SUMMARY
Phosphorus, which is currently being extracted from mines in the form of phosphate rocks, is a finite resource unable to maintain in the long term an extraction level as at present. Due to the fact that much of the phosphorus used as fertilizers, foods, and cleaning products reaches wastewaters, the key to minimizing the extraction consists of the recovery of the phosphorus present in those wastewaters. At present, the phosphorus, after being removed chemically or biologically from the wastewater, is lost by incineration or deposition except when the biological sludge is used for composting.
Struvite crystallisation, ammonium magnesium phosphate crystallisation, appears as one of the most promising processes to recover phosphorus from wastewaters. The application of struvite crystallisation processes at full scale in a Wastewater Treatment Plant (WWTP) not only removes phosphorus from the process but allows to recover it in a form that could be reused as a fertilizer or as a raw material in the phosphorus industry.
The main objective of this PhD thesis has been to understand the struvite crystallisation principles with the aim to improve the phosphorus recovery as struvite from the sludge lines supernatants of a WWTP. Due to the difficulty to model all the processes involved in the struvite precipitation, the supersaturation distribution inside the reactor, the dominating nucleation mechanisms at a certain operational conditions, and the phosphate thermodynamics and chemistry, it has been necessary to carry out some experiments to verify the feasibility of working with an agitated reactor as the one employed in this work. Firstly, solutions prepared in the laboratory were employed to ensure some theoretical aspects of struvite precipitation. Afterwards, some experiments were carried out employing the supernatants generated in the anaerobic digestion pilot plant located in the Cuenca del Carraixet WWTP, Valencia. 
The results obtained from the experiments working with solutions prepared in the laboratory showed that it was possible to achieve phosphorus precipitation and recovery efficiencies up to 80%. It was observed that, as the pH in the reactor and the Mg/P and N/P molar ratios in the influent increased, both efficiencies increased, whereas a decrease in the hydraulic retention time had no effect on the efficiencies achieved. The pH at which the reactor operated had a great influence on the process. A fuzzy logic control algorithm was developed with the aim to control the pH. It allowed high stable pH conditions in the reactor to be achieved. 
In the working conditions, the reactor employed was able to retain the majority of the solids formed inside. Therefore, its design proved to be adequate from the phosphorus recovery point of view. Nevertheless, a slight loss of phosphorus in the form of small crystals was observed in all the experiments carried out. This fines formation was related directly with the presence in the reactor of zones of high local supersaturation, as no relation was found between the recovery efficiency, the particle size obtained and the quantity of phosphorus lost with the effluent with the global supersaturation inside the reactor. Struvite precipitation at the exit of the reagent tubes confirmed that a high local supersaturation existed in those points. This high local supersaturation promoted primary nucleation instead of crystal growth. Sometimes, due to the proximity of the feed points, this local supersaturation was so high that it caused fouling in the reactor. Fouling reduced the precipitation and recovery efficiencies and promoted the formation of fine crystals. Furthermore, the reactor got soiled which was an additional operational problem. Therefore, it was important to pay attention, when applying the process at full scale, to the way and place of discharge of the reagents in order to optimize the process and to reduce the problems associated with fouling. 
The presence of calcium in the precipitant solutions caused the formation of a mixture of struvite and amorphous calcium phosphate, and a higher loss of phosphorus with the effluent when working with high calcium concentrations in the influent. The same results were obtained when working with the supernatants generated in the anaerobic digestion plant. 
Precipitation and recovery efficiencies values of up to 95 and 87% respectively were obtained when the precipitant solutions were the supernatants of the anaerobic digestion process. The precipitates formed were mainly struvite, followed by amorphous calcium phosphate and, in some experiments, by calcite. The aeration of the supernatants allowed to achieve a pH value of 8,7 due to CO2 desorption. The control algorithm developed also showed its feasibility when the pH adjustment was carried out by aeration instead of by sodium hydroxide addition. Aeration caused higher phosphorus loss with the effluent, although on the other hand it allowed struvite to be obtained clean of suspended solids, making the formed struvite crystals easier to separate and clean. 
Lastly, the crystals obtained showed the typical tubular struvite morphology. The main difference found between the crystals formed from the supernatants generated in the anaerobic digestion plant, and the ones obtained from solutions prepared in the laboratory, was the absence of agglomeration in the formers. In general, fine needled crystals were not found when working with the supernatants, instead a rectangular shape similar to a trapezium was obtained. 
This thesis work implies a better understanding of the struvite crystallisation process. It has allowed to conclude that this is a very sensible process to changes in pH, characteristics of the supernatants, the presence of impurities and the inadequate initial mixing of reagents. Therefore, once the principles of the process had been established, it was expected that the results obtained in this work would serve as a guide to optimize the process at full scale.