 223.34 213.78 1292 9100 370 422 test 4 208.82 202.70 1013 8240 289 278

test 1-2 133.81 126.66 97 339 121 52 apple test 3 144.13 134.60 239 385 152 127 test 4 111.70 105.04 216 371 95 76

3. DETERMINATION OF THE ACIDIFYING POWER FOR EACH CO-SUBSTRATE COMPARED TO SUCROSE

During biological acidification of pig slurry with sucrose, the lowest pH reached decreases linearly with the initial amount of sucrose added down to pH 4. Biological acidification of digested swine slurry was conducted with 7 different sucrose concentrations (in simplicate): 0, 10, 20, 30, 40, 50, and 60 g/L (Figure 34). The digested pig slurry was obtained from an 87L in-house digester that provides a stable effluent all year long. Samples were taken at 24, 48, 72 and 96 h for pH measurement. A linear equation (Equation 16) was obtained, linking lowest pH reached during the 96 h batch with the initial sucrose concentration.

Equation 16. Calculation of lowest pH based on initial pH and sucrose concentration lowest pH = pH0 –A*S0

S0: Initial sucrose concentration (g/L) A: slope R2=0.9931

125

Figure 34. Lowest pH reached at different initial sucrose concentrations. Example of sucrose-equivalent determination for apple

A biological acidification with each organic co-substrate at 50g-VS per liter of digested slurry was performed in simplicate. Based on the lowest pH obtained with the co substrate, Eq.1 could be solved for the corresponding sucrose concentration. The sucrose equivalent (g-sucrose/g-VS) was calculated as the corresponding sucrose concentration (g-sucrose/L) divided by 50 (g-VS/L). The test was repeated with a lower concentration of co-substrate in the case of carrots and apples (30 and 36g-VS/L for carrot and apple respectively) due to an acidification that went beyond the linear pH range. The sucrose equivalent of each co substrate can be found in Table 13.

Table 13. Sucrose equivalents of the co-substrates

carrot green bean pea apple VS added (g/L) 30 50 50 36 lowest pH reached 5.49 5.44 5.42 4.66 corresponding sucrose concentration (g/L) 20.52 20.95 21.08 26.73 sucrose eq. (g-sucrose/g-VS) 0.69 0.42 0.42 0.74

4. BIOLOGICAL ACIDIFICATION OF PIG SLURRY IN BATCH TESTS USING COMPLEX ORGANIC CO-SUBSTRATES

The tests were conducted in 1L bottles inoculated with 400 mL of raw pig slurry. Three different concentrations for each organic co-substrate and sucrose were added to the bottles before sealing them with a lid. The three concentrations of co-substrate (“low”, “medium”, “high”) were calculated using a simple empirical model adapted from Piveteau et al. (2017) linking initial concentration of sucrose, buffer capacity, initial and lowest pH (Equation 17, Equation 18 and Table 14), with the objective to target three minimum pH between 4 and 6, a range in which phosphorus is largely dissolved. No

126 oxygen removal was performed in the headspace since it has no effect acidification (Piveteau et. al. 2017). The bottles were placed in a thermostatic shaker, at 120 rpm and 38 °C. Samples were taken every 24 hours during 4 days for analysis. The same protocol was applied to the four slurries.

Equation 17. Calculation of initial sucrose concentration necessary to reach a targeted pH

Sucrose0 (g/L) = -105.404 – 16.879 pHtargeted+ 27.271*pH0 + 0.283*Buffer (mEq/L)

Equation 18. Calculation of initial co-substrate concentration based on its sucrose equivalent

Co-substrate (g-VS/L) = Sucrose0 (g/L) / sucroseeq(g-sucrose/g-VS)

Table 14. Co substrates and sucrose concentrations in each batch test (g/L)

carrot green bean pea apple sucrose Concentration fresh VS fresh VS fresh VS fresh VS

high 1058 88 1348 136 700 144 600 76 58 medium 815 68 1038 104 539 111 462 58 44

slurry 1 slurry low 571 48 728 73 378 78 324 41 31

high 616 51 511 51 410 84 379 48 36 medium 386 32 320 32 257 53 237 30 22

slurry 2 slurry low 156 13 129 13 103 21 96 12 9

high 789 68 649 63 520 111 481 65 45 medium 554 48 458 45 367 78 340 46 32

slurry 3 slurry low 324 28 269 26 215 46 198 27 19

high 382 36 589 68 291 59 265 28 25 medium 237 22 365 42 180 37 164 17 15

slurry 4 slurry low 91 9 139 16 69 14 63 7 6

5. SEMI CONTINUOUS REACTOR

A semi-continuous reactor for biological acidification was set up with the objective of studying the effect of semi-continuous feeding on acidogenic biomass and potential improvements in term of pH decrease, organic acids production, and P dissolution. Slurry 4 (3 L), carrots (560 g) and peas (440 g) were introduced in a 4 L reactor under anaerobic conditions with a helix for mixing. After 4 days, semi continuous operation of the reactor was started with one fourth of the volume withdrawn and replaced once a day with the same proportion of pig slurry, carrots and peas (HRT = solid retention time = 4 days). The reactor was operated during 16 days, 4 days as batch and 12 days (3 HRT) in semi- continuous mode. pH, organic acids and ionic concentrations were measured daily in the effluent.

6. CRYSTALLIZATION PROCESS

The effluent collected during the last five days of semi continuous operation in the semi continuous reactor was used for struvite precipitation. The effluent was centrifuged immediately at 10000 RPM

127 during 30 min at 10°C with the addition of 1.5 g of coagulant per liter of acidified slurry (coagulant FLOPAM EM 840 TBD). 400 mL of supernatant was poured in a beaker equipped with a pH probe and a helix. Mixing was set at 40 RPM and a solution of Mg(OH)2 at 100 g/L was introduced progressively until pH 8 was reached. The mixing was maintained during 2 hours to allow crystal growth. Then the content of the beaker was poured through a sieve with pores of 100 µm, previously weighed. The sieve was left to dry at room temperature during 48 hours and weighed again to measure the amount of solid recovered. Analysis of the solid’s ionic content was done through acidic 3- + 2+ 2+ dissolution and measurement of PO4 , NH4 , Mg , Ca ions (see section IV.7).

7. ANALYSIS

After sample collection, pH was measured using a WTW probe. Total nitrogen (TN), total solids (TS) and volatile solids (VS) were measured in triplicates on the raw slurry with standard methods (APHA, 1998). Total phosphorus (TP), total calcium (T-Ca) and total magnesium (T-Mg) were measured after acidic dissolution of the slurry’s ashes. 200 mg of ashes were added to 0.5 g of K2SO8 and 5 mL of a mix of H2SO4 / HNO3 (75:25) in triplicate. The liquid obtained was autoclaved at 110°C during one hour at 1 bar. The concentrations in TP, T-Ca and T-Mg (expressed in gram per liter of raw manure in Table 11) were measured using automated colorimetric methods on a spectrophotometer (Gallery, Thermoscientific). The references for the respective methods are 984366, 984361 and 984360.

The samples collected daily were centrifuged during 20 minutes (4°C, 20,000g) and the supernatant obtained was filtered through a 0.45 µm polypropylene membrane. Phosphate concentration was measured with automated colorimetric methods on the Gallery. Cationic concentrations were measured by ionic chromatography Metrohm 940 Professional Vario IC with a Metrosep C4 -250/4,0 column. Sucrose, glucose, fructose and organic acids composition of the supernatant was measured by HPLC (Ultimate 3000, Dionex) with a Hiplex-H column (Agilent).

For solid analysis from the crystallization process, 100 mg of dried solid recovered after crystallization and filtration was introduced in in a beaker with 150 mL of distilled water and pH was adjusted at 1.8 3- with sulfuric acid (1 mol/L). Volume was adjusted to 200 mL with distilled water. P04 concentration 2+ + 2+ was measured on the Gallery while Mg , NH4 and Ca were measured by ionic chromatography.

V. RESULTS AND DISCUSSION

1. MODEL FITNESS

The deviation of the model was calculated for each co-substrate as the average absolute difference between the measured lowest pH and the targeted lowest pH (Table 15 and Figure 35). The average relative difference between measured lowest pH and targeted lowest pH, also displayed in Table 15, reflects the overall overestimation (positive average pH difference) and underestimation (negative

128 average pH difference) of the acidifying capacity of the co-substrate. The model demonstrated a moderate fitness (absolute deviation) in the case of sucrose and for most of the co-substrates but showed a relatively poor fitness in the case of pea. The deviation could be due to an overestimation of the sucrose equivalent for peas. Indeed, the lowest pH obtained was almost always above the targeted pH. This overestimation could itself be explained by the fact that the fermentation products of pea differs from those of sucrose. Lactate was the main molecule produced during biological acidification of pig slurry using sucrose (Figure 36B) and this organic acid has a low pKa compared to usual volatile fatty acids (VFA). On the contrary, fermentation of pig slurry with pea as organic co-substrate produced mainly acetate, propionate, butyrate and valerate (Figure 36A). As a result, while sucrose had a linear effect on minimum pH down to pH 4.5-4 (Figure 37A), that was not the case for pea, with an inflexion around 6-5.5 (Figure 37B). This makes the calculated sucrose equivalent for pea a function of pH, not a fixed value as initially postulated. The sucrose equivalent of pea was determined during a batch test in which the lowest pH reached was 5.4 (Table 13). Consequently, when lower pH values were targeted, the amount of co-substrate necessary was underestimated. To a lesser extent, the same conclusion applied to green beans, which also did not produce any lactic acid. However, the batch tests with carrots and apples did have lactic acid production and were able to reach lower pH. The absolute error of the model for pea and green bean was significantly lower when only the targeted pH above 5 were taken into account (Table 15).

The batch test with slurry 4 led systematically to a minimum pH higher than the targeted value. Based on the proportion of added VS converted to organic acids, it could be postulated (no gas analysis was performed) that methanisation occurred, converting some of the organic acids into acetate and then methane, limiting the acidification. Indeed, as seen in Figure 38A, the proportion of co-substrate converted into organic acids was significantly lower for every co-substrate at high initial concentration during batch 4 compared to the others. Besides, remaining sucrose (plus glucose and fructose) concentration was null after 48 hours, indicating that sucrose had been consumed but not converted into organic acids, confirming further the hypothesis of methanisation.

Sucrose was entirely converted into organic acids during the first three batchs (98-116% of initial sucrose added, Figure 38B), mostly as lactate, which confirms why sucrose is the most acidifying co- substrate. Apple had the second highest conversion into organic acid (58-73% of VS) with lactic acid as the main product (65-85% of organic acids produced in mol/mol) when minimum pH was reached. A secondary fermentation occurred when pH did not decrease below 5, with lactic acid being converted into VFAs. The phenomenon of secondary fermentation leads to a pH re-increase (McDonald, 1982), which implies that a targeted acidification meant to reach a given pH requires frequent pH monitoring in order to stop the process when the desired pH has been reach. Batch tests with carrots had the third highest conversion into organic acids (41-58% of VS). Lactate was once

129 again the main acid when minimum pH was reached but in lower proportion than sucrose and apples (30-75%).Secondary fermentation of lactate occurred systematically at medium and low initial concentration Batch tests with green beans had 41 to 48% of the added VS converted to organic acids, none of it lactate. Finally, batch tests with pea had the lowest conversion to organic acids with only 18 to 34% of the added VS converted to organic acids. The hierarchy in acidifying capacity of each co- substrate, expressed via their “sucrose equivalent” (Table 13) was confirmed by the similar ranking of their conversion to organic acid and their proportion of lactic acid.

As described in the case of pea and bean, the use of a fixed sucrose equivalent to define the acidifying capacity of an organic co-substrate relatively to sucrose was inaccurate when a low pH (4-5) was targeted, due to the higher pKa of the organic compounds produced compared to lactate. However, when biological acidification of pig slurry is used for maximum P recovery as struvite, a pH of 5.5-6 was considered as optimal by Piveteau et.al. (2017). As a result, this sucrose model expanded to organic co-substrate could still be accurate enough for limited biological acidification.

Table 15. Average relative and absolute error of the model for sucrose and each co-substrate

relative error absolute error absolute error (pH>5) sucrose 0.03 0.37 carrot 0.23 0.34 green bean 0.23 0.38 0.23 pea 0.72 0.79 0.45 apple -0.15 0.43

Figure 35. Measured minimum pH plotted against model-predicted pH for sucrose, apple, carrot, pea and green bean batch tests 130

Figure 36. Organic acids produced during biological acidification of slurry 2 with pea (A) and sucrose (B) as co- substrates

Figure 37. Minimum pH reached during biological acidification using sucrose (A) and peas (B) as organic co-substrate

Figure 38. Maximum proportion of co-substrate converted into organic acids during batch tests with high initial concentration of co-substrate (A). Mean proportion and standard deviation of initial VS converted into organic acids for each co-substrate and initial concentration (B) (slurry 4 excluded)

2. DISSOLUTION OF P, MG, CA AND N a. PHOSPHORUS

The dissolution of phosphorus was closely linked to pH in slurry 2, 3 and 4 (Figure 39B, C and D) and very loosely linked to pH in slurry 1 (Figure 39A). This difference could be explained by the relative

131 proportion of P originating from the slurry compared to the co-substrates themselves. Slurry 1 had only 230 mg-TP/L. As a result, P from co-substrates represented up to 72% of TP (Table 16). It could be postulated that orthophosphate concentration was determined by both the kinetics of hydrolysis/degradation of the co-substrates and pH (Figure 40). Indeed, as seen in Figure 40, phosphate concentration for P-rich co-substrates (in decreasing order: pea, bean, carrot) evolved similarly to a first order hydrolysis reaction. In the case of sucrose, P concentration and pH remained stable from 24h until the end, indicating logically that only inorganic P was dissolved, driven by pH. Low phosphate concentration measured in the batch tests with apple, carrots and sucrose despite the low pH value (4-5) was likely due to P incorporation during biomass growth. Indeed, considering a biomass chemical formula as C5H7NO2P0.074 (Droste, 1997), and a yield on sucrose of 0.14 (biomass COD.sucrose COD-1) (Gujer and Zehnder, 1983), 121, 93 and 65 mg-P/L were necessary at high, medium and low sucrose concentration for biomass growth, representing 52, 40 and 28% of TP in slurry 1.

In slurry 2, 3 and 4, dissolved P concentration increased with decreasing pH down to pH 5.5 and then increased slower until pH 4 in slurry 2 and 4, and decreased in slurry 3. The lower phosphate concentration in slurry 3 at pH 4-5 was associated with the high initial concentration of carrots, apple and sucrose. Consequently, biomass growth and associated need for P was high in those batch tests while carrot, apple and sucrose provided little or no P. The maximum proportion of dissolved P/TP was 85, 80, 94 and 73% in slurry 1, 2, 3 and 4 respectively (Figure 41A, slurry 1 was not included since P was not directly correlated with pH). Those values are relatively homogenous however it should be noted that the high maximum of dissolved-P / TP in slurry 1 was mostly due to P originating from peas and beans, not the slurry. This indicates that in the case of pig slurries with low P content, the choice of co-substrate should not only be guided by its acidifying capacity but also its P content.

Table 16. Proportion of total phosphorus in each batch test originating from the co-substrate expressed as percentage of TP

concentration carrot bean pea apple

High 47.3 64.8 72.2 20.1 Medium 40.9 58.7 66.6 16.2

slurry 1 slurry Low 32.7 49.9 58.3 11.9

High 10.9 14.0 26.2 3.6 Medium 7.1 9.3 18.2 2.3

slurry 2 slurry Low 3.0 4.0 8.2 0.9

High 24.0 33.9 58.2 19.2 Medium 18.1 26.6 49.5 14.4

slurry 3 slurry Low 11.4 17.5 36.5 8.9

High 7.3 12.6 15.9 3.5 Medium 4.7 8.2 10.5 2.2

slurry 4 slurry Low 1.8 3.3 4.3 0.9

132

Figure 39. Phosphate concentration plotted against pH in slurry 1 (A), slurry 2 (B), slurry 3 (C) and slurry 4 (D)

Figure 40. Percentage of dissolved P and pH plotted against time in slurry 1 at medium initial concentration

133

3- 2+ 2+ + Figure 41. Percentage of PO4 , Mg , Ca and NH4 plotted against pH

b. MAGNESIUM CALCIUM AND NITROGEN DISSOLUTION

Mg2+ concentration expressed as percentage of T-Mg increased rapidly with decreasing pH in all the slurries (Figure 41B). Below pH 5.5, between 80 and 100% of T-Mg was dissolved. In slurry 1 the concentration of Ca2+ expressed as percentage of T-Ca appeared to be both pH and time dependent. Indeed, in the case of carrots batchs, pH dropped in the three reactors during the first 24 h and remained stable afterwards, while Ca2+ concentration kept on increasing. At low initial concentration, Ca2+ increased from 60% at 24h to 80% of T-Ca at 96h (pH stable at 5.5), and at high initial concentration, Ca2+ increased from 80 to 100% of T-Ca (pH stable at 4.3). While pH dependency tends to indicate dissolution of inorganic solids and time dependency usually results from hydrolysis and degradation of organic matter, it is also possible that inorganic solids dissolution was simply not instantaneous. However, 48, 41 and 33% of T-Ca came from the carrots added to the slurry (at high, medium and low initial concentration respectively), and 103, 90, 84% of T-Ca was dissolved at the end of the batch test, indicating that a large part of the dissolved calcium came from the carrots. As a result, it seems likely that decomposition of organic matter participated in Ca2+ increase, not solely pH-driven dissolution of inorganic solids. In the other slurries, calcium dissolution was driven by pH change (Figure 41C). The trends were similar in slurry 2 and 3 with a linear dissolution from approximately 10% at initial pH, 30-50% at pH 5.5 and 50-80% at pH 4.5. The dissolution of calcium in slurry 4 was slightly different, with no dissolution from initial pH (6.5) to pH 5.5 (less than 10% of

134

T-Ca) and a sudden increase to 60% at pH 5. The inorganic, solid forms of Ca in slurry 4 may have been different than those in slurry 2 and 3, with different dissolutions constants.

Dissolution of nitrogen was not influenced by pH in any of the slurries (Figure 41D). Some mineralization of organic matter from the complex co-substrates occurred but not from the slurries themselves. Indeed, in the batches with sucrose as co-substrates, ammonia concentration never increased, while in the case of N-rich co-substrates ammonium concentration increased over time gaining progressively 10 to 30 additional percent of TN during the four days of experiment in the batch tests with peas in each slurry.

3. MOLAR RATIOS OF INTEREST FOR STRUVITE PRECIPITATION

An equimolar amount of P, N and Mg is necessary to precipitate all the phosphorus as struvite. To favor the precipitation of phosphorus as struvite and prevent calcium phosphate formation, a high molar ratio N:P (>3) as well as an excess of Mg compared to Ca are required (Capdevielle et al., 2013). Ammonium was always largely in excess, with an N:P ratio above 4.5 in all the slurries at any sampling time. Mg2+ was often the limiting ions (i.e. a molar concentration lower than phosphates), with Mg:P ratio sometimes as low as 0.5. Yet the amount of Mg necessary to match the phosphate concentration was never higher than 185 mg/L. When Mg(OH)2 was used to precipitate struvite, 3.4g/L on average were necessary to reach pH 8, i.e. 1.4 g-Mg2+/L, more than 8 times the amount of Mg required for equimolarity (see section V.4).

4. SEMI-CONTINUOUS REACTOR OPERATION AND PHOSPHORUS RECOVERY

As seen in Figure 42, pH decreased rapidly during the first 4 days (reactor operated in batch mode) to reach 5.7, and progressively decreased to 5.4 at the end of the experiment of semi continuous operation. Phosphate concentration increased quickly from 150 to 680 mg/L during the four days of batch operation and kept on increasing during semi-continuous operation, albeit at a slow pace, to reach 747 mg/L at the end. This experiment indicates a semi-continuous process for biological acidification of swine slurry with organic waste is at least as efficient as a batch process in terms of pH reached and dissolved phosphorus.

The effluent withdrawn daily during the last five days of operation of the semi-continuous reactor was used for five identical struvite crystallization experiments. Centrifugation was efficient thanks to the coagulant, with 83% (± 2.1) of the initial volume recovered in the supernatant. The detailed composition of the biologically acidified slurry can be found in Table 17. At pH 5.4, the concentration of P, Ca, Mg and N were similar to those found during the batch experiment with slurry 4, with approximately 50% of TP (including the phosphorus contained in the co-substrates). The N:P:Mg:Ca molar ratio was 6.5:1:0.9:0.4, indicating both a favorably large excess of ammonium yet also a slight deficit in magnesium and a non-negligible amount of calcium. The dried solid obtained after

135 crystallization had an Mg:N:P:Ca ratio of 1.32:1:1.19:0.05 (standard deviation of 0.1:0:0.04:0.02). The relative proportion of each compound is relatively close to struvite, with a small excess of P and Mg (the excess of magnesium could have resulted from undissolved magnesium hydroxide in the solid), and very low contamination by calcium. The proportion of struvite in the solid could be extrapolated based on ammonia concentration measured from the dissolved solid. Indeed, struvite is the only inorganic precipitate containing N that could have been formed during the pH increase from 5.4 to 8. Based on ammonia, 69.1% (± 6.9) of the solid recovered consisted in struvite crystals. The rest of the solid was most likely composed of other phosphorus precipitates, undissolved magnesium hydroxide and organic matter.

In order to precipitate phosphorus, 3.4 g-Mg(OH)2 (± 0.3 g) per liter of supernatant were necessary to reach pH 8. As a result, a large amount of Mg2+ remained in the liquid after crystallization (Table 17). Nevertheless, addition of magnesium hydroxide enabled the precipitation of 99.2% (± 0.3) of the phosphates from the supernatant (calculation based on the difference between phosphate concentration in the acidified slurry and in the liquid after precipitation and filtration).Not all precipitated P was retained by the 100 µm sieve, with 4.5g of dried solid recovered per liter of supernatant (± 1g/L), most likely due to heterogeneous crystal sizes (Figure 43). As a result, despite the complete precipitation of P, “only” 58% (± 9) of dissolved-P or 26% (± 5) of TP was effectively recovered as solids. A simple optimization of the crystallization process (e.g. mixing adjustment, rinsing of the solids to remove adsorbed organic acids, longer time allowed for crystal growth and dissolution of magnesium hydroxide) should enable a larger and more homogeneous crystal size, leading to near complete 3- recovery of the 750 mg P04 -P/L or 50% of TP with low contamination of inorganic and organic matter, only limited by the loss of the liquid trapped in the solid phase during centrifugation.

Table 17. Composition of acidified slurry and filtrate after crystallization, including mean value and standard deviation

pH P Ca Mg N mg/L

acidified mean 5.42 747 408 508 2174 slurry std 0.06 8 18 29 156

mean 8.30 6 432 893 1610 filtrate std 0.12 2 58 331 86

136

Figure 42. Phosphate concentration and pH plotted against time in the semi-continuous reactor

Figure 43. Recovered crystals from the precipitation step

VI. CONCLUSIONS

The batch experiments demonstrated that the organic co-substrates used were effective to acidify pig slurry and dissolve a large proportion of TP. The release of P, Ca and N due to hydrolysis of organic co-substrate and P assimilation for bacterial growth were non negligible phenomena in the slurries with lower solid content.

137

The semi-continuous process for biological acidification proved to be as efficient or slightly more compared to batch operation. The crystallization tests demonstrated complete phosphorus removal, resulted in a solid with low calcium contamination and an N:P:Mg ratio very close to the one of struvite. The process should be optimized to obtain larger crystals and thus a higher recovery rate. The effects of biological acidification and P-recovery on the subsequent anaerobic digestion in terms of methane potential should be investigated.

VII. FUNDING SOURCE

S. Piveteau is the beneficiary of a PhD scholarship funded equally between Irstea and the Brittany Region. This PhD is realized within the framework of the Valodim project financed by BPIFRANCE PSPC call 2013

138

VIII. REFERENCES

Abma, W.R., Driessen, W., Haarhuis, R., van Loosdrecht, M.C.M., 2010. Upgrading of sewage treatment plant by sustainable and cost-effective separate treatment of industrial wastewater. Water Sci. Technol., 61(7), 1715-1722. APHA. 1998. Standard methods for the examination of water and wastewater. American Public Health Association, Washington, DC. Barnett, G.M., 1994. Phosphorus forms in animal manure. Bioresour. Technol., 49(2), 139-147. Burns, R., Moody, L., Celen, I., Buchanan, J., 2003. Optimization of phosphorus precipitation from swine manure slurries to enhance recovery. Water Sci. Technol., 48(1), 139-146. Capdevielle, A., Sýkorová, E., Biscans, B., Béline, F., Daumer, M.-L., 2013. Optimization of struvite precipitation in synthetic biologically treated swine wastewater—Determination of the optimal process parameters. J. Hazard. Mater., 244–245(0), 357-369. Christensen, M.L., Hjorth, M., Keiding, K., 2009. Characterization of pig slurry with reference to flocculation and separation. Water Res., 43(3), 773-783. Cordell, D., Drangert, J.O., White, S., 2009. The story of phosphorus: Global food security and food for thought. Glob. Environ. Chang., 19(2), 292-305. Daumer, M.-L., Béline, F., Spérandio, M., Morel, C., 2008. Relevance of a perchloric acid extraction scheme to determine mineral and organic phosphorus in swine slurry. Bioresour. Technol., 99(5), 1319-1324. Daumer, M.L., Picard, S., Saint-Cast, P., Dabert, P., 2010. Technical and economical assessment of formic acid to recycle phosphorus from pig slurry by a combined acidification–precipitation process. J. Hazard. Mater., 180(1–3), 361-365. Desmidt, E., Ghyselbrecht, K., Zhang, Y., Pinoy, L., Van der Bruggen, B., Verstraete, W., Rabaey, K., Meesschaert, B., 2015. Global Phosphorus Scarcity and Full-Scale P-Recovery Techniques: A Review. Critical Reviews in Environ. Sci. Technol., 45(4), 336-384. Droste, R.L., 1997. Theory and practice of water and wastewater treatment. John Wiley & Sons Incorporated. Giesen, A., 1999. Crystallisation Process Enables Environmental Friendly Phosphate Removal at Low Costs. Environ. Technol., 20(7), 769-775. Gujer, W., Zehnder, A.J.B., 1983. Conversion Processes in Anaerobic Digestion. Water Sci. Technol., 15(8-9), 127-167. Günther, L., Dockhorn, T., Dichtl, N., Müller, J., Urban, I., Phan, L., Weichgrebe, D., Rosenwinkel, K., Bayerle, N., 2008. Technical and scientific monitoring of the large-scale seaborne technology at the WWTP Gifhorn. Water Pract. Technol., 3(1). Hjorth, M., Fernandez, M., Jayaram, S., Sørensen, J.A., Adamsen, A.P.S., 2015. Bio-acidification of manure–By supplying manure with 2-3% sugar or cellulose. RAMIRAN 2015–16th International Conference. TuTech Innovation GmbH, Hamburg. pp. 96-96. Liu, Y., Kwag, J.-H., Kim, J.-H., Ra, C., 2011. Recovery of nitrogen and phosphorus by struvite crystallization from swine wastewater. Desalination, 277(1), 364-369. Mattenberger, H., Fraissler, G., Jöller, M., Brunner, T., Obernberger, I., Herk, P., Hermann, L., 2010. Sewage sludge ash to phosphorus fertiliser (II): Influences of ash and granulate type on heavy metal removal. Waste Manag., 30(8), 1622-1633. McDonald, P., 1982. Silage fermentation. Trends Biochem. Sci., 7(5), 164-166.

139

Piveteau, S., Picard, S., Dabert, P., Daumer, M.-L., 2017. Dissolution of particulate phosphorus in pig slurry through biological acidification: A critical step for maximum phosphorus recovery as struvite. Water Res., 124(Supplement C), 693-701. Rahman, M.M., Liu, Y., Kwag, J.-H., Ra, C., 2011. Recovery of struvite from animal wastewater and its nutrient leaching loss in soil. J. Hazard. Mater., 186(2), 2026-2030. Schauer, P., Laney, B., 2013. Full-scale implementation of the WASSTRIP™ process: plant-wide impact of struvite recovery. in: WEF/IWA nutrient removal and recovery. Vancouver. Schuiling, R.D., Andrade, A., 1999. Recovery of Struvite from Calf Manure. Environ. Technol., 20(7), 765-768. Toor, G.S., Hunger, S., Peak, J.D., Sims, J.T., Sparks, D.L., 2006. Advances in the Characterization of Phosphorus in Organic Wastes: Environmental and Agronomic Applications. Advances in Agronomy, 89, 1-72. Vanotti, M.B., Szogi, A.A., Hunt, P.G., Millner, P.D., Humenik, F.J., 2007. Development of environmentally superior treatment system to replace anaerobic swine lagoons in the USA. Bioresour. Technol., 98(17), 3184-3194. Ye, Z.-L., Chen, S.-H., Lu, M., Shi, J.-W., Lin, L.-F., Wang, S.-M., 2011. Recovering phosphorus as struvite from the digested swine wastewater with bittern as a magnesium source. Water Sci. Technol., 64(2), 334-340. Yin, J., Yu, X., Wang, K., Shen, D., 2016. Acidogenic fermentation of the main substrates of food waste to produce volatile fatty acids. Int. J. Hydrog. Energy, 41(46), 21713-21720.

140

DISCUSSION AND PERSPECTIVES

The main objective of this PhD was to use acidogenesis, a biological process instead of chemical acidification to dissolve phosphorus in raw and digested pig slurry for struvite recovery.

Using sucrose as a model organic co-substrate, lactic acid fermentation took place in swine slurry, leading to a pH decrease proportional to the initial amount of sucrose added down to pH 4. A simple model predicting the lowest pH reachable was developed based on initial pH and sucrose concentration as well as buffer capacity of the slurry. 60 to 90% of TP was dissolved when pH decreased to the range 5-6. Because calcium dissolution was more progressive and increased down to pH 4, it could be determined that an optimal acidification for struvite recovery should not go further than 5.5, balancing the need for high phosphorus dissolution and low calcium concentration. Secondary fermentation of lactic acid toward acetate, propionate, butyrate and valerate occurred when pH did not drop below 5 during lactic acid fermentation. A pH re-increase followed, re-precipitating most of the phosphorus together with calcium.

Since the process could potentially be implemented in parallel of anaerobic digestion, biological acidification was also tested on digested swine manure and similar lactic acid fermentation took place. Various organic wastes were tested but lactic acid fermentation did not occur systematically. Only the wastes with high, easily degradable carbohydrates content led to lactic acid and very low pH, while the other co-substrates led to mild acidification with mostly acetate, butyrate, propionate and valerate as final products. The microbiological analysis using 16S rRNA gene high throughput sequencing confirmed that lactic acid bacteria (Lactobacillus) were responsible for the lactic acid production while Clostridiales dominated during more moderate acidification.

Several co-substrates were ranked in term of acidification capacity compared to sucrose in order to expand the initial model predicting lowest pH to these co-substrates. Experimental results indicated that such predictions were accurate when lactic acid production occurred (e.g. apple or carrots) or for mild acidification only (pH 5-6). Interestingly, part of the dissolved-P obtained came from the co- substrates themselves as a result of organic matter hydrolysis. Consequently, during biological acidification of pig slurry with low P content, P dissolution was only partly a function of pH, with time dependent increase or decrease of P concentration as a result of hydrolysis and biomass assimilation respectively. Therefore the P content of waste should be taken into account in low P content slurry, with longer acidification time to allow its release.

A semi continuous reactor fed with raw swine slurry, carrots and peas (HRT = 4 days) led to a progressive decrease in pH from 5.7 to 5.4 and a slight improvement in phosphate concentration from 45 to 50% of TP. While this dissolution rate is slightly lower than previous results of biological 3- acidification (60-90% TP), this corresponds to 750 mg PO4 -P/L, i.e. a very concentrated stream compared to most processes reviewed in the Context section. However, no lactic acid was present,

141 even though carrot has been shown to lead to lactic acid fermentation in batch tests. It appears likely that the 4 days retention time and limited acidification has led to secondary fermentation of lactate in the semi-continuous reactor. This reactor had been operated for 16 days only. Possibly shorter retention time or longer operation of the reactor could have led to the selection of LAB versus other acidogenic bacteria. The supernatant obtained after centrifugation was used to recover P as struvite. 83% of the volume centrifuged was recovered as supernatant thanks to an efficient coagulant. More than 99% of phosphates precipitated, corresponding to 41.5% of initial TP. The actual recovery rate was not 41.5% but 26% because some of the struvite crystals were too small to be retained by the 100 µm mesh sieve. However, a simple system with longer retention time and recirculation of the effluent should lead to near complete recovery of precipitated P. The organic content of the solid recovered was relatively high 10-20%. A washing process of the crystals with clean water might help remove some of the organic matter contamination. Even though no heavy metal analysis had been performed when the work in chapter 2 was submitted for publication, certain heavy metal contaminants have been measured since. Zinc and copper represented 152 and 157 mg/Kg-DM, below the 200 mg/Kg- DM recommended for organo-mineral fertilizer by the European Sustainable Phosphorus Platform. Legislative efforts should be pursued in order to obtain a certification for struvite fertilizer derived from animal manure. Certified struvite with lower contamination rates compared to conventional fertilizers would increase the marketability of the product and its selling price.

Current estimations put the selling price of struvite between 188 and 763 $/T (Molinos-Senante et al., 2011). The process described in this PhD would reduce the price of struvite recovery by up to 65% compared to chemical acidification. Taking into account the cost of energy, generic flocculant/coagulant and magnesium hydroxide, the total cost would be between 500 and 1500 $ per ton of struvite produced, depending on quality of chemicals used and the rebate obtained on bulk purchase. As a result, the process would be profitable if the selling price of struvite is in the upper range and if coagulant and magnesium hydroxide can be bought at minimum price. As described in section II.1. of the context (agronomic value of struvite), the physical aspect of the product obtained is critical for its re-use as fertilizer. The precipitation pilot developed by IRSTEA should be used to optimize the parameters applied to obtain a product with sufficient length and crushing strength to be directly usable as a mineral P fertilizer.

A potential source of improvement to lower the cost of struvite recovery would be to minimize the need for magnesium hydroxide. This chemical was used to bring the pH from 5.4 to 8 to enable struvite crystallization. The supernatant contained approximately 24g of organic acids per liter of acidified slurry. Because this was not a lactic acid fermentation, acetate propionate butyrate and valerate were the main organic acids, representing respectively 42, 21, 25 and 10% of all acids produced (g/g). As a result, 63 mmol or 20% of these acids were under an undissociated form at pH 5.4. Deprotonation of VFAs accounted for 70% of the acid buffer capacity contained in the

142 supernatant, with ammonium representing 9% and phosphates 21%. At such initial pH, only 3% of lactic acid would have been protonated. By optimizing HRT and cultivating progressively a biomass rich in lactic acid bacteria, the amount of magnesium hydroxide could be significantly reduced.

Further investigation should also be conducted in order to assess the impact of this P-recovery process on subsequent / parallel digestion process. It has been shown in chapter 2 and 3 that consequent hydrolysis of the co-substrate occurred. This could potentially increase the methane potential during anaerobic digestion or reduce the hydraulic retention time in the digester.

To conclude, this work has demonstrated that a large fraction of total phosphorus in pig slurry (40- 90%) could be dissolved by adding organic waste collected locally and with relatively short retention time (48 hours for sucrose, apples and carrots, 96 hours for beans and peas). Lactic acid fermentation provides the quickest and most efficient dissolution process with little acidic buffer to overcome when precipitating struvite. Implementing this P recovery process in some of the many digestion units existing in Brittany for the treatment of livestock, industrial and municipal organic waste would significantly reduce the impact of animal breeding on the local environment and provide a clean P- fertilizer for crop producers, thus participating in progressively closing the phosphorus cycle.

143

ABSTRACT

Phosphorus is a crucial nutrient for life, implicated in cellular bioenergetics as well as storage and processing of genetic information. It is also one of the limiting nutrients in agriculture with nitrogen and potassium. Since the green revolution in the middle of the 20th century, agriculture has relied on increasing amounts of cheap mineral P-fertilizers produced from a fossil resource to improve crop yields and sustain population growth. However, the resource is depleting and its use efficiency is poor: less than 20% of extracted P is actually consumed in food. One of the reasons for this is the specialization of entire regions into on type of agricultural production or another. Thus, regions focusing on high-yield crops require large applications of fossil mineral fertilizers while intensive livestock breeding areas cannot find an output for their P-rich manure due to the distance with crop fields in need of P fertilization. Over application of animal manure in Brittany is the main cause of eutrophication in the area. Phosphorus could be recovered from pig manure as struvite, a concentrated, slow-release mineral fertilizer easily transported to crop- oriented regions in need of P fertilization. P in pig slurry is mostly under a solid inorganic form, requiring dissolution prior to precipitation as struvite. Because chemical acidification is too expensive and harmful to the environment, the process developed in this PhD relied on acidogenesis, a biological process in which organic matter is converted to organic acids under anaerobic conditions, thus naturally acidifying the swine slurry. Various organic wastes were tested as organic co-substrates on raw and digested pig slurry, leading to lactic acid fermentation when the co-substrate had a high content in easily biodegradable carbohydrates and a fermentation with diverse organic acids produced at low content in easily biodegradable carbohydrates. Lactobacillus was the genus responsible for lactic acid fermentation and various Clostridiales dominated otherwise, producing acetate, propionate, butyrate and valerate. A reactor was operated with semi-continuous feeding of raw swine slurry and carrot/pea, leading to the dissolution of 50% total-phosphorus or 750 mg-P/L. After centrifugation, struvite was precipitated in the supernatant by adding magnesium hydroxide to increase the pH to 8. 99% of dissolved P precipitated. The solid recovered contained 70% of struvite, a slight excess of P and Mg as well as organic matter. Because hydrolysis of organic matter and production of organic acids occur during acidogenesis, the process could be implemented in the many anaerobic digestion units installed in Brittany treating animal manure and agricultural, industrial and municipal organic waste. The struvite recovered could be sold to regions in need while the digestate impoverished in P and rich in organic matter could be kept locally. Such process would reduce eutrophication due to over application of pig manure and reduce as well the reliance on fossil P-fertilizer by offering an alternative source with equivalent fertilizing performances.

Keywords: phosphorus recovery, struvite, swine manure, pig slurry, lactic acid fermentation