• Aucun résultat trouvé

Optimization of struvite precipitation in synthetic biologically treated swine wastewater - Determination of the optimal process parameters

N/A
N/A
Protected

Academic year: 2021

Partager "Optimization of struvite precipitation in synthetic biologically treated swine wastewater - Determination of the optimal process parameters"

Copied!
14
0
0

Texte intégral

(1)

O

pen

A

rchive

T

OULOUSE

A

rchive

O

uverte (

OATAO

)

OATAO is an open access repository that collects the work of Toulouse researchers and

makes it freely available over the web where possible.

This is an author-deposited version published in :

http://oatao.univ-toulouse.fr/

Eprints ID : 9736

To link to this article : DOI:10.1016/j.jhazmat.2012.11.054

URL :

http://dx.doi.org/10.1016/j.jhazmat.2012.11.054

To cite this version :

Capdevielle, Aurélie and Sýkorová, Eva and Biscans, Béatrice and

Béline, Fabrice and Daumer, Marie-Line Optimization of struvite

precipitation in synthetic biologically treated swine wastewater -

Determination of the optimal process parameters. (2013) Journal of

(2)

Optimization

of

struvite

precipitation

in

synthetic

biologically

treated

swine

wastewater—Determination

of

the

optimal

process

parameters

Aurélie

Capdevielle

a,d,∗

,

Eva

S ´ykorová

b

,

Béatrice

Biscans

c

,

Fabrice

Béline

a

,

Marie-Line

Daumer

a

aIRSTEA/Cemagref,17avenuedeCucillé,CS64427,35044RennesCedex,France

bICTPrague,DepartmentTechnologyofwater,Technická5,Prague6,16628,CzechRepublic

cUniversitédeToulouse,LaboratoiredeGénieChimique,UMR5503CNRS/INP/UPS,SitedeLabège,BP84234,CampusINP-ENSIACET,4alléeEmileMonso,31030ToulouseCedex4,

France

dUniversitéeuropéennedeBretagne(UEB),5BoulevardLaënnec,35000Rennes,France

h

i

g

h

l

i

g

h

t

s

StruviteprecipitationismainlyinfluencedbythequantityofMgOadded.

Optimizedparametersfavouredstruviteformationdespiteofhigh[Ca2+]withoutaddingotherreagents. 90%ofthetotaldissolvedphosphorusisrecoveredaslargecrystalsofstruvite.

Ramanspectroscopyandsoliddissolutionshowco-precipitationofACPandpresenceofCaCO3insolid.

Keywords: Struvite P-recovery MgO Calciumphosphate Swinewastewater Raman

a

b

s

t

r

a

c

t

Asustainablewaytorecoverphosphorus(P)inswinewastewaterinvolvesapreliminarystepofP dis-solutionfollowedbytheseparationofparticulateorganicmatter.Thenexttwostepsarefirstlythe precipitationofstruvitecrystalsdonebyaddingacrystallizationreagent(magnesia)andsecondlythe filtrationofthecrystals.Adesignofexperimentswithfiveprocessparameterswassetuptooptimize thesizeofthestruvitecrystalsinasyntheticswinewastewater.Morethan90%ofPwasrecoveredas largecrystalsofstruviteinoptimalconditionswhichwere:lowMg:Caratio(2.25:1),theleading param-eter,highN:Pratio(3:1),moderatestirringrate(between45and90rpm)andlowtemperature(below 20◦C).Theseresultswereobtaineddespitethepresenceofalargeamountofcalciumandusingacheap reactant(MgO).ThecompositionoftheprecipitateswasidentifiedbyRamananalysisandsolid disso-lution.Resultsshowedthatamorphouscalciumphosphate(ACP)co-precipitatedwithstruviteandthat carbonateswereincorporatedwithsolidfractions.

1. Introduction

Thesoilin Brittany(France)isextremelyrichin phosphorus (P)becauseofintensivepigfarming.Pmovesintosurfacewaters duetosoilerosion,andtakespartintheireutrophication.Whilst needsofPasfertilizerincreaseeveryyear,phosphatereservesare estimatedtorunoutin100years[1].Phastoberecycledfrom ourwastes. Therefore, someprocesses havebeen developed to recycledissolvedPasmineralfertilizerfromurbanwastewater. Calciumphosphateandstruvitearethetwoformsofphosphate fer-tilizerproduced.Struviteisknownasaslowreleasefertilizerwith

∗ Correspondingauthorat:IRSTEA/Cemagref,17avenuedeCucillé,CS64427, 35044RennesCedex,France.

E-mailaddresses:aurelie.capdevielle@irstea.fr(A.Capdevielle),

eva.sykorova@vscht.cz(E.S ´ykorová),beatrice.biscans@ensiacet.fr(B.Biscans).

comparableperformancesto,orevenhigherthan,superphosphate fromore[2].However,struviterecoveryprocess,thatrecyclesa significantamountofP,isseldomappliedtoconcentratedeffluents suchasswinewastewaters.Oneofthelimitationsistheformthat Pactuallypresentsinswinemanure:80%ofPisintheparticulate part[3].

AsustainablewaytorecoverPinswinewastewaterinvolvesa preliminarystepofPdissolutionatpH4.5followedbythe separa-tionofresidualparticulateorganicmatter.Lessreactantisneeded whenbuffereffectisdecreasedbypreviouslytreatingthe efflu-entfornitrogenremoval.Thenexttwostepsare:theprecipitation ofstruvitecrystals(MgNH4PO4·6H2O)byaddingacrystallization reagent(magnesia,MgO)andthefiltrationofstruvitecrystalsin filterbagswitha100mmcut-off[4].

MgOwaschosenastheprecipitationreagentasitscostislow, itisaby-productfromtheanimalfoodindustry,itissafeanditfits toagriculturalrestrictions.Ithastwomainfunctions:toincrease

(3)

thepHvalue(6.5–11)andtogetamolarratioofMg:N:Pequalto 1:1:1inordertoreachtheconditionstoprecipitatestruviteinthe solution[5,6].Nevertheless,thecontrolofthepHvalueandthe molarratioofMg:Cacannotbecontrolledindependentlyofeach other.

Thecontrolofthemechanismsofprecipitationisthemainissue toobtain:highvalue-addedproducts,predictablemineral compo-sitionandbigstruvitecrystals.

Theamountandsizeofcrystalsdependonnucleationrateand crystalgrowth.Thesetwomechanismshavealreadybeenmodelled

[7,8].Nucleationrateisdescribedby(1). J1,hét[nb m−3s−1]=A1,hétexp −f.˚v V2mg3g,l (kT)3ln2b

!

(1) b = aMg2+× aNH+ 4 × a PO3− 4 Kstruvite (2) SI= log(b) (3)

A1,het is the nucleii agglomeration rate (nb.m−3.s−1), Vm

is the nucleii molar volume (m3), k the Boltzmann

con-stant=1.38×10−23J.K−1, T the temperature (K), gg,l the

nuclei/liquid interfacial energy (N.m−1), b the

supersatura-tion(2) andfand

Ф

v areshapefactors.ai aretheionactivities

andKstruvite is thesolubilityproduct of thestruvite. SI (3),the

saturationindexisusedtopredicttheprecipitationpotentialof struvite.bisstronglydependantonthepHduetothedissolution ofMgO,tothepKaofthetriproticphosphoricacid(2.12,7.21and

12.67at25◦C)andthepK

aofNH3/NH4+(9.25).

Thegrowthrateofacrystalisdescribedbyatwostepmodel: thetransportofthesolutestothecrystalsandtheirintegrationinto thecrystalstructure[7–10].

The diffusion-controlled (Gd), theintegration-controlled (Gr)

andtheoveralllineargrowthrate(Gg)aredescribedbythe

fol-lowingequations(4),(5),(7)[7–9].GgisthesumofGdandGrand

canbeexpressedinfunctionofthemassfluxdensity(dm/dt)(7). Gd= L× A 3× Vp× rc × kd× (c− ci) (4) Gr= L ×A 3× Vp× rc× kr× (ci− c∗) r (5) kr= kr0× exp



−Er RT



(6) Gg= L ×A 3× Vp× rc × kg× (c− c∗)g= L 3× Vp× rc ×dm dt (7)

Wherekd(4)isthemasstransfercoefficient,kr(6)isthe

reac-tionrateconstant,kgisthegrowthrateconstant,ristheorder

oftheintegrationreaction,kr0isthereactionconstant,1Eristhe

activationenergyandg(7)istheoverallorderofthegrowth pro-cess(g=1athighsupersaturationand2atlowsupersaturation). c,ciandc*aretheconcentrationofthesolute,itsconcentrationat

thecrystal-solutioninterfaceanditssupersaturatedconcentration respectively.cciisthedrivingforcefordiffusionandcic*isthe drivingforceforreaction.Listheaveragediameterofthecrystal;A istheareaoftheparticle,Vpitsvolumeandrcthecrystaldensity;

misthemassofthestruviteformed.

In acontinuousreactor,thestruvite growthseemstofollow asizedependentgrowthmodel[11]:thelargerthecrystalsare, thefasterthecrystalgrows.Nucleationandgrowthmechanisms arestronglydependentonpH,supersaturation,temperature,ionic strengthofsolution,presenceofforeignsubstances,designof crys-tallizer,stirring,residencetimeofcrystals[7,8].Whentheinfluent containsbothcalciumandmagnesium,apHrangeof7to11enables theprecipitationofbothstruviteandcalciumphosphate[12–16].

Table1

Ionicconcentrationsofsyntheticandrealeffluents.

Ionicconcentrations(mgL−1) Realeffluent Syntheticeffluent

Na+ 524± 163 2083± 152 K+ 1993 ± 453 1559± 120 Mg2+ 361± 35 345 ± 19 Ca2+ 694± 108 645 ± 38 Cl− 1277 ± 262 2198± 40 NO2−-N 356± 87 293 ± 3 NO3−-N 68± 10 115 ± 2 PO43−-P 654± 232 629 ± 3 SO42−-S 113± 31 111 ± 1

Competitionbetweenthetwoproductswilldependonseveral fac-torsamongstwhichtheMg:CaandN:Pmolarratios[17–19].

Mostofthestudies[20–22]onstruviteandcalciumphosphate crystallizationhave beenperformedonpure syntheticmedium containingonlytheionsdirectlyinvolved in theircomposition. Realinfluentshavealsobeenstudiedbutwithlowerionicstrength comparedtotheswinewastewaterafterPdissolution.In these studies,magnesiumandpHincreasewereusuallyprovidedby solu-blereactants[21,23,24].Magnesiadissolutionkineticandammonia volatilizationwerenottakenintoaccountwhiletheydodetermine theMg:CaandN:Pmolarratios.

Thisstudyfocusesontheinfluenceoftheprocessparameters ontheprecipitation of struvite in complexmedium likeswine wastewater.Precipitation was carriedout in syntheticmedium withanioniccompositionclosetotheacidifiedswineslurry enter-ing thecrystallization step ofthe process. Thepurpose was to improvetheprecipitationofstruviteinsteadofcalciumphosphate ina stirredbatchreactor.Thestudiedprocessparameters were stirringrate,temperature,Mg:Caand N:Pmolarratio.Adesign ofexperimentswassetuptounderstandtheinfluenceofthese variablesoverstruviteprecipitationandcrystalsize.Thesurface responsewasemployedtooptimizethequantityoflargecrystals ofstruvite.

2. Experimental

2.1. Syntheticswinewastewaterpreparation

40Lofsyntheticswinewastewaterwerepreparedbymixing differentsaltsindistilledwatertoobtaintheionicconcentrations observedinthestudiedswinewastewater(Table1).Formicacid (85%,CarlosErba)wasaddedinthesolutionuntilpHreached4.5.

Thesaltsusedfor syntheticeffluentpreparation were:KOH, K2SO4,NaNO2,KNO3,CaCl2·2H2O(CarloErba), MgCl2·6H2Oand Na3PO4·12H2O(Merck).

Theionicconcentrationsofsyntheticswinewastewaterwere controlledbyionchromatographyonDionexDX-120. Concentra-tionsforsyntheticandrealeffluentsareshowninTable1.

TheadditionofavariableamountofNH4Clpowder(99%,Carlo

Erba)wasperformed5minbefore thebeginningofeach runto minimizethepossiblevolatilizationofNH3.Theadditionwas

per-formedunderhighstirring. 2.2. Magnesiasuspension

Previousexperiments(datanotshown)showedthata suspen-sionofMgOinwaterwasmuchmoreeffectivethanMgOpowder directlyintroducedintothereactor.MgOsuspensionwasbetter dispersedanditsreactivitywasimproved.Thiskindofsuspension wasalsousedinasimilarstudybyMünchetal.[25].TheMgO sus-pensionwaspreparedbymixing500gofMgOpowder(97%,VWR Prolabo)in5Lofdistilledwater.Magnesiawaskeptinsuspension undercontinuousstirringat300rpmandwasthermostatedatthe

(4)

sametemperatureastheruns.Therunsbeganwhenaddingthe suspensionofMgO.

2.3. Equipment

TherunswereperformedinaFC6sflocculatorfromVelp Sci-entifica.Sixaxialmixingbladesareoperatedbysixmotorswith variousstirringratesfrom10to200rpm.Eachbeakerwasfilled with800mLofsyntheticeffluent.TheadditionofMgOwas car-riedoutundercontinuousstirring at90rpmfor 30s. Then, the stirringratewasmaintainedattheconsignedvalueduring24h. Therunswerecarriedoutinvesselsthermostatedbywater circu-latingthroughaconstanttemperaturebath(5◦C,20Cor35C).

pHand temperaturewererecordedbya8channelsdatalogger PH/MV/ION/O2fromFisherScientific.

2.4. Sampling

Samples(4mL)werecollectedfrombeakersevery30min dur-ing4handthenafter24hfordissolvedcompoundanalysis.At4h and24h,a250mLsamplewascollectedforparticleanalysis.

After24h,samplesweresievedon25and100mmfilters.The threesolidfractions(<25mm,25–100mmand>100mm)weredried atroomtemperature.Eachfractionwasthenweighedandanalyzed asdescribedbelow.

2.5. Analysis

Thecompositionofthecollected fractionswereanalyzedby ionchromatographyaftersoliddissolutionbyformicacidandby Ramanspectroscopy.Theprecipitationratewasevaluatedby mon-itoringthedissolvedconcentrationsofCa2+,PO

43−,Mg2+andNH4+. 2.5.1. Ionicconcentration

TheconcentrationsofdissolvedPO4-P,NH4-Nwereanalyzedby

automatecolorimetricmethodsonQuikChem®FIA+fromLachat InstrumentswithQuikChemmethod10-115-01-1-Pand 10-107-06-1-Jrespectively.ThedissolvedMg2+andCa2+weredetermined

bycationchromatography.

Thethreesolidfractionswereacidifiedwithformicacid(85%, CarloErba) and dilutedin 50mLof distilled water. Theywere analyzedbycationandanionchromatographyonDionexDX-120 withanIonPacTMCS12AcolumnandanIonPacTMAS9-HCcolumn

respectively.

Themolarbalancewascheckedbyanalysingthesolidandliquid compositionatthebeginningandtheendoftherun.

2.5.2. Ramanspectroscopy

AnalysiswerecarriedoutatroomtemperatureusingaRaman microscope(KaiserOptical Systems,Leica DM1)equipped with a thermoelectrically cooledCCD detector coupled to the spec-trometer with multi-mode fibre optic cables; 50mm diameter wasemployedforexcitation,and62.5mmdiameterforcollection. Theexcitation ofRamanscatteringwasoperated bya 400mW diodelaseroperatingat785nm,withanaverage poweroutput of100–200mWthroughanobjectivelenswithamagnificationof 20Xandaworkingdistanceof11mm.Thespotsizeofthelaser beamthroughtheobjectivewasapproximately20mm.About1mg ofsolid wasintroduced onamicroscope slide.Themicroscope wasfocusedvisually,usingawhitelightsourcetomaximizethe Ramansignal.Thecollectionrangewas100–3100cm−1.Thetime

ofcollectionwasthenadjustedat2mininordertoobtainagood signal-to-noiseratio.

2.6. Experimentaldesign

TheresponsesurfaceplanusedisaBox–Behnkenmodifiedplan with48runs:12centrepointsandonerepetitionforthe extrem-ities.Fiveprocessparameterswerestudied:A,thestirringrate,B, thetemperature,C,theN:Pmolarratio,D,theinitialMg:Camolar ratioandE,theconcentrationofmagnesiumremainingasMgO particles.

DandEwereoriginallygroupedinoneparameter:theMg:Ca molarratio.However,themeasurementofdissolvedMg2+ inthe

MgOsuspensionshowedvariationsatthebeginningofeachblock ofruns.Tobetterdescribetheresultsandunderstandtheeffects, MgOwasconsideredastwoparametersforstatisticaltreatment oftheexperimentaldata:theinitialdissolvedMg:Caratioandthe remainingMg2+ asMgO,[Mg-MgO].Thesetwoparameterswere

calculatedfromthedissolvedMg2+measuresinMgOsuspension.

Thecorrelationmatrixoftheexperimentaldesign showedthat, evenwiththischange,therewasnoriskofconfusion.

Thedesignmatrixandexperimentallevelofeachvariableare describedin

Table2.Thestudiedresponseswere:ionicconcentrations(Ca2+,

PO43−andNH4+)after24h,theproductionofparticles>100mm

andtheproportionofstruviteversuscalciumphosphateindried solids.

2.7. Dataevaluation

TheBox–Behnkenmodifiedplanusesasecondorderpolynomial model(8)topredicttheionicconcentrationvaluesasafunction ofacombinationoftheprocessparameters(stirringrate,initial dissolvedMg:Caratio,[Mg-MgO],N:Pratioandtemperature)[26]. Y= b0+ 5

X

i=1 biXi+

X

i<j bijXj+ 5

X

i=1 biiX2i + « (8)

WhereYisthepredictiveresponse.Inthispaper,sixresponses werepredictedandfivewereusedfortheoptimizationstep.b0is

theconstantterm.biarethecoefficientsofthelinearparameters

Xi(A,B,C,D,E).bijrepresentsthecoefficientsoftheinteraction

parametersXiandXj(forinstance:AB,AC,AD...).biirepresents

thecoefficientsofthequadraticparametersXi2 (AA,BB,CC,DD

andEE)whichshowthequadraticdependenceoftheresponsesto aparameter.

Statgraphics Centurion XVI, 16.1.11 software was used for experimentaldesignanddataevaluation.Thestatisticalmodelsare saidtofittheexperimentaldatawhentheadjustedR2weregreater

thanorequalto65%.AlowprobabilityFvalue(“Prob.>F”)lessthan 0.05indicatesprocessparametersaresignificant.

2.8. DeterminationofthefinalsolidphasesbyPHREEQC®

Thesaturationindex(SI)ofthevariousmineralswascalculated usingthegeochemicalsoftwarePHREEQC®withtheMinteq.V4 database.Thedatabasewasmodifiedtoincludethestruviteand amorphouscalciumphosphate(ACP)phases.TheKsp valuesfor

struvitewereadaptedtothetemperatureoftherunsaccordingto Hanhounetal.[27].TheKspvalueforACPwassetat26.83,themean

valueoftheKspforACPfromallthestudiesdescribedbyMa ˜nasetal.

[28].TheKspforbrucitewassetat17.01accordingtothestudy of Altmaieretal.[29].Severalminerals wereremovedfromthe databaseastheycannotprecipitatethroughtheconditionsofthe runs:themonetite(CaHPO4),themagnesite(MgCO3),thearagonite

(CaCO3), thedolomite(CaMg(CO3)2,thehuntite(Mg3Ca(CO3)4),

the b-tricalciumphosphate (Ca3(PO4)2) as they can only form

athightemperature[30,31],thehydroxyapatite(Ca5(PO4)3(OH))

(5)

Table2

ThemodifiedBox–Behnkendesignforoptimizationof5variableseachatthreelevels.

Runs Stirringrate(rpm) InitialMg:Camolarratio [Mg-MgO](mgL−1) N:Pmolarratio Temperature(C)

Low 10 0.63 475 0.71 5 Middle 45 1.06 1228 2.03 20 High 90 1.66 2617 3.11 35 1 10 1.08 475 2.86 5 2 45 1.10 475 1.96 5 3 90 1.59 797 2.00 5 4 10 1.66 797 1.00 5 5 10 1.66 797 3.00 5 6 45 1.57 797 2.00 5 7 45 0.84 1607 2.00 5 8 45 1.11 975 2.00 5 9 45 1.14 1606 2.00 5 10 45 1.09 975 3.00 5 11 45 1.09 975 2.00 5 12 90 0.71 975 3.00 5 13 10 1.02 1657 2.00 20 14 10 1.09 1026 3.00 20 15 10 1.11 1026 2.13 20 16 45 1.02 500 1.03 20 17 45 0.99 1044 2.14 20 18 45 1.00 1044 1.00 20 19 45 1.13 991 1.89 20 20 90 1.15 991 2.31 20 21 45 0.91 1004 0.90 20 22 45 1.02 1635 2.35 20 23 45 1.08 1004 2.20 20 24 45 1.07 478 2.15 20 25 45 0.81 1052 2.14 20 26 45 0.80 1683 2.10 20 27 45 1.05 1052 1.93 20 28 90 0.80 1052 1.95 20 29 90 0.92 552 1.92 20 30 45 0.95 1085 2.14 20 31 45 0.63 552 2.14 20 32 45 0.85 1085 2.89 20 33 45 0.90 1087 2.25 20 34 90 0.91 1718 2.14 20 35 45 0.97 1087 2.13 20 36 45 0.85 1087 1.19 20 37 10 0.94 1637 2.48 35 38 90 0.90 1637 0.71 35 39 45 1.00 2608 2.86 35 40 45 1.00 1160 1.96 35 41 45 0.94 872 2.00 35 42 45 0.96 1637 1.00 35 43 90 1.18 1613 3.00 35 44 10 1.18 1613 2.00 35 45 45 1.43 2618 2.00 35 46 45 1.30 2159 2.00 35 47 45 1.35 2324 2.00 35 48 45 1.18 2324 3.00 35

accordingtotheOstwald’srules,thelessstableforms(amorphous calciumphosphate(ACP)ordicalciumphosphatedihydrate(DCPD, CaHPO4·2H2O))precipitatefirst[7,12,32,33].TheSI valueswere calculatedusingtheinitialconcentrations.Theconcentrationof Mg2+wasdefinedbythetotalMgpotentiallydissolved.The

con-centrationofcarbonatesadjustedthechargebalance.WhenSI>0 (>0.5withthesecuritymargin[28]),thesupersaturationisreached andmineralsareabletoprecipitate.

3. Resultsanddiscussions

The experimental design evaluates the influence of process parametersonP-recoveryaslargestruvitecrystals.

Ineachrun,themolarbalanceswerecontrolledforphosphorus, nitrogen,calciumandmagnesium.Eachmolarbalancewasasum of4or5analyticaldata.Theammoniavolatilizationwascalculated thankstotheliquid/solidmolarbalanceforammonium(12)inpart 3.1.3.Errorsinmolarbalancewerebelowthesumofexperimental errors.

3.1. Phosphateremovalconsideringammoniavolatilization 3.1.1. Phosphateremovalismainlyinfluencedbystirringrate andadditionofMgO

Theevolution ofthephosphate concentrationfollows afirst orderkinetic(9)[34].

−d[PO4− P]t

dt = K ([PO4− P]t− [PO4− P]e) (9) Theexpressionof[PO4-P]t(10)istheresultoftheintegrationof

(9).

[PO4− P]t=([PO4− P]e− [PO4− P]0) ×(1− e−Kt)+ [PO4− P]0

(10) Where[PO4-P]t is theconcentrationof thetotally dissolved

phosphateatt.[PO4-P]0and[PO4-P]earetheconcentrationbefore

theadditionofMgOandatequilibriumrespectively.Kisakinetic constant.Forrunsat10rpm,theconcentrationsofthefinalPO4-P

(6)

Fig.1. TheevolutionofPO43−-Pfollowsafirstorderkinetic,temperature=5◦C,

initialMg:Ca=1.2–1.7,[Mg-MgO]=800–1000mgL−1.

werebelowthelimitcalculatedbythefirstorderkineticequation (Fig.1,1).At4and24h,sampleswerecollectedfor morphogran-ulometriccharacterization.Thestirringratewasincreasedduring 5mininordertocollectahomogeneous quoteofparticles.This stirringperturbedtheresultsofthe10rpmstirredruns.Therefore, theconcentrationsofthefinalPO4-Pwerecalculatedthankstothe

limitof(10)whichequalsto[PO4-P]e.[PO4-P]ewasthenchosento

evaluatetheinfluenceofprocessparametersonPremoval. Over48runs,44weredescribedby(10)(meanR2was0.9998

andstandarddeviationwas2.47mgL−1).4runsweredescribedby

anaffinefunctionbecausetheconcentrationsofPO4-Pat30min

werebelowthequantificationlimit(10mgL−1)andmorethantwo

pointsareneededtoevaluate[PO4-P]e.Therefore,thestudywas

performedonthe44runsdescribedby(10).

AftertheadditionofMgOinthebeakers,thepHvalueincreased from4.5to7–8inthefirstmin.ThepHvalues remainedat7–8 during15–25minutes.Afterwarditincreasedfrom7–8to9–10. Fortherunsat5◦C,itremainedconstantuntiltheendoftheruns.

Fortherunsat20◦Cand35C,thepHdecreasedslightlyby0.5

to1unitafter12h(Fig.2).Thetemperatureandthe concentra-tionofMg2+ remainingasMgO([Mg-MgO])influencedthefinal

pHvalues.FinalpHwasabout10–10.5with800mgL−1 of

[Mg-MgO]at5◦C,about9.5with1050mgL−1at20Candabout9with

1700mgL−1at35C(Fig.3).ThefinalpHvaluedependsonvarious

phenomena,NH3volatilization,incorporationofcarbonates,MgO

dissolution...,whichweredependentonthefiveprocess parame-tersstudied.

Fig.2.ThepHfunctionoflogarithmbase2ofthetimeevolveddifferentlyfunction oftemperature–5◦Crun10,20Crun30,35Crun42.

Fig.3.ThefinalpHisstronglyrelatedto[Mg-MgO]andtothetemperature.pH reachesitsmaximumwith800mgL−1of[Mg-MgO]at5C,with1050mgL−1at

20◦Candwith1700mgL−1at35C.

Table4

ANOVAforthefinalconcentrationofphosphates.

Processparameters Prob.>Ffor[PO4-P]e

AdjustedR2 66.28%

A:Stirringrate 0.0003(–)

B:Temperature 0.2083(–)

D:initialMg2+:Ca2+molarratio 0.2158(–)

E:Mg2+asMgO 0.0000(–)

AA 0.0002(+)

AB 0.0005(+)

EE 0.0000(+)

(+)indicatesa positiveinfluenceand (–)anegativeinfluence onthestudied responses.

[Mg-MgO]andstirringrateweretheleadingprocessparameters whichhad anegativeinfluenceon[PO4-P]e andconsequentlya

positiveinfluenceonPremoval(Table4).

Theinteractionbetweenthetemperatureandstirringratewas significant.Whateverthestirringrate,Pwasentirelyremovedat theend of theexperiments at 35◦C, while [PO

4-P]e wasup to

450mgL−1at5Candastirringrateof10rpm(Fig.4).

When the stirringrate wasat 10rpm, the MgO settledand onlythesurfaceofMgOcouldreact.Thecrystallizationofstruvite wasloweredandthepHdidnotincrease.Thiseffectwaspartially counterbalancedbytheincreaseoftemperaturethatimprovedthe diffusionandthedissolutionofMgOin themedium.Therefore,

Fig.4. Temperatureandstirringratenegativelyinfluencethefinaldissolved phos-phate[PO43−-P]e,thegridrepresentsthesurfaceresponseofthestatisticalmodel.

(7)

Fig.5.[Mg-MgO]andstirringratenegativelyinfluencethefinaldissolved phos-phate[PO43−-P]e,thegridrepresentsthesurfaceresponseofthestatisticalmodel.

[PO4-P]eremainedhighforthe10rpmstirredruns(Fig.5).The

pos-itiveeffectoftheconcentrationofmagnesiumwasalsoobserved inliteraturewiththeadditionofMg(OH)2[25]orMgCl2plusNaOH

[15].

3.1.2. PhosphateremovalasstruviteismainlyinfluencedbyN:P ratio

Pwassupposedtoformeitherstruviteorcalciumphosphate whichwasconfirmedbyRamananalysis(part3.3.3).The percent-ageofphosphateasstruviteiscalculatedaccordingto(11)where MNandMP arethemolecularmassofnitrogenandphosphorus

respectively.[NH3-N]volatilizediscalculatedusing(12)inpart3.1.3.

%PO4− PStruvite

= ([NH4− N]initial− [NH4− N]final− [NH3− N]volatilized) ×MP MN× [PO4− P]initial

(11) TheANOVAinTable5indicatesthatN:Pmolarratio(p<0.0001) andtemperature(p<0.01)weretheleadingprocessparametersthat influencedtheproportionofPasstruvite.

%PO4-PStruviteincreasedfrom20%to90%withanincreaseinN:P

molarratiofrom1to3.Abbonaetal.[13]showedthatahighMg:Ca ratio(4:1)combinedwithalowpH(around7)improvedstruvite formationovercalciumphosphateformation.Inthisexperimental design,theminimumMg:Camolarratiowassetat2.25toincrease thepHvaluetoatleast7.Itappearedthatabovethisratio,theMg:Ca molarratiowasnotthelimitingfactortostruviteformation,itwas theN:Pmolarratio.

TherearetwopossibleexplanationsfortheinfluenceoftheN:P molarratioontheproportionofPasstruvite.Firstly,NH4+

partici-patestostruviteprecipitation:thesupersaturationincreasedwith Table5

ANOVAfortheproportionofPasstruvite(%PO4-Passtruvite).

Processparameters Prob.>Ffor%PO4-Passtruvite

AdjustedR2 65.27% B:Temperature 0.0089(−) C:N:Pmolarratio 0.0000(+) BB 0.0740(–) BC 0.1000(–) CC 0.0472(–)

(+)indicates apositiveinfluence and(–)anegative influenceonthestudied responses.

Table6

ANOVAforthepercentageofammoniavolatilization.

Processparameters Prob.>Ffor%NH3volatilization

AdjustedR2 87.28% B:Temperature 0.0006(+) E:Mg2+asMgO 0.0000(+) BB 0.0005(–) BE 0.0004(+) EE 0.0002(–)

(+)indicatesapositive influenceand(–) anegativeinfluenceonthestudied responses.

theN:Pmolarratioandthenucleationrateofstruvitewasspeeded up(1).Secondly,NH4+improvesthebuffercapacityofthesolution

[18,35,36].ThenucleationofstruviteoccursatalowerpHthanthe oneneededfortheprecipitationofcalciumphosphate[13]. Conse-quently,struviteprecipitationisfavouredbyahighconcentration ofNH4+.However,asstruviteprecipitates,NH4+bufferingcapacity

decreases.Therefore,theeffectofN:Pmolarratioonthebuffering capacitycannotbeobservedjustbymonitoringthepH.

%PO4-PStruvite decreasedfrom80% to50%when temperature

rosefrom5to35◦C.

Withhightemperature,thepKaofNH4+/NH3decreases

favou-ring the NH3 form and NH3 volatilization as described below.

Thereforethesupersaturationofstruvitedecreasesandthe phos-phatesareabletoprecipitateascalciumphosphate.

3.1.3. Ammoniavolatilizationismainlyinfluencedby temperatureandadditionofMgO

Molar balance for nitrogen highlighted a strong ammonia volatilizationattheendoftheruns.Foreachrun,totalammonia volatilizationwascalculatedusing(12).

[NH3− N]volatilized= [NH+4− N]initialinliquid− [NH +

4 − N]finalinliquid

− [NH+

4− N]finalinsolid (12)

[Mg-MgO]andtemperature(p<0.0001)weretheleading pro-cessparameterswhichinfluencedammoniavolatilization(Table6). Theinteractionbetweentemperatureand[Mg-MgO]was sig-nificant. Whatever concentration [Mg-MgO] had at 5◦C, NH

3

volatilizationwasbelow10%attheendoftheexperiments,while theeffectof[Mg-MgO]washighathightemperature(Fig.6).

In literature,ammonia volatilization is function ofpH, tem-peratureandammoniumconcentration[37].Inthisexperimental

Fig.6.[Mg-MgO]andtemperaturepositivelyinfluencetotalNH3volatilization,the

(8)

Fig. 7. Calcium apparent dissolution follows a first order kinetic, tempera-ture=20◦C,N:P=3,initialMg:Ca=1.0–1.2,[Mg-MgO]=1000–1600mgL−1.

design,initialN:Pmolarratioandstirringratehadnosignificant influenceonammoniavolatilization.

3.1.4. Phosphateprecipitationascalciumphosphatesismainly influencedbystirringrateandN:Pratio

Thecalciumprecipitationisverycomplextoexplainasitcould precipitateinseveralformsdependingonthekineticofstruvite precipitation,pH,supersaturation,temperature...

Furthersolidanalysisneedstobeperformedhourlyto deter-mineexactlythecompositionofthesolid.Intheseexperiments, alargepartofthecalciumwasfirstremovedwithin30min,and then,weobservedanapparentdissolutionoftheremovedcalcium whichfollowedafirstorderkinetic(Fig.7and(13)).

Over48runs,43weredescribedby(13)(meanR2was0.99and standarddeviationwas5.36mgL−1).5runswereexcludedfrom

thestatisticalanalysisbecausetherewasnoapparentdissolution ofthecalcium.

[Ca2+]

t= ([Ca2+]e− [Ca2+]0)× (1− e−Kt)+ [Ca2+]0 (13)

[Ca2+]

tand[Ca2+]earetheconcentrationofthecalciumtotally

dissolvedattandatequilibriumrespectively.[Ca2+]

0,the

concen-trationwhentimeequals0h,hasnochemicalreality.Kisthekinetic constantofdissolution.

TheN:Pmolarratioandstirringrateweretheleading parame-tersthatinfluenced[Ca2+]

e(Table7andFig.8).

Whenstirringratewasbelow10rpm,precipitationsor disso-lutionsofcalciumphosphateandstruviteandMgOwerediffusion controlled[7,8].Therefore,lowstirringratessloweddown,oreven preventedthedissolutionofcalciumphosphate.

Table7A

ANOVAfortheconcentrationofcalciumatequilibrium.

Processparameters Prob.>Ffor[Ca2+] e

AdjustedR2 70.67%

A:Stirringrate 0.0003(+)

B:Temperature 0.0152(–)

C:N:Pmolarratio 0.0000(+)

D:initialMg2+:Ca2+molarratio 0.4254(–)

E:Mg2+asMgO 0.0093(–) AA 0.0006(–) BB 0.0000(+) BD 0.0075(–) CD 0.0171(+) DD 0.0115(–)

(+)indicatesapositive influenceand(–)a negativeinfluenceonthestudied responses.

Table7B

ANOVAforparticles>100mm.

Processparameters Prob.>Fforparticles> 100mm(g/L)

Prob.>Ffor%particles >100mm AdjustedR2 69.43% 72.36% A:Stirringrate 0.3431(+) B:Temperature 0.0062(–) 0.0460(–) C:N:Pmolarratio 0.0029(+) 0.0361(+) E:Mg2+asMgO 0.0000(–) 0.0000(–) AE 0.0474(–) EE 0.0000(+) 0.0000(+)

(+)indicatesa positiveinfluenceand (–)anegativeinfluence onthestudied responses.

TheinfluenceofN:Pmolarratioon[Ca2+]

econfirmsthe

previ-ousresults:highN:Pmolarratioimprovedstruviteprecipitation insteadofcalciumphosphateprecipitation.

3.2. Influenceofprocessparametersondriedproducts

Thethreesolid fractions(<25mm,25–100mmand >100mm) wereweighedandanalyzedbyionchromatographyafteracid dis-solutionandbyRamanspectroscopy.

Inordertoassessmassbalanceindriedsolids,phosphatesare consideredasstruviteandACP(part3.3),calciumwasthoughtto precipitateasACPandCaCO3andmagnesiumwasconsideredas

MgOand struvite(14). Asrunswereallcarriedout inanopen beaker,carbonatewaseitherincorporatedtothecrystalsor pre-cipitated as calcite [38]. The linear formula used for ACP was Ca2P2O7·H2O.

mdriedsolid= mMgO+ mStruvite+ mACP+ mCaCO

3+ mNa++ mK+ + mCl−+ mNO

2 + mNO −

3 + mSO2−4 (14) For80%oftheruns,theabsoluteerroronmassbalancewas belowthesumof experimentalerrors(20%).These results con-firmedthehypothesisthatcalciumcouldprecipitateasACPandas calciumcarbonate.Theabsoluteerroronmassbalancewas signif-icantinathirdofthefineparticlesandinafifthoftheparticles >100mmobtainedat35◦C.TheRamananalyses(3.3.3)showthat

magnesiumprecipitatedasMg(OH)2(brucite)insteadofremaining

asMgOinsmallparticlesorinsteadofCaCO3inlargeparticles.

Fig.8. StirringrateandN:Pmolarratiopositivelyinfluence[Ca2+]

e,thegrid

(9)

Fig.9. [Mg-MgO]andtemperaturenegativelyinfluenceconcentrationofparticles biggerthan100mm,thegridrepresentsthesurfaceresponseofthestatisticalmodel.

3.2.1. Particles>100maremainlyinfluencedbytheadditionof MgO

Theinfluenceofprocessparameterswasstudiedonthefractions >100mm;thefiltrationofstruvitecrystalsarecarriedoutinfilter bagswitha100mmcut-offintheprocessplant.TheANOVAin

Table7indicatedthat[Mg-MgO](p<0.0001),temperatureandN:P molarratioweretheleadingprocessparameterswhichinfluenced theparticles>100mm(Fig.9).

Thestruvitegrowthisassumedtobesizedependent[11]:small nucleiproducesmallcrystalsultimately.Moreover,thehigherthe supersaturation(b)is,thesmallernucleiare[5,7,8].Therefore,as highconcentrationsofMg2+increasethesupersaturationof

stru-vite,thequantityoffinecrystalsisincreasedtoo.Theinfluenceof [Mg-MgO]onthestruvitegrowthisverycomplex.[Mg-MgO] influ-encedthepH(andsothephosphateremoval),theconcentrationof Mg2+(andsothesupersaturation),thecrystallizationasnucleation

ofstruvitecouldoccurontheparticlesofMgO...Furtherstudies areneededtobetterunderstandtheinfluenceofthisparameteron thekinetics.

When temperature rises, the supersaturation of struvite decreases,sothesizeofthecrystalsincreases[39].Thisassertion istrueonlywhenNH4+isnotalimitingfactorinstruvite

crystal-lization.Otherwise,astemperatureincreases,Henry’slawconstant (KH)anddissociationconstant(Ka)ofammoniaevolvetofavour

greaterNH3/NH4+ratioinliquidandammoniavolatilization[37].

Therefore,whendissolvedNH4+isalimitingfactor,NH3

volatiliza-tionwillslowdownstruvitecrystallizationandstruvitecrystals willdissolveinordertorestorethebalancetoNH4+inthesolution.

Onthisaccount,thetemperaturehasanegativeimpactoncrystal sizes.

TheeffectofstirringistomaintainMgOinsuspension (interac-tionAE).Withoutit,thepHwillnotrisetotheminimalvalueforP precipitation.However,highmixingrateshaveanegativeimpact onstruvitecrystalsizes.Twomechanismscouldbeinvolved: stru-vitedissolutionorcrystalbreak.Asthefullamountofstruvitedid notchange,breakingwasmoreprobable.

3.2.2. Theproportionofstruviteversuscalciumphosphateis mainlyinfluencedbyparticlessizeandN:Pratio

TheexperimentaldesigninStatgraphicssoftwarewasmodified toincludetheinfluenceofparticlessize(<25mm,25-100mmor >100mm)ontheproportionofstruviteversuscalciumphosphate. R2ofstatisticalmodelinTable8wasbelow65%soitwassaidto

onlyindicatetrends.

TheANOVAinTable8indicatedthatparticlesize(p<0.0001), N:P molar ratio (p<0.001) and temperature (p<0.05) were the

Table8

ANOVAforproportionofstruviteversuscalciumphosphateindriedsolid(%). Processparameters Prob.>Fforproportionofstruviteversus

calciumphosphateindriedsolid(%)

AdjustedR2 62.78% B:N:P 0.0004(+) D:[Mg-MgO] 0.2961(–) E:Temperature 0.0177(–) F:Particlesize 0.0000(+) BE 0.0495(+) DD 0.0477(+) FF 0.0000(–)

(+)indicatesapositive influenceand(–) anegativeinfluenceonthestudied responses.

Fig.10.ParticlesizeandN:Pmolarratiopositivelyinfluencetheproportionof stru-viteversuscalciumphosphateindriedsolid,thegridrepresentsthesurfaceresponse ofthestatisticalmodel.

leadingprocess parameters which influencedthe proportionof struviteversuscalciumphosphateindriedsolids(%).

Struvitewasthemaincomponentofthefractions>25mmfor alllevelofvariables(Fig.10).

TheinfluenceofN:Pmolarratiosontheproportionofstruvite versuscalciumphosphateconfirmstheresultsofthepreviouspart: withahighN:Pmolarratio,struviteprecipitationwasfavoured overcalciumphosphateprecipitation.

3.3. Analysesofthesolidfractions

3.3.1. PotentialsolidphaseswithPHREEQC®

ThepotentialfinalprecipitatesaredefinedbyaSI>0.5;theSI valuesforthe48runsaregiveninTable10.ACP,calciteandstruvite couldprecipitateinalltheruns.Brucite,brushite,calciumand mag-nesiumphosphatecouldprecipitateinmostoftherunsasshown inTable11.Hydromagnesiteandartinitecouldprecipitate,but,in literature,therewereeithernotobservedinassociationwith stru-viteprecipitationortheycouldonlybiomineralizewiththehelpof bacteria[40,41].

3.3.2. CalciumphosphateprecipitatesasACP

Theformsofcalciumphosphateprecipitatedweredetermined bycalculatingtheCa:Pmolarratiointhethreesolidfractionsusing (15):

Ca:P= CaPrecipitated

(10)

Table10A

SIvaluesforthepotentialprecipitatesafter24hforthe48runs.

Minerals Formula Min Max Median NbrunSI>0.5

pH 6.84 10.47 9.20

Temperature(◦C) 5.00 35.00 20.00

Artinite MgCO3·Mg(OH)2·3H2O -7.12 3.18 1.64 36

Brucite Mg(OH)2 -0.84 6.51 5.26 42

Brushite(DCPD) CaHPO4·2H2O 0.12 1.00 0.58 38

Calciumphosphate CaHPO4 0.40 1.36 0.80 46

Calcite CaCO3 0.63 3.09 2.76 48

Calcitehydrate CaCO3·H2O −0.71 1.75 1.42 40

Hydromagnesite Mg5(CO3)4(OH)2·4H2O −12.62 8.83 4.71 40

K-Struvite MgKPO4·6H2O −2.09 1.06 0.47 26

Magnesiumphosphate Mg3(PO4)2 −1.28 5.29 4.03 41

Newberyite MgHPO4·3H2O −0.33 0.64 0.26 10

Struvite MgNH4PO4·6H2O 0.47 3.07 2.52 48

ACP Ca3(PO4)2·xH2O 1.33 6.74 5.42 48

Table10B

Ca:Pmolarratioinvariouscalciumphosphatesprecipitates[45].

Ca:Pmolarratio Name Formula pHstabilityrangein

aqueoussolutionsat25◦C

1.00 Brushite CaHPO4·2H2O 2.0–6.0

1.33 Octacalciumphosphate Ca8(HPO4)2(PO4)4·5H2O 5.5–7.0

1.00–2.20 ACP CaxHy(PO4)z·nH2O 5.0–12.0

1.50–1.67 Calcium-deficienthydroxyapatite Ca10−x(HPO4)x(PO4)6−x(OH)2−x 6.5–9.5

1.67 Hydroxyapatite Ca10(PO4)6(OH)2 9.5–12.0

WhereCaPrecipitated,PPrecipitatedandNPrecipitatedarethecalcium,

phosphorusandnitrogenprecipitatedinmol.g−1 of

solidrespec-tively.Nitrogenisassumedtoonlyprecipitateasstruvite. Inthefractions<25mm,Ca:Pmolarratiowasnear0.90which indicatesthepossiblepresenceofbrushiteorACP(Table10).The formofcalciumphosphatewasprobablyACPastheminimumpH valuesoftherunswerehigherthan6.0.Theliteratureconfirmsthe resultthatACPco-precipitateswithstruvite[14,19,21,42,43].The incorporationofMg2+ intothestructureofACPcouldexplained

whytheCa:Pmolarratiowasinferiorto1andwhyACPparticles weremainlyinthefraction<25mm[12,32].

Inthefractions>25mm,Ca:Pmolarratioswerebelow0.7. Cal-ciummightbeincorporatedinstruvitestructure.

Ca:Pmolarratiowashigherthan1.2forthemajorityofthe frac-tions>100mmandobtainedat35◦C.Ramanspectrumforthese

fractionsshowedthecharacteristicpeaksofCaCO3 (Fig.11,part

3.3.3).Moreover,theFig.2inpart3.1.1showsthatthepHvalues decreasedafter12hintherunsat35◦C.TheprecipitationofCaCO

3

couldexplainthisdecrease[44].

3.3.3. SolidanalysesbyRamanspectroscopy

Calcite, calcium phosphate, HAP, MgCO3, magnesium

phos-phate, newberyite, brucite, MgO, ammonium phosphate and struvite(analyticgrade)wereanalyzedbyRamanspectroscopyto havereferencespectra.

TheRamanspectrumforstruviteshowedtwomainpeaksfor phosphatevibrations,oneat950cm−1(totalsymmetricstretching

modeof–PO4)andoneat565cm−1(n4antisymmetricbendmodeof

–PO4)andtwopeakscorrespondingtoammoniumvibrations,one

at1440cm−1(n

4modeof–NH4)andoneat1700cm−1(n2modeof

–NH4)(spectrum(f)inFig.11).TheseRamanshiftsareinlinewith

literatureresults[38,46–48].

Presence,absenceandheightratioofthesefourpeaksindicate whethertheproductmainlycontainsstruviteornot.

TheRamanspectrumforbruciteshowed3characteristicpeaks: 3strongbandsat278,443(symmetricEgandA1gstretchingmode)

and1085cm−1andonebroadpeakat1044cm−1whichiscoherent

withliterature[49].

TheRamanspectraofthesolidfractionsconfirmedtheresults ofTable10.ACP,struviteandbrucitewerethemaincomponentof thefractions<25mm(spectrum(a)inFig.11andinTable11).

At5and20◦C,struvitewasthemaincomponentofthefractions

>25mm(spectra(d)and(e)inFig.11andinTable11).At20◦C,the

compositionofthefractions>100mmdependedonotherprocess parametersratherthanonparticlessizeortemperature(spectra (c)inFig.11).

At35◦Cthemaincompoundwasnotstruvitewhichwasinline

withsoliddissolutionanalysis.

Itisnoticeablethatthespectrumat5◦Cforfineparticles(a)

wassimilartothespectrum at35◦C forparticles >100mm(b).

Thosespectrapresented threecharacteristic peaksof carbonate precipitation: oneat 1080cm−1 (n

1 symmetric stretchmode of

–CO3),abroadoneat715cm−1 (n2modeof–CO3)[50]andone

at2906cm−1 (couldbeattributedtoCaCO

3).Comparedto

stru-vitespectrum, peaksat 435cm−1 (n

2 symmetric bendmode of

–PO4)shiftedby+15cm−1 whichcouldindicatethepresenceof

calciumphosphate[51].Thefeaturepeaksat278,and442cm−1

couldalsoindicatethepresence ofbrucite[49].Themainpeak at945cm−1 eitherwasveryloworhadvanished.Moreover,two

peaksappearedoneat1050cm−1(n

3antisymmetricstretchmode

Table11

Struviteisthemaincomponentofparticles>25mmobtainedat5and20◦C.

Temperature (◦C) 5◦C 20C 35C Particlessize (mm) <25 25–100 >100 <25 25–100 >100 <25 25–100 >100

Composition ACP=MAP*>Br*>Calcite MAP>CaP*=Br MAP>Br MAP>ACP>Br MAP>Br MAP>Br Br>ACP>MAP Br=MAP>ACP Br>ACP>MAP *MAP:struvite,Br:Brucite,CaP:Calciumphosphate,=:sameproportion,>:higherproportion.

(11)

Fig.11.Ramanshiftspectraofparticles<25mm(a)and>100mm(b)to(e)functionoftemperature:theruns3and12at5◦C(a)andtheruns42and47at35C(b)may

containACP,bruciteandcalcite,theruns28at20◦C(c)maycontainstruviteandbrucite,therun29at20C(d)andtherun2at5C(e)maycontainonlystruvite.(f)ispure

struvitefromCarloErba(99%)Optimizationofprocessparametersformaximizingstruviteprecipitation.

Table12

Processparameterslevelsforoptimaldesirability.

Processparameters Levels Levelsat15◦Cand60rpm

A:Stirringrate 80rpm 60rpm

B:Temperature 5◦C 15.0C

C:N:Pmolarratio 3 3

D:initialMg2+:Ca2+molarratio 0.6 1.7

E:Mg2+asMgO 520mgL−1 500mgL−1

of–PO4)andanotherat1370cm−1.Spectra(a)and(b)showedthe

presenceofthecharacteristicpeaksofphosphate,carbonateand brucite.Calciumcarbonate,bruciteandcalciumphosphatecould haveprecipitatedinthesefractions.TheprecipitationofCaCO3has

beenalreadyobservedinstruviteprecipitationprocesses[15,52]. Theoptimizationaimedtominimize[PO43−-P]eandammonia

volatilization,andmaximizetheproductionofbigparticlesof stru-vite.TheoptimumlevelsofparametersaregiveninTable12.

Alowtemperatureisdifficulttoapplyinanon-farmprocessing plant.Theprocessplantcouldstayinaclosedroombuttemperature variationscannotbeavoided.Therefore,theoptimalresponsefor thedesignexperimentwasestimatedat15◦C,theannualaverage

temperatureinBrittany(Table12andTable13).Stirringratewas loweredat60rpm,comparedtotheoptimal80rpm,duetofacility restrictions.

Themultipleresponseoptimizationproblemsweresolvedby combiningtheresponsesintoasingleindex:thedesirability func-tion.

[Mg-MgO]wastheleadingvariable thatinfluenced the pre-cipitationof struvite(Fig.12). When [Mg-MgO]waslow, small variationsof temperature,stirring rateorinitial Mg:Cadid not affectthedesirabilitywhichstayedinarangeof0.8–1.However, variationsoftheN:Pmolarratioaffectedthedesirabilitywhich decreasedto0.6foraN:Pmolarratioof1.

Inliterature,theoptimalpHforthestruviteprecipitationisin arange8to10[22,24,53].Weshowedthatafter24h,in“swine wastewaterconditions”, theoptimal pH wasnear7. TheFig.3

Table13

Predictedandexperimentalresponsesforoptimaldesirability.

Responses Weight Impact Optimalpredicted

response Predictedresponseat 15◦Cand60rpm Experimentalresponse at15◦Cand60rpm Desirability 3 0.90 0.86 [PO4-P]e 1 3 80mgL−1 60mgL−1 43 ± 12mgL−1 Particles>100mm 1 3 3.6gL−1 3.4gL−1 3.5± 0.3gL−1 %particles>100mm 1 3 67% 60% 87 ± 1% %PO4-Passtruvite 1 3 100% 96% 97 ± 3% %ammoniavolatilization 1 3 2% 8% 1 ± 1%

(12)

Fig.12.[Mg-MgO]istheleadingvariablethatinfluencesthestatisticalresponsesofthedesignofexperiments.

showsthatthispHvaluecorrespondedtoabout500mgL−1of

[Mg-MgO].ThisiscoherentwithastudyofBattistonietal.whichsaid thatthepresenceofcalciumspeedsupthestruviteprecipitationat lowerpHvalues[54].

3.4. Precipitationwiththeoptimizedlevelsforvariables

The precipitation of struvite was performed again with the best conditions. The experiments were triplicated. One more

experimentwasperformedtomonitortheammoniavolatilization. The N:P molar ratio was 3.2 and the Mg:Ca molar ratio was 2.3. MaximalpHwasthen 6.7.The experimentalresponses are presentedinTable13.

After24h,allparticleswereinthefractions>25mm.TheMgO entirelydissolvedandtheparticlesofcalciumphosphatewerein thefractions>25mm.

Theproportionofstruviteversuscalciumphosphateindried solidswas76%± 9%inthefraction25-100mmandincreasedto

Fig.13.Ramanshiftspectraofparticlessievedafter24hat15◦Csuggestthatstruviteisthemaincomponent:ofparticlesbiggerthan100mm(a)andofparticlesbetween

(13)

86%± 2%inthefraction>100mm.Theseresultsconfirmtheresults ofthestatisticalmodel.

Theoptimallevelsforvariablesmighthavenotonlyimproved thesizeofstruvitecrystalsbutalsoincreasedthesizeofcalcium phosphateparticles.Inpart3.3,aCa:Pmolarrationear1indicated thepresenceofACP.Inthe48runs,thisratiowasobservedinthe fractions<25mm.Intheoptimizationrun,thisratiowasobserved inthefraction25-100mm.ACPcouldhaveprecipitatedinthissolid fraction.HoweverRamananalysis(spectrum(b)inFig.13)didnot showthecharacteristicpeaksofcalciumphosphate:theproportion ofstruviteindriedsolidswasmuchlargerthancalciumphosphate. Intheoptimizationrun,Ca:Pmolarratiowas0.57± 0.04inthe fraction>100mmwhichwassimilartoresultsobtainedin experi-mentaldesign.

4. Conclusions

The optimizationof theprecipitationof struvitein a stirred beakermadeitpossibletorecyclemorethan90%ofphosphorusin largecrystalsofstruvitedespiteofthepresenceoflargeamounts ofcalciumandusingacheapreactantsuchasMgO.

Theoptimalconditionsfortheremovalofphosphorusaslarge crystalsofstruviteinsyntheticswinewastewaterwere:lowMg:Ca molarratio(2.25:1),theleadingparameter,highN:Pmolarratio (3:1),moderatestirring rate(between45 and 90rpm)and low temperature(below20◦C).

High N:P molarratio improved theprecipitation of struvite insteadofcalciumphosphate.LowconcentrationsofMgOinduced alowsupersaturationwhichimprovedthesizeofthestruvite crys-tals;italsominimizedammoniavolatilization.

Ramananalysisandsoliddissolutioninacidmadethe identifi-cationofthecompositionoftheproductpossible.Theseanalyses revealedthat:ACPco-precipitatedwithstruvite,carbonateswere incorporatedintosolidfractionsandbruciteprecipitatedinalmost alloftheruns.TheRamananalyseswerecoherentwiththeSIvalues calculatedbyPhreeqc.

Therunslastedfor24hinordertoreachequilibrium.Thesize andthecompositionoftheprecipitatedsolidscouldevolvebefore reachingtheequilibrium.Furtheranalysisisneededtounderstand evolutionofthesolidphasesover24-hours.

Acknowledgements

Thisworkissupportedbythe“ConseilRégionaldelaBretagne”, bytheFrenchResearchNationalAgency(ANR)andbyspecific uni-versityresearch(MSMTNo21/2012).

References

[1]A.Smit,P.Bindraban,J.Schröder,J.Conjin,H.VanderMeer,Phosphorusin Agriculture:GlobalResources,TrendsandDevelopments,Reporttothe Steer-ingCommitteeTechnologyAssessmentoftheMinistryofAgriculture,The Neetherlands,Wageningen,2009.

[2]R.GonzalezPonce,Evaluationofstruviteasafertilizer:acomparisonwith traditionalPsources,Agrochimica51(2007)301–308.

[3]M.-L.Daumer,F.Béline,M.Spérandio,C.Morel,Relevanceofaperchloricacid extractionschemetodeterminemineralandorganicphosphorusinswine slurry,Bioresour.Technol.99(2008)1319–1324.

[4]M.-L.Daumer,S.Picard,P.Saint-Cast,P.Dabert,Technicalandeconomical assessmentofformicacidtorecyclephosphorusfrompigslurrybyacombined acidification-precipitationprocess,J.Hazard.Mater.180(2010)361–365. [5]K.S.LeCorre,UnderstandingStruviteCrystallisationandRecovery,Cranfiel

Uni-versity,SchoolofAppliedScience,DepartmentofSustainableSystem,Center forWaterScience,2006.

[6]N.Hutnik,K.Piotrowski,B.Wierzbowska,A.Matynia,Continuousreaction crys-tallizationofstruvitefromphosphate(V)solutionscontainingcalciumions, Cryst.Res.Technol.46(2011)443–449.

[7]J.Mullin,Crystallization,fourthed.,ButterworthHeinemann,2001. [8]A.Mersmann,CrystallizationTechnologyHandbook,second,CRC,2001. [9] M.I.H.Bhuiyan,D.S.Mavinic,R.D.Beckie,Nucleationandgrowthkineticsof

struviteinafluidizedbedreactor,J.Cryst.Growth310(2008)1187–1194.

[10]K.Byrappa,InternationalSchoolonCrystalGrowthofTechnologically Impor-tantElectronicMaterials,AlliedPublishers,2003.

[11]J. Koralewska, K. Piotrowski, B. Wierzbowska, A. Matynia, Kinetics of reaction-crystallizationofstruviteinthecontinuousdrafttubemagmatype crystallizers—influenceofdifferentinternalhydrodynamics,Chin.J.Chem.Eng. 17(2009)330–339.

[12]F.Abbona,M.Franchini-Angela,Crystallizationofcalciumandmagnesium phosphatesfromsolutionsoflowconcentration,J.Cryst.Growth104(1990) 661–671.

[13] F.Abbona, H.E.L.Madsen, R. Boistelle, Theinitialphases ofcalciumand magnesiumphosphatesprecipitatedfromsolutionsofhightomedium con-centrations,J.Cryst.Growth74(1986)581–590.

[14]F.Abbona,H.E.LundagerMadsen,R.Boistelle,Thefinalphasesofcalciumand magnesiumphosphatesprecipitatedfromsolutionsofhightomedium con-centration,J.Cryst.Growth89(1988)592–602.

[15] I.Kabdas¸li,O.Tünay,M.S¸.C¸etin,T.Ölmez,Assessmentofmagnesium ammo-niumphosphateprecipitationforthetreatmentofleathertanningindustry wastewaters,WaterSci.Technol.46(2002).

[16]O.Tünay,I.Kabdasli,D.Orhon,S.Kolc¸ak,Ammoniaremovalbymagnesium ammonium phosphateprecipitationin industrialwastewaters,Water Sci. Technol.36(1997)225–228.

[17]M.L.Daumer,F.Béline,S.A.Parsons,Chemicalrecyclingofphosphorusfrom piggerywastewater,in:IntlWaterAssn,Vancouver,2009,pp.339–350. [18]I.Lopez-Valero,C.Gomez-Lorente,R.Boistelle,Effectsofsodiumand

ammo-niumionsonoccurrence,evolutionandcrystallinityofcalciumphosphates,J. Cryst.Growth121(1992)297–304.

[19]D.Crutchik,J.M.Garrido,Struvitecrystallizationversusamorphousmagnesium andcalciumphosphateprecipitationduringthetreatmentofasalineindustrial wastewater,WaterSci.Technol.64(2011)2460.

[20]K.S.LeCorre,E.Valsami-Jones,P.Hobbs,S.A.Parsons,Kineticsofstruvite pre-cipitation:effectofthemagnesiumdoseoninductiontimesandprecipitation rates,Environ.Technol.28(2007)1317–1324.

[21]Z.-L.Ye,S.-H.Chen,S.-M.Wang,L.-F.Lin,Y.-J.Yan,Z.-J.Zhang,etal., Phospho-rusrecoveryfromsyntheticswinewastewaterbychemicalprecipitationusing responsesurfacemethodology,J.Hazard.Mater.176(2010)1083–1088. [22] J.D.Doyle,R.Philp,J.Churchley,S.A.Parsons,Analysisofstruviteprecipitation

inrealandsyntheticliquors,ProcessSaf.Environ.Prot.78(2000)480–488. [23]A.Uysal,Y.D.Yilmazel,G.N.Demirer,Thedeterminationoffertilizerquality

oftheformedstruvitefromeffluentofasewagesludgeanaerobicdigester,J. Hazard.Mater.181(2010)248–254.

[24]D.Kim,H.-D.Ryu,M.-S.Kim,J.Kim,S.-I.Lee,Enhancingstruvite precipita-tionpotentialforammonianitrogenremovalinmunicipallandfillleachate,J. Hazard.Mater.146(2007)81–85.

[25] E.V.Münch,K.Barr,Controlledstruvitecrystallisationforremovingphosphorus fromanaerobicdigestersidestreams,WaterRes.35(2001)151–159. [26]J.Goupy,Lesplansd’expériencepoursurfacederéponse,Dunod,1999. [27]M.Hanhoun,L.Montastruc,C.Azzaro-Pantel,B.Biscans,M.Frèche,L.Pibouleau,

Temperatureimpactassessmentonstruvitesolubilityproduct:a thermody-namicmodelingapproach,Chem.Eng.J.167(2011)50–58.

[28]A.Ma ˜nas,M.Pocquet,B.Biscans,M.Sperandio,Parametersinfluencingcalcium phosphateprecipitationingranularsludgesequencingbatchreactor,Chem. Eng.Sci.77(2012)165–175.

[29] M.Altmaier,V.Metz,V.Neck,R.Müller,T.Fanghänel,Solid-liquidequilibriaof Mg(OH)2(cr)andMg2(OH)3Cl4H2O(cr)inthesystemMg-Na-H-OH-Cl-H2Oat 25◦C,Geochim.Cosmochim.Acta67(2003)3595–3601.

[30]J.C.Deelman,Low-TemperatureFormationofDolomiteandMagnesite, Com-pactDiscPublications,2003.

[31]X.L.Deng,Y.M.Luo,M.M.Xu,X.Y.Hu,Influenceofsolventpolarityonsynthesis ofbeta–tricalciumphosphate,KeyEng.Mater.330–332(2007)143–146. [32]F.Abbona,A.Baronnet,AXRDandTEMstudyonthetransformationof

amor-phouscalciumphosphateinthepresenceofmagnesium,J.Cryst.Growth165 (1996)98–105.

[33]P.Battistoni,A.DeAngelis,P.Pavan,M.Prisciandaro,F.Cecchi,Phosphorus removalfromarealanaerobicsupernatantbystruvitecrystallization,Water Res.35(2001)2167–2178.

[34]M.Quintana,E.Sanchez,M.F.Colmenarejo,J.Barrera,G.Garcia,R.Borja,Kinetics ofphosphorusremovalandstruviteformationbytheutilizationofby-product ofmagnesiumoxideproduction,Chem.Eng.J.111(2005)45–52.

[35]A.A.Szögi,M.B.Vanotti,Removalofphosphorusfromlivestockeffluents,J. Environ.Qual.38(2009)576.

[36]M.B.Vanotti,A.A.Szogi,P.Hunt,Extractionofsolublephosphorusfromswine wastewater,Trans.ASAE46(2003)1665–1674.

[37]F.Montes,C.A.Rotz,H.Chaoui,Processmodelingofammoniavolatilization fromammoniumsolutionandmanuresurfaces:areviewwithrecommended models,Trans.ASABE52(2009)1707–1719.

[38]M.Kazanci,P.Fratzl,K.Klaushofer,E.P.Paschalis,Complementary informa-tiononinvitroconversionofamorphous(precursor)calciumphosphateto hydroxyapatitefromramanmicrospectroscopyandwide-angleX-ray scatter-ing,Calcif.TissueInt.79(2006)354–359.

[39]M. Ronteltap, M. Maurer, R. Hausherr, W. Gujer, Struvite precipitation from urine –influencing factorson particle size,Water Res. 44 (2010) 2038–2046.

[40]M.Rivadeneyra,R.Delgado,J.Párraga,A.Ramos-Cormenzana,G.Delgado, Precipitationofmineralsby22speciesofmoderatelyhalophilicbacteriain artificialmarinesaltsmedia:Influenceofsaltconcentration,FoliaMicrobiol. 51(2006)445–453.

(14)

[41]M.Sánchez-Román,C.S.Romanek,D.C.Fernández-Remolar,A.Sánchez-Navas, J.A.McKenzie,R.A.Pibernat,etal.,AerobicbiomineralizationofMg-rich carbo-nates:Implicationsfornaturalenvironments,Chem.Geol.281(2011)143–150. [42]L.Pastor,D.Mangin,J.Ferrer,A.Seco,Struviteformationfromthesupernatants ofananaerobicdigestionpilotplant,Bioresour.Technol.101(2010)118–125. [43]N.Martí,L.Pastor,A.Bouzas,J.Ferrer,A.Seco,Phosphorusrecoverybystruvite crystallizationinWWTPs:Influenceofthesludgetreatmentlineoperation, WaterRes.44(2010)2371–2379.

[44]J.A.Wojtowicz,Calciumcarbonateprecipitationpotential,J.SwimmingPool SpaInd.2(2001)23–29.

[45]S.V. Dorozhkin, Amorphous calcium(ortho)phosphates, Acta Biomater. 6 (2010)4457–4475.

[46]R.L.Frost,M.L.Weier,W.N.Martens,D.A.Henry,S.J.Mills,Ramanspectroscopy ofnewberyite,hannayiteandstruvite,Spectrochim.ActaA62(2005)181–188. [47]V.Stefov,B. ˇSoptrajanov,I.Kuzmanovski,H.D.Lutz,B.Engelen,Infraredand Ramanspectraofmagnesiumammoniumphosphatehexahydrate(struvite) anditsisomorphousanalogues.III.Spectraofprotiatedandpartially deuter-atedmagnesiumammoniumphosphatehexahydrate,J.Mol.Struct.752(2005) 60–67.

[48] V.Stefov,B. ˇSoptrajanov,F.Spirovski,I.Kuzmanovski,H.Lutz,B.Engelen, InfraredandRamanspectraofmagnesiumammoniumphosphatehexahydrate (struvite)anditsisomorphousanalogues.I.Spectraofprotiatedandpartially

deuteratedmagnesiumpotassiumphosphatehexahydrate,J.Mol.Struct.689 (2004)1–10.

[49]F.Pascale,S.Tosoni,C.Zicovich-Wilson,P.Ugliengo,R.Orlando,R.Dovesi, Vibrationalspectrum of brucite, Mg(OH)2: a periodic ab initio quantum

mechanicalcalculationincludingOHanharmonicity,Chem.Phys.Lett.396 (2004)308–315.

[50]H.G.M.Edwards,S.E.J.Villar,J.Jehlicka,T.Munshi,FT–Ramanspectroscopic studyofcalcium-richandmagnesium-richcarbonateminerals,Spectrochim. Acta,PartA61(2005)2273–2280.

[51] C.Paluszkiewicz,M.Gałka,W.Kwiatek,A.Parczewski,S.Walas,Renalstone studiesusingvibrationalspectroscopyandtraceelementanalysis, Biospec-troscopy3(1997)403–407.

[52]P.Battistoni,R.Boccadoro,F.Fatone,P.Pavan,Auto-nucleationandcrystal growthofstruviteinademonstrativefluidizedbedreactor(FBR),Environ. Technol.26(2005)975–982.

[53] H.Huang,C.Xu,W.Zhang,Removalofnutrientsfrompiggerywastewaterusing struviteprecipitationandpyrogenationtechnology,Bioresour.Technol.102 (2011)2523–2528.

[54]P.Battistoni,P.Pavan,M.Prisciandaro,F.Cecchi,Struvitecrystallization:a fea-sibleandreliablewaytofixphosphorusinanaerobicsupernatants,WaterRes. 34(2000)3033–3041.

Figure

Fig. 2. The pH function of logarithm base 2 of the time evolved differently function of temperature – 5 ◦ C run 10, 20 ◦ C run 30, 35 ◦ C run 42.
Fig. 5. [Mg-MgO] and stirring rate negatively influence the final dissolved phos- phos-phate [PO 4 3− -P] e , the grid represents the surface response of the statistical model.
Fig. 8. Stirring rate and N:P molar ratio positively influence [Ca 2+ ] e , the grid repre- repre-sents the surface response of the statistical model.
Fig. 10. Particle size and N:P molar ratio positively influence the proportion of stru- stru-vite versus calcium phosphate in dried solid, the grid represents the surface response of the statistical model.
+3

Références

Documents relatifs

Similar to soluble P concentration in the rhizosphere, total P uptake by plants was primarily controlled by N forms whereas P treatments had no significant effect (Table 4)..

The Q-band EPR spectrum of sample Cu_150 heated to 1000ºC exhibited two signals, characteristic for Cu(II) in sites with rhombic and axial distortions, respectively..

The experimental system included four types of microcosm upflow CWs depending on the type of bed media and the presence of vegetation: vegetated columns with

The CWs filled with organic media (UF-PP) reduced the Ni concentration with higher efficiency than observed for this metal in the UF-GP columns, probably due to

The absence of antibodies against ASF in warthog sera confirms the lack of circulation of ASFV among warthog populations in Senegal, even in areas where the tick is present..

Oxidation of flumequine in aqueous solution by UV-activated peroxymonosulfate: Kinetics, water matrix effects, degradation products and reaction pathways.. Enhancement

les pathologies cardiovascu- laires les plus fréquentes (et les plus fréquemment rencon- trées en salle d’urgence) sont le syndrome coronarien aigu et la fibrillation auriculaire