Syneresis of rennet-induced milk gels as influenced by cheesemaking parameters

CENTRALE LANDBOUWCATALOGUS

0000 0298 5626 Promotor: dr. ir. P. Walstra, hoogleraar in de Zuivelkunde H. J. C. M. van den Bijgaart

Syneresis of rennet-induced milk gels as influenced by cheesemaking parameters

Proefschrift ter verkrijging van de graad van doctor in de landbouwwetenschappen, op gezag van de rector magnificus, dr. H. C. van de Plas, in het openbaar te verdedigen op woensdag 21 september 1988 des namiddags te vier uur in de aula van de Landbouwuniversiteit te Wageningen BIBLIOTHEEK LANDBOUWUNIVERSITEIT: WAGENINGEN

ISW i'i^.L/i iy ABSTRACT

Van den Bijgaart,H.J.CM . (1988). "Syneresis of rennet-induced milk gels as influenced by cheesemaking parameters". Ph.D. Thesis, Laboratory of Dairying and Food Physics, Department of Food Science, Wageningen Agricultural University (141 pp, Englishan dDutc h summaries). key-words: syneresis, rennet-induced milk gels, endogenous syneresis pressure, permeability, mechanical pressure, cheese- makingparameters .

The syneresis behaviour of rennet-induced skimmilk gels was studied at closely controlled conditions. The origin of the endogenous syneresis pressure and changes therein and in the permeability of the gel during syneresis are discussed. The latter appears to be of predominant importance for the one- dimensional shrinkage rate under quiescent conditions. The influence of cheesemaking parameters such as milk pretreatment, rennet concentration, acidification procedure and pH, addition of CaCl2, additiono f NaCl, and temperature on the permeability and the initial endogenous syneresis pressure were separately determined. The syneresis behaviour is clearly related to the size and the reactivity of the aggregated paracasein micelles and to the changes in the microstructure and the rheological propertiesdurin ggelation . A method for the accurate determination of the important effect of mechanical pressure is outlined, indicating a square root proportionality of the shrinkage with time after load appli­ cation forth eone-dimensiona l case.Base d onexistin g theories for the compression of porous media, the results obtained were related to some structural characteristics of the protein matrix. Available informationwa suse d toconside r someaspect s relevant to practical cheesemaking. Attention was paid to hydrodynamic effects during cutting and stirring, shrinkage of spherical particles,syneresi swit hchangin gp Ha swel la st osyneresi so f gelsfro mpreconcentrate d milk. (JNo$Zo\, \13\

STELLINGEN

De synerese van met leb gevormde melkgelen wordt mede bepaald door de reologische eigenschappen van het caseïne- netwerk. Dit proefschrift.

2. Dedoorstroombaarhei d vanme tle bgevormd emelkgele n tijdens synererenword t onderschat indien daartoe de resultaten van experimenten met gelen uit voorgeconcentreerde melk worden gehanteerd. Dit proefschrift.

Gelet op de experimenteleomstandighede n bij deproeve n van Marshall moet de door hem gevonden snellere synerese indien het gel werd gesneden op een later tijdstip na stremsel- toevoeging, worden toegeschreven aan betere mogelijkheden voor uitstroming van vloeistof dan aan een hogere endogene syneresedruk. Marshall. R.J. (1982). J. Dairy Res. 49. 329-336.

4. Bij onderzoek naar de effecten van een temperatuur­ verandering op de fysisch-chemische eigenschappen van caseïnemicellen dient rekening gehouden te worden met een vertraagdeaanpassin gva nd e zwellingsstoestand.

5. Melk bevat nog niet-geïdentificeerde componenten die een stimulerende invloed hebben op de vlokaktiviteit van verdund kalfsstremsel en die bovendien de inaktivering bij invriezentegengaan .

6. De geringe specificiteit van de door Castaneda et al. beschreven methode voor het aantonen van droge stof uit kaaswei in melkpoeder, door middel van bepaling van een vriespunt,beperk td ebruikbaarhei dervan . Castaneda, R.. Fernandez, G.. Caló. M. & Pasqualini, A. (1987). Neth. Milk Dairy J. 41. 69-79.

7. De intensiteit van de zure-smaakgewaarwording van oplos­ singen van verschillende voedingszuren correleert beter met demolaritei tda nme td egewichtsconcentratie . 8. In beschouwingen over de rentabiliteit van het eventuele gebruik van boviene somatotropins in de melkveehouderij krijgend emogelijk econsequentie svoo rd eafze tva nmel k en zuivelprodukten teweini gaandacht .

De uitvaardiging van normen voor "warmtebehandelde" melk, zoals vastgelegd in EG richtlijn 85/397, dient geen enkel wezenlijk doel; een verantwoorde toepassing is bovendien nietgoe dmogelij kzolan gd ebijbehorend e onderzoeksmethoden nietzij nvastgesteld .

10.He t gebruik van een beeldscherm betekent lang niet altijd eenverruimin gva nhe tblikveld .

H.J.C.M,va nde nBijgaar t Syneresiso frennet-induce d milkgel s asinfluence db ycheesemakin gparameters . Wageningen,2 1septembe r1988 . WOORDVOORA F

Bij het beginva n dit proefschrift hoortee nwoor d vandan k aan al diegenen, die in belangrijke mate hebben bijgedragen aan de totstandkomingervan . Jeanne, ondanks dewa t lijdzamewe g diepartner s van promovendi noueenmaa lbeschore n lijktt ezijn ,he bj edoo rj evoortdurend e aanmoediging en begrip wezenlijk aan de voltooiing van dit proefschriftbijgedragen . Het Kaascontrolestation 'Friesland' te Leeuwarden heeft via de Stichting J. Mesdagfonds de uitvoering van het beschreven onderzoek financieelmogelij kgemaakt . Beste Pieter,hooggeleerd e promotor, ikwi l jebedanke nvoo r de gelegenheid die je me hebt geboden om dit onderzoek te ver­ richten. Jouw kennis en aanpak hebben de afgelopen periode tot eenzee rleerzam evoo rm egemaakt . Mijndan kgaa tverde rui tnaa rTo mGeurt svoo rzij nvoortdurend e beschikbaarheid als praatpaal,voo r het aandragen vand e kennis vanuit het kaasmaken en voor het wijzen op de kleine maar vaak zeerwezenlijk e zakenbi jhe tuitvoere nva nexperimenten . Ton van Vliet en Nel Zoon hebben via hun uitleg en discussie voormi jee nwe ggebaan d inhe twou dva nreologisch easpecten . Annemarie Spaargaren heeft een bijdrage geleverd in het kader van haar doktoraalstudie. Aliza de Groot wil ik bedanken voor hetverrichte nva nenkel eaanvullend emetingen . De uitwerking van een aantal computermatige berekeningen is geschied met aanzienlijke hulp van dr. R. Koopmans (vakgroep Cultuurtechniek), Prof. J. Schenk (emiritus-hoogleraar van de vakgroep Technische Natuurkunde) en Harrie Kamphuis (KSEPL te Rijswijk). Demedewerker sva nd esecti eZuive le nLevensmiddelennatuurkund e hebben op uiteenlopende wijze bijgedragen, van de directe technische ondersteuning tot en met de benodigde afleiding via allegedachtenwisselinge n overniet-zuivelonderwerpen . De heer C. Rijpma van de tekenkamer in het Biotechnion en de medewerkers van de fotolokatie De Dreyen hebben op bijzonder vlottee nnauwgezett ewijz ehe tteken -e nfotower kverzorgd . Voor de afwerking van dit proefschrift mocht ik een bijdrage ontvangenui the tLEB-fonds . Mijn huidige werkgever, de Coöperatieve Vereniging voor Melk- onderzoek "Zuid-Nederland" teVeldhoven ,be ni kerkentelij k voor degelegenhei d diemi jwer dgebode no md everslagleggin gva ndi t onderzoek aft eronden . CONTENTS

Page 1INTRODUCTIO N 1.1Genera l 1 1.2Rennet-induce dcoagulatio no fmil k 1 1.3Syneresi so frennet-induce dmil kgel s 5 1.4Outlin eo fthi sstud y 6

2MATERIAL SAN DMETHOD S 2.1Material s 7 2.1.1Reconstitute dskimmil k 7 2.1.2Concentrate dmil k 8 2.1.3Ra wmil k 8 2.1.4Renne t 9 2.1.5Whe y 9 2.1.6Aprotini n 9 2.1.7Chemical s 10 2.2Method s 10 2.2.1Clottin gtim e 10 2.2.2Shrinkag erat eo fcur dslab s 10 2.2.2.1Microscop emetho dwithou tmechanica l pressure 10 2.2.2.2Microscop emetho dwit hmechanica l pressure 12 2.2.3Long-ter mresidua lheigh to fcur dslab s 12 2.2.3.1Residua lheigh twithou tmechanica l pressure 12 2.2.3.2Residua lheigh twit hmechanica lpressur e1 3 2.2.4Permeabilit ymeasurement s 13 2.2.4.1Genera l 13 2.2.4.2Permeabilit yo fnon-synerese dgel s 14 2.2.4.3Permeabilit yo fsynerese dgel s 15 2.2.5Viscosit yo fth ewhe y 15 2.2.6Ge lelectrophoresi s 15 2.2.7p H 16 2.2.8Calciu mio nactivit y 16 3 ONE-DIMENSIONAL SYNERESIS:SOM ECONSIDERATION S 3.1 Introduction 17 3.2 Syneresispressur e 19 3.2.1 Syneresispressur e ina non-synerese d gel 19 3.2.1.1Endogenou ssyneresi spressur e 19 3.2.1.2Influenc eo ftim eafte rrenne tadditio n 22 3.2.1.3Influenc eo fmil kpreconcentratio n 24 3.2.1.4Gravitation-induce d syneresispressur e 25 3.2.2 Syneresispressur e ina syneresin gge l 26 3.2.2.1Endogenou ssyneresi spressur e 26 3.2.2.2Gravitation-induce d syneresispressur e 28 3.3 Permeability 29 3.3.1 Permeabilityo fgel sfro mpreconcentrate d milk 29 3.3.2 Permeability ofsynerese d gels 30 3.4 Endpoin to fsyneresi s 35 3.5 Resultso fmode lcalculation s 37 3.5.1 Theeffec to fvariou stria l functions forth e endogenoussyneresi spressur e 37 3.5.2 Theeffec to fdifferen t relations forth echang e of thepermeabilit ycoefficien t 39 3.6 Calculationo fth einitia lendogenou s syneresis pressure fromexperimenta l results 39 3.7 Conclusions 44

4 EFFECTO FVARIATIO N INCONDITION SDURIN GRENNETIN GAN D COAGULATIONO N SYNERESIS 4.1 Introduction 46 4.2 Resultsan ddiscussio n 48 4.2.1 Treatmento freconstitute d milk 48 4.2.2 Rennetconcentratio n 52 4.2.3 Acidification 55 4.2.3.1Effec to fth eacidificatio nprocedur e 56 4.2.3.2p H 59

4.2.4 CaCl2 65

4.2.4.1Tim eo fCaCl 2 addition 66

4.2.4.2Amoun to fCaCl 2 added 68 4.2.5 NaCl 68 4.2.6 Temperature 71 4.2.7 Fat 75 4.3 Conclusions 76

5 INFLUENCEO FMECHANICA L PRESSUREO NSYNERESI S 79 5.1 Introduction 79 5.2 Somegenera l remarkso nth eexperimenta l results 80 5.2.1 Courseo fth eshrinkag ewit htim eafte r loadapplicatio n 80 5.2.2 Contact surface 82 5.3 Resultswit hvariou sp Han dtemperatur e 84 5.4 Gelsfro mpreconcentrate dmil k 86 5.5 Applicationo fexistin g theories forcompressio n 89 5.5.1 Terzaghimode l 89 5.5.2 Amode l forelasti cdispers esystem s 90 5.5.3 Resultso fcalculation swit hexperimenta l data 95 5.5.4 Someremark sconcernin gvisco-elasti csystem s 98 5.6 Conclusions 102

6 SOMEASPECT SRELEVAN TT OPRACTICA LCHEESEMAKIN G 103 6.1 Introduction 103 6.2 Mechanicalpressur ean dhydrodynami ceffect sdurin g cutting and stirring 103 6.2.1 Cutting 104 6.2.2 Stirring 105 6.3 Intermittentpressur e 113 6.4 Modelcalculation swit hchangin gp H 114 6.4.1 Resultso fcalculation swithou tmechanica l pressure 115 6.4.2 Resultso fcalculation swit hmechanica l pressure 117 6.5 Milkpreconcentratio n andsyneresi s 118 6.6Preliminar y calculations for three-dimensional syneresis 121 6.7 Conclusions 125

LITERATUREREFERENCE S 126 LISTO FSYMBOL S 130 SUMMARY 133 SAMENVATTING 137 CURRICULUMVITA E 141 1 INTRODUCTION

1.1 General

Cheesei son eo fth emai ndair yproducts .Durin gcheesemakin g the larger part of thenutritiou s components of milk isconcen ­ trated. The composition and properties of the final cheese are determined by the characteristics of the raw material and the processing conditions. Thisha sresulte d into a widevariet y of types. Optimizationo fth echeesemakin g processimplie scontro l over theprocessin g steps:mil k treatment,ge l formation,dehydratio n and possible ripening. Although qualitatively the effects of variations (sucha smil kpretreatment ,renne tconcentration ,pH , temperature) on the processing steps are known, little effort has been spent in relating this to the changes of the building blocks and the structure at the microlevel. However, todays automation of the production requires a better understanding of theunderlyin gprinciples . In the past years our knowledge about the renneting of milk and the influence of cheesemaking parameters has considerably increased (recently reviewed by Dalgleish, 1987). Basic infor­ mation about the subsequent gelation and syneresis is rather scarce. Van Dijk (1982) performed experiments under closely controlled conditions and successfully related the syneresis behaviour to the microstructure of the gel. It was felt worth­ while to extend on his work and to study the influence of cheesemaking parameterso nsyneresi si nmor edetail . This study particularly concerns rennet-type gels, in which coagulation of the milk is induced by rennet enzymes. However, the outlined physical principles also apply to the other major groupo facid-typ egels .

1.2 Rennet-inducedcoagulatio no fmil k

The casein fractiono f milk,whic h constitutes about 80 %o f total milk protein, is the main component of interest for the changes in the physical state of milk during the cheesemaking 1 process. One can dividebetwee n fourmai ncharacteristi c groups of caseins: otsl-, CLB2-, ß-an d K-casein. In each group several genetic variants can be distinguished. Also differences in the molecular residues attached to amino acid side groups exist (Swaisgood, 1982;Walstr a& Jenness , 1984). Under normal conditions (uncooled milk, pH = 6.7) almost all casein appears in aggregates of colloidal sizewit h a diameter between 20 and 300 nm, the casein micelles. These aggregates also contain some 8 g of inorganic matter per 100 g casein, consisting of colloidal calcium phosphate (CCP) and counter ions, which includes some Mg, Na, K and citrate. The CCP is essential in keeping the integrity of the micelles (Schmidt, 1982). For our purpose thecasei nmicell e isthough t to consist of small subunits, the submicelles, which contain the casein molecules. These are mainly kept together by hydrophobic bonds and saltlinkage s (Walstra &Jenness , 1984). Caseinmicelle s are voluminous particles.The y hold a considerable amount ofwater , « 3 g/g casein (Walstra, 1979). Without going into any detail about the possible structure of the casein micelle (see for instance Schmidt, 1982), thehighl y idealized picture in Figure 1.1 can serve for this study. As outlined, several equilibria withseru mcomponent sexist ,indicatin g thedynami ccharacte ro f caseinmicelles .Th e locationo f submicelleswit h a high amount of K-casei n on the outside of the micelle is of main interest (Waugh, 1971;McGan n et al., 1980;Schmidt , 1982). A consider­ able amount of evidence has been presented for its role in the stabilizationo fth emicelle sunde rnatura lconditions .Th epro -

Co.2* * HP0«f" ^^ 1 — submicelles t casein molecules

Fig. 1.1. Model of a casein micelle, consisting of spherical submicelles, kept together by small patches of colloidal calcium phospate, and showing protruding chains of the caseino-macropeptide part of K-casein. Some equilibria with building blocks in the serum are schematically depicted (From: P. Walstra & T. van Vliet, 1986). trudinghydrophili c ic-caseinmacropeptid emoiet y (CMP),whic h is subject to Brownian motion, imparts steric and electrostatic repulsion between closely approaching micelles (Holt, 1975; Walstra, 1979;Walstr a et al., 1981;Home , 1984;Home , 1986; vanHooydon k &Walstra , 1987). Therennetin go fmil k isclearl yth eresul to ftw oprocesses , theattac ko n K-caseinb yproteolyti cenzyme s (mainlychymosin) , and the flocculation of the destabilized micelles. During the first reaction the CMP is split off by the specific attack of rennet enzymes on the Phej50 -Met j06 bond (Joliese t al., 1968). In milk this reaction is essentially first order, since the diffusivity of the micelles is negligible compared to that of theenzym emolecule s (vanHooydon k et al., 1984;va nHooydon k & Walstra, 1987). Before a particle becomes subject to floccu­ lation, a considerable amount of the surface charge and steric repulsion is lost (Green & Crutchfield, 1971; Pearce, 1976; Darling & Dickson, 1979;Walstr a & al., 1981;Dalgleish , 1984; Holt & Dalgleish, 1986). Only after about 70 - 90 % of the K- casein on amicell e hasbee n split, flocculationca noccu r when twoparticle smee teac hothe r (Dalgleish,1979 ;Chapli n& Green , 1980; van Hooydonk et al., 1986b). From that moment theaggre - gatibility increases fairly rapidly with the extent of proteo­ lysis (Darling & van Hooydonk, 1979; Dalgleish, 1980; van Hooydonke t al-, 1986b;Dalgleish , 1987). Theaggregatio nca nb e described with Smoluchovski kinetics, but in milk the reaction proceedsmuc hslowe rtha ni nth ediffusion-controlle d limit (van Hooydonk &Walstra , 1987). Thecaus e for theaggregatio n isno t fully understood. Besides van der Waals attraction, specific ion-pair formation and hydrophobic effects have been held responsible (Dalgleish, 1987). The influence of variation in experimental conditions on the separate reactions hasbee n sum­ marized byWalstr a &va nVlie t (1986)an dva nHooydon k &va nde n Berg (1987). Asa resul to fongoin g flocculationconglomerate san dthread ­ like structure are formed, as schematically outlined in Figure 1.2 (Mulder et al., 1966; Henstra & Schmidt, 1970; Green et al., 1978). When flocculation proceeds undisturbed (no stir­ ring), acontinuou s network is formed. Electron microscopy 1 oo O o o o

Fig. 1.2. Schematic representation of the changes from stable casein micelles to aggregated paracasein micelles during the renneting of milk. reveals an inhomogeneous structure,whic h ismad eu p of strands of micelles and thicker nodes, leaving openings up to 10 urni n diameter (Knoop & Peters, 1975; Green et al., 1978; Glaser et al., 1980). Generally, fatglobule sar etrappe d inth epore san d thusac ta sa non-reactiv efiller . After flocculation and gel formation the contact area or junction between neighbouring particles changes from 'touching' to 'fusion',a sdepicte d inFigur e1.3 . Initially,on eo ronl y a few bonds per junction exist. Afterwards, the particles become attached at more sites. This results in thickening and more or less smoothing of strands. Long-term rearrangement processes in thestrand s (Knoop& Peters ,1975 ;Gree ne tal. , 1978)ultimate ­ ly prevent the differentiation of bonds between micelles to those within micelles. During this process the contribution of the various types of bonds may change (Walstra & van Vliet, 1986). Macroscopically, during several hours after rennet addition an increase of the dynamic moduli is found in rheological measurements (vanDijk , 1982;Tokit ae t al., 1983;Bohlin ,1984 ; Dejmek, 1987). These moduli depend on the number, the strength andth erelaxatio nbehaviou ro fth ebond s (Zoone t al., 1988) .

Fig. 1.3. Schematic picture of the change in conformation of flocculated paracasein micelles during ageing of a rennet-induced milk gel (From: P. Walstra &T . van Vliet, 1986). 1.3 Syneresiso frennet-induce dmil kgel s

Rennet-induced milk gelsma y show syneresis, i.e. theexpul ­ sion of whey. This can result from an endogenous tendency to contract or be due to external forces exhibited on the gel. In bothcase spressur e isexerte d onth e liquid inth epores .Thi s pressure, and the resistance against flow through the matrix, that can be expressed as a permeability coefficient, determine theloca lchang ei nth evoluminosit ywit htime ,i.e . theshrink ­ age. This was modelled, based on the equation of Darcy (van Dijk, 1982;va nDij ke tal. , 1984).Walstr ae tal . (1985)argue d that the rearrangement in the network of paracasein particles must be the main driving force for syneresis to occur under quiescent conditions. This rearrangement can occur because paracasein micelles probably are reactive over their entire surface. Inth einitia lstage safte rge l formation rearrangement mayb epromote d bynon-aggregate d particles,whic hattac ht oth e existingnetwork .Ther ethe ybecom ea nextende d targetpoin t for danglingo rmovin g strands.I nlate rstages ,breakin go f someo f thestrand swil lb eneede d forextensiv erearrangemen t tooccur . Ifa ge li sconstrained , therearrangin g tendencyresult si nth e formation of dense regions and less dense regions elsewhere. This has been designated microsyneresis (van Dijk, 1982; van Dijk & Walstra, 1986). Furthermore, changing conditions during syneresis (e.g. pH and temperature)ma y affect the size of the building blocks (Walstra, 1979)an d contribute to the shrinking tendency of the matrix. Taking into account the nature of the resulting endogenous syneresis pressure, it is obvious that it mustdepen do nth estag eo fgelatio nan do nsevera l experimental conditions.Som eresult so nth einfluenc eo fpH ,temperatur ean d added CaCl2 on this parameter have already been given by van Dijk (1982)an d vanDij k &Walstr a (1986). According toexperi ­ mental conditions,value s between 1an d 3 Pawer e found for the initial endogenous syneresis pressure. The large effect of externalo rmechanica lpressur ei stherefor eno tsurprisin g (van Dijk, 1982). However, it appeared difficult to obtain reliable quantitative information fromexperiments . 1.4 Outlineo f this study

First,a novervie w isgive no fth eexperimenta l apparatusan d procedures. Except for one experiment, reconstituted skimmilk wasuse d throughout thisstudy . The one-dimensional approach did ask for a more thorough analysis of the syneresis pressure and the permeability. In Chapter 3 the original hypothetical description of van Dijk (1982) is defined in some more detail, thereby taking into account the implications of additional experimental results. Attemptsar edescribe d todetermin eth epermeabilit y coefficient of syneresed gels and to find an "endpoint "o f syneresis, i.e. themaximu m shrinkagepossible . In Chapter 4 results of syneresis and permeability measure­ ments under various conditions, such as milk pretreatment, rennetconcentration ,pH ,measurin g temperature,an dadditio no f

CaCl2 or NaClar egiven . The separateeffect s onth epermeabil ­ ity coefficient and the endogenous syneresis pressure are evaluated. The important effect of mechanical pressure on syneresis is treated inChapte r 5.Afte radaptatio no fth emicroscop emethod , accurate and reproducible results on the shrinkage rate with various pressures were obtained. The results are considered in the light of existing theories for the compression of porous media. Thereby, it is tried to subtract some rheological para­ meters from the available experimental results and to link the syneresis behaviour to the rheological characteristics of the proteinmatrix . In Chapter 6 some remarks are made with respect to the results in relation to practical cheesemaking. Hydrodynamic effects during cutting and stirring are briefly described. Calculational results for intermittent pressure and changing pH are given. Also syneresis of gels from preconcentrated milk is evaluated. At the end of this chapter a preliminary three- dimensionalcalculationa lmode li spresented . 2MATERIAL SAN DMETHOD S

2.1Material s

2.1.1Reconstitute dskinmil k

Formos to fth eexperiment sreconstitute dskimmil kwa sused , which wasmad e from low-heat skimmilk powder (Krause,Heino) . Thecompositio no f thispowde r (seeTabl e 2.1)wa sgive npre ­ viouslyb yRoef s(1986) .Th edifferenc ebetwee nth etru eprotei n fractionan d thesu mo f thecasei nan d serumprotei n fraction must forth egreate rpar tb e ascribed toth eproteose-pepton e components.

Table2.1 .Compositio n(i nwt .% )o fth elow-hea t skimmilkpowde remploye di nthi sstudy .

drymatte r 96.8 ash 6.1 fat 0.6 trueprotei n 1) 33.8 casein 28.3 serumprotei n 3.4 NPN 1.8 WPNinde x(ADMI,1971 ) 6.45 1) Mos treference sgiv etota lo rcrud eprotein .Thi s includestru eprotei nan dNPN .Fo rmor einformatio n seee.g .Karma n& va nBoeke l(1986) .

For thepreparatio n of standard skimmilk 10.4g powderwa s dissolved per 100g demineralized water.Whe n other additions hadt ob emad eafterward s(acid ,CaCl 2),th eamoun to fwate rwa s correspondingly adjusted. 100pp mthiomersa l(CjHjHgSCgl^COONa , BDHChemical sLtd )wa suse da sa preservative .Va nDij k (1982) foundn oeffec to fthi spreservatio no nth esyneresi sbehaviou r ofmil kgels .Durin gthi sstud y thepossibl eeffec to fthiomer ­ salo n therennetin g behaviourwa schecked . Theclottin g time wasno taffected . The dispersions were stirred for 20 to 28 hours at 30° C priort orenne taddition ;wit hthi streatmen t afurthe rstorag e ofth ereconstitute dskimmil ka t3 0° Cdi dno tmateriall yalte r 7 itsproperties . For comparison, in some experiments dispersions werekep tovernigh ta t4 °Cafte ra previou streatmen to f 1hou r at 30 °Co r 45 °C.Prio r to rennet addition these samples were storeddurin g 1hou ra t3 0 °C.

2.1.2 Concentratedmil k

Concentrated milk was obtained by ultrafiltering reconsti­ tuted skimmilk at 30 °C in an Amicon concentrator model CH3, equipped with a hollow fiber cartridge (type H1P10, with a molecular cut-off of about 10000 Daltons). The calculated con­ centration factorwa sbase d on the residual volume of retentate andwa schecke db yKjeldah lanalysis .Goo dagreemen twa sfound .

2.1.3 Rawmil k

Rawmil kwa sobtaine d fromth edair y farmo fth eAgricultura l University. To study the influence of milk pretreatment on the permeability of rennet skimmilk gels (see Section 4.2.1), a portion of fresh uncooled raw milk was used. After centrifu- gation the skimmed milk was pasteurized (30 min at 63°C) , cooled to 30 °C and divided into two samples after addition of 100pp mthiomersal .On esampl ewa suse d thesam eda y forexperi ­ ments. The other sample was stored overnight at 4 °C and was used the next day after storage for 1hou r at 30 °C prior to rennetaddition . For an experiment on the influence of fat on the syneresis behaviour (see Section 4.2.7), raw milk, stored overnight at 4 °C, was divided into two portions of which one was centri­ fugea. Bothth e skimmed and theunskimme d milk were pasteurized

Table2.2 .Compositio n (% w/w)o funskimme d and skimmed milktha twa suse di nth eexperimen to nth e influence of fato n the syneresisbehaviour .

unskimmedmil k skimmedmil k

fat 3.45 0.07 protein 3.65 3.84 lactose 4.47 4.66 (30mi n 63 °C) and cooled to 30 °C. After addition of 100 ppm thiomersal, theywer ekep t at 30 "Cunti l rennetwa s added. The composition of both portions, as determined with a Milkoscan 104A/B (A/SNFos sElectric ,Denmark) , isgive ni nTabl e2.2 .

2.1.4 Rennet

Commercial calf rennet (Leeuwarder kaasstremsel) with a strengtho f 10800Soxhle tunit swa sused . Itwa sdilute d shortly beforeus ewit hdemineralize dwater .

2.1.5 Whey

In permeability experiments at pH = 6.7 whey from low-heat whey powder was used. This powder was obtained by renneting a batch of raw milk at 30 °Cwithou t any means of acidification, cutting the gel and collecting the whey. The whey was skimmed, concentrated, spray dried and stored in cans. For experiments 7.3 g whey powder was dissolved per 100 g demineralized water with 100 ppm thiomersal. After dissolving, the whey was clar­ ified in a Christ UJ3 centrifuge (2 0mi n at 2300 x g) . The wheywa s filteredbefor euse . Inothe r caseswher ewhe y wasneeded , reconstituted milk was renneted under conditions corresponding to those during the intended experiment. The gel was cut 30 min after rennet addi­ tion. Thewhe y was separated bycentrifugatio nan d clarified by filtration as described above. Some information about the com­ position of the whey prepared at pH = 6.7 by both methods is giveni nTabl e2.3 .Onl ymino rdifference swer efound .

2.1.6 Aprotinin

In one experiment aprotinin (Sigma Chemicals) was used to check on thepossibl eeffec to fprotei nbreakdow n by indigenous milkproteinase so nth epermeabilit ycoefficien to f thegel .Th e final activity of aprotinin in the reconstituted milk amounted to0.19 8 TIU/ml. Table2.3 .Som ecompositiona ldat ao fth ewhe y(p H= 6.7 ) whenprepare dfro mlow-hea twhe ypowde ran d fromreconstitute dskimmilk .

wheyfro m wheyfro m wheypowde r skimmilk drymatte r( %w/w) 1 6.65 6.67 totalprotei n( %w/w) 2 1.11 1.14 fat( %w/w) 2 0.14 0.13 lactose( %w/w) 2 4.93 4.94 Ca+ M g(mM) 3 12.30 12.52 totalphosphoru s( %w/w) 4 0.44 0.42 1) accordingt oID F21 :196 2 2> determine dwit ha Milkosca n10 4A/ B 31 accordingt ova nde rHav e(1954 ) 4) accordingt oLan g& Miethk e(1933) ,ID F33A :197 1

2.1.7Chemical s

All low molecular chemicals were analytical grade.Appro ­ priatesolution swer eprepare dwit hdemineralize dwater .

2.2Method s

2.2.1Clottin gtim e

Forth edeterminatio no fth eclottin gtim e1 m lo fa dilute d rennet solution was added to 10 ml of milk at 30.0°C .Th e clotting timewa sdetermine d visually bydrawin g a filmalon g the inside wall of a beaker until the first floes appeared. Determinationswer ecarrie dou ti nquadruplicate .

2.2.2Shrinkag erat eo fcur dslab s

2.2.2.1Microscop emetho d without mechanical pressure

Formeasurin gth eshrinkag erat eo fcur dslab si nth eearl y stageso fsyneresis ,th emicroscop emetho ddescribe db yva nDij k (1982)an dva nDij k& Walstr a(1986 )wa sused .Th eprincipl eo f thismetho di soutline di nFigur e2.1a .

10 -whey i—curd slab glass filter plate water immersion objective

water water

thermostatted jacket

Q.

Fig. 2.1. Principle of the microscope method used to measure the shrinkage rate of a curd slab. The radius of the slab was 5 cm, its initial thickness 5 mm. a) without mechanical pressure, b) with mechanical pressure exerted by a sintered glass filter plate.

According toth eintende dexperiment ,HCl ,salt so rNaO Hwer e added five minutes before rennet addition to milk at 30 "C, unless described otherwise. After rennet addition the milk was brought into the thermostatted vat and left for renneting. Shortly before starting syneresis some corundum grains were sprinkled on the surfaceo f thegel .Fo r starting syneresis the surface was wetted, first by spraying and then flooding to preventdamag eo fth egel .Th emicroscop ewa s focuseda ssoo na s possible (generallywithi n6 0s afte rstartin g syneresis)o non e ofth ecorundu mgrains .Th echang ei nheigh to fth eslab ,deter ­ minedb yrefocusin go nth esam egrai nwit hshor ttim eintervals , was read on the scale at themicromete r knobo f themicroscope . Thedept ho f focuswa sabou t 1u man don euni to nth emicroscop e scale corresponded with 2 urn. To find the change in height beforeth e first reading, thereading swer eextrapolate d to the momento fwettin g (seeSectio n3.6) . Inexperiment so nth einfluenc eo fth emeasurin gtemperature , thetemperatur ewa schange dtwent yminute safte rrenne tadditio n byconnectin ganothe rwater-bat ht oth eapparatus . To determine the initial height of the slabs, in separate experiments a gelwa scoole d to 18 °C6 0mi nafte r rennet addi­ tion; syneresisrat ewa s almost zeroa t thistemperature . After focusing on thecorundu m grainso nth esurfac eo f the slab,th e gelwa s sucked away from under theobjective . Subsequently, the microscopewa s refocused on thebotto m of thevat .Thi sexperi ­ ment was carried out six times.Th e average initial height was 4829u mwit ha standar ddeviatio no f2 6um . 11 2.2.2.2 Microscope method with mechanical pressure

Mechanical pressurewa s applied onto a curd slabb ymean s of highly permeable sintered glass filter plates. The radius of these plates was slightly smaller than the radius of the curd slab (se e Figure 2.l b) . The pressure exerted on the curd slab was calculated from the dimensions and the mass of theplates, assuming the density of the glass to be 2600 kg«m"3. Applied pressureswer e8 ,2 7an d 62Pa . During preliminary experiments itoccurre d that covering the bottom of the glass filter plates with filter paper led to somewhat higher shrinkage rates (se e Section 5.2.2). Filter paperwa suse d inal l furtherexperiment scarrie dout . After starting syneresis by spraying and flooding with demineralized water the prewarmed wet glass filter plate was carefully placed on top of the slab 60 s after starting syne­ resis. The change in height was measured by focusing on the surface of theglas s filter plate andwa s read off as described inSectio n2.2.2.1 .

2.2.3 Long-termresidua lheigh t ofcur d slabs

2.2.3.1 Residual height without mechanical pressure

Besidesdeterminin g theshrinkag erat ei nth eearl y stageso f syneresis, itwa s tried to find anequilibriu m stateo f shrink­ age for a curd slab. If the microscope method was used, as described in Section 2.2.2.1, thecontent s of the thermostatted vat markedly changed at times longer than 20 to 25 hours after rennet addition due tobacteria l activity. Therefore, analter ­ native method was followed for the case without mechanical pressure.Cur d slabswer eforme d inPetr idishes, mad eo fglass , witha radiu so f9 cm .Th edishe swer estore d ina thermostatti - callycontrolle d watertan kusin gweighting .Clos eattentio nwa s paid to keep the dishes water-level. The air under the lid prevented the inflow of the surrounding water. Measures were taken to minimize the disturbance due to vibrations during storage. The slabs were flooded with the corresponding whey at 12 one hour after rennet addition. After 25 to 50 hours the resi­ dual height of a slab was determined with a microscope by focusing on the surfaceo f the slab,carefull y removing thege l from under the objective and refocusing on the bottom of the dish. At times longer than 50 hours the results became unre­ liable because of the occurrence of cracks and slits in the slab.Problem sdu et obacteria lactivit ywer eno tencountered . The average initial height of the slabs with a given amount of milk was determined analogous to the description in Section 2.2.2.2.Fo r1 1sample sthi sheigh twa sfoun d tob e480 7u mwit h a standarddeviatio no f3 2um . The repeatability of the method was checked by thedetermi ­ nation of the residual height in 10 dishes at 24 hours after rennet addition for pH = 6.68 and 30 °C. The average residual heightwa s 1489ur nwit ha standar ddeviatio no f3 0urn .

2.2.3.2 Residual height with mechanical pressure

Pressure was exerted and the shrinkage was measured as out­ lined in Section 2.2.2.2.Afte rhavin g removed theglas s filter plate at 25 hours after flooding, the final measurement of the residualheigh twa scarrie dou ta sdescribe d inSectio n2.2.3.1 .

2.2.4 Permeability measurements

2.2.4.1 General

A measure of the resistance exerted on a liquid flowing through a fixed matrix incas eo f a laminar flow isth epropor ­ tionality constant in the equation of Darcy, which states for theflo wi non edirection :

v =(-B/n )• V P (2.1)

v =liqui d flux (i.e.volum eflo w rate/cross-sectional area)(m^s" 1) B =permeabilit ycoefficien t (m2) Tl =viscosit yo fth eflowin g liquid (Pa»s) VP =pressur egradien t (Pa»m_1) 13 The critical value of Re for Eq. (2.1) to hold is at least 0.1 (Scheidegger, 1960). The corresponding value for the cri­ ,r tical liquid flux vcrlt equalsRe«(l-q>) l/(P-6), where 6 is a diameter associated with theporou smediu m inm andwher e(1-

2.2.4.2 Permeability of non-syneresed gels

The tube method developed by van Dijk (1982) and also des­ cribed by van Dijk &Walstr a (1986)an d Roefs (1986)wa s used. Glasstube swit ha lengtho f 25c m and an innerdiamete ro f 4.0 mmwer ecleane dthoroughl ybefor eus ean dseale dwit h laboratory film inwhic h a pinholewa smade .Fo rth eexperiment s they were lowered intest-tube s filled withmil k towhic h rennetha d been added. Themil k was left for renneting at 30 °C.Afte r the gel had become firm enough, the tubes were unsealed and put in a thermostatted plexiglassmeasurin gvat , filledwit hwhey . The whey started flowing through the gel after the level of the whey was raised. The change in the level of the whey in the tubes and the length of the gels was

-A-.'.v.,..;.;-.-..* . determined with a cathetometer. whey Generally the length of the gel ; h(t2) was about 12cm . The initial fc hit,) pressure gradient generally amounted to 5-103 Pa-nr1. Varia­ gel- H tion of the initial pressure ' gradient did not significantly influence the calculated permea­ bility coefficient, as was also Fig. 2.2. Principle of a permeability measurement. For foundb yva nDij k (1982). explanation see text. Van Dijk (1982) has derived the following relationship for 14 the calculation of the permeability coefficient B half-way betweentw oreading s(se eFigur e2.2) :

h(~)-h(t2) { }• n . H hM-hlti) B= (2.2) p .g • (t2-tx) ft(co)= heigh to fth ewhe yleve li nth ereferenc etub e(m ) h(t) =heigh to fth ewhe yleve li nth ege ltub e(m ) H =lengt ho fth ege l(m ) g =gravitationa lacceleratio n(m«s~ 2)

Thepermeabilit y increased linearlywit htim ewithi nth eaccu ­ racyo fth eexperimenta lresults ,a swa sals ofoun db yva nDij k (1982).Consequently ,Be ,th epermeabilit ycoefficien to fa ge l at themomen t thepressur egradien twa sapplied ,coul d beob ­ tained by extrapolation. Results shown are averagevalue s for threeo rfou rtubes .

2.2.4.3 Permeability of syneresed gels

Fora descriptio no fth edeterminatio no fth epermeabilit yo f syneresedgels ,th ereade ri sreferre dt oSectio n3.3.2 .

2.2.5Viscosit yo fth ewhe y

Whenusin gpreoonoentrate dmilk ,th eviscosit yo fth ecorres ­ pondingwhe ywa sdetermine d usinga KP GUbbelohd eviscosimete r (SchottGeräte ,BRD )i na thermostatte dwater-bath .

2.2.6Ge lelectrophoresi s

The proteolytic actiono f indigenous milk proteinases as a functiono ftim ewa schecke db ya quantitativ evarian to fPoly - Acrylamidege lElectrophoresi s (PAE)accordin gt oth eprocedur e describedb yD eJon g(1975) .

15 2.2.7 pH

The pHvalue s were determined after cooling the samples to room temperature with a Radiometer pH meter (type PHM 62), equippedwit ha Schot telectrod e (typeN61) .

2.2.8Calciu m ionactivit y

The calcium ion activity was determined by making use of an Orion 701A digital ion meter, equipped with an Orion 93-20-01 Ca2* ion-selective electrode and anOrio n 90-01 single junction reference electrode, as described by Geerts et al. (1983). Measurements were carried out at 30 °C. The activity coeffi­ cients for the standard solutions were calculated with the Debeije-Hückelequation ,whic hstates :

2 A .z £ •/ I - logY t = (2.3) 1 +B • at • /I

Yj =activit ycoefficien t 2 zL =valenc yCa * 2 a± =io nsiz eparamete r forCa * I =ioni cstrengt hi nm M A,Bar econstant s

Calculations were performed with a± = 4.944, A = 0.5141 and B =0.329 7 (Zoon, personal communication). The slope of the electrodewa sdetermined , asdescribe db yGeert se t al. (1983).

16 3 ONE-DIMENSIONALSYNERESIS :SOM ECONSIDERATION S

3.1 Introduction

The concept of one-dimensional syneresis was introduced by vanDij k (1982),thereb yofferin gpossibilitie sfo rquantitativ e comparison betweentheor y and experimental results.Fo r sakeo f clearness,i nthi ssectio na brie fovervie wwil lb egiven .Fo ra more extensive treatment of the analytical and the numerical approach the reader is referred to van Dijk et al. (1984). His theoretical approach was based on an analytical description of thetranspor to fwhe yinsid eth egel .Thi swa scombine dwit hth e equationo fDarc y (seeals oEq .2.1) :

v =(-B/n )• V P (3.1)

Forconstan tB , r\an d VP thisresulte d ina differentia lequa ­ tion, which is identical with the second equation of Fick for diffusion processes (Crank, 1975). The Boltzmann analytical solution of this equation indicates a proportionality of the shrinkagewit h the squareroo to f timeafte r starting syneresis (t°•5)fo rth einitia lstage so fth esyneresi sprocess .However , during syneresis the pressure on the liquid phase and the permeability of the matrix change with time and place. This precluded finding an analytical solution forth edescriptio n of thewhol eprocess . It was possible to account for these effects by using a finiteelemen tmethod . Inthi snumerica lapproac ha sla bo fcur d with an initial thickness H0 was divided intom thin sliceso f thickness t^ 0. Durin g the shrinkage process the thickness of the slices changes with time and position in the slab (see Figure 3.1). The relative remaining volume i of a slice is definedas :

actualvolum e 1 = (3.2) volumea tt= 0

17 t = 0 t>0 2j AH istrkt smS f Ük.t-Eïk hk.t

Fig. 3.1. Curd slab divided into slices with the parameters for the numerical calculation of the shrinkage rate according to van Dijk (1982).

Using theequatio no f Darcyon eca nwrit e forth eliqui d volume

fluxstr k fromslic ek intoslic ek-1 :

(pk.t -p k-i.t) Vt Ac-i.t , .( + )-i str,k. t (3.3) 0.5 • n Bk. t Bk -1.t

As the shrinkage of the slab is exclusively caused by the loss ofwhey ,on ema ystate :

AVt •< strk.i.t -str k.t) 'A t (3.4)

These equations permit the calculation of the change in thick­ ness of any slice during the shrinkage process. In the calcu­ lationssevera lprecaution sar etake nt oascertai nth ereliabil ­ ityo fth eresult swit hthi snumerica lmode l (vanDijk , 1982). Thenumerica l approach alsoresulte d ina proportionality of the shrinkage to t°• 5 in the early stages of the syneresis process. Together with information about thepermeabilit y coef­ ficient at the very beginning of the syneresis process, this allows the calculation of the initial endogenous syneresis pressurePjj ,fo rwhic hva nDij k (1982)derived :

dAH/dt ( )2 (3.5) Be 0.5 •Q

whereB e isth epermeabilit y coefficient at themomen t of star­ ting syneresis inm 2 anddAH/d ti sth e shrinkage rate inm-s" .1

18 Frommode l calculations avalu eo f about0. 6 forth econstan t Q wasobtaine d (vanDijk , 1982). In this chapter syneresis pressure, permeability and their dependence on time after rennet addition and degree of concen­ tration will be separately discussed. Moreover, the largely hypothetical description of van Dijk (1982) is combined with someadditiona lexperimenta lresults ,leadin gt oa mor edetaile d picture of what happens during syneresiso f rennet-induced milk gels. Compression of thematri xunde r the influenceo f a mechanically exerted pressurewil lb etreate d inChapte r5 .

3.2 Syneresispressur e

The syneresis pressure, i.e. the pressure on the liquid phase, is the cause for syneresis to occur. For a clearunder ­ standing of thecontributin g processes,i t isusefu l todistin ­ guish between the situation before macroscopic syneresis has started and the situation thereafter in a shrinking gel. How­ ever,mos to f thephenomen amentione d forth e initial situation mayals opla ya par ti na shrinkin ggel .

3.2.1 Syneresispressur e ina non-synerese d gel

For the one-dimensional case in the absence of mechanical pressure, the syneresis pressure is made up of the endogenous syneresis pressure and a gravitational contribution asa result of thedensit ydifferenc ebetwee nth ecasei nmatri xan dth ewhe y (vanDijk , 1982).

3.2.1.1. Endogenous syneresis pressure

After a gel (i.e.a continuous network of paracaseinmicel ­ les)ha s been formed still many more junctions can come about, since the paracasein micelles are probably reactive over their entire surface. The formation of new contacts results in local network stresses,wherea s at the same timechange s elsewhere in the network may result in stress relaxation. The height of the 19 pressure on the liquid, the endogenous syneresis pressure, is governed by thebalanc e of theseprocesses .On e can distinguish between: -Th ereactivit y of theparticles .Thi sca nb eexpresse d as an activation free energy for attaining contact between fully renneted micelles as such. It is predominantly due to elec­ trostatic and steric repulsion and depends on experimental conditions (Walstra & van Vliet, 1986). Higher temperatures up to 60 °C result in lower values for the activation free energyo ffull yrennete dmicelle s (Dalgleish, 1983). Thisca n not be fully explained by changes in the surface charge of themicelle s (vanHooydon k &Walstra , 1987). Itwa s suggested that a temperature-dependent steric repulsion betweenmicel ­ lesfro mprotrudin gchain so f ß-caseinma yserv ea sa partia l explanation. Furthermore, the reactivity of renneted parti­ cles becomes higher with a higher Ca2* concentration and lowerwit ha highe rioni cstrengt h (Dalgleish, 1983) . Afterrenne tadditio nreactiv esite sar eformed .Thei rnumbe r gradually increases until all ic-casein has been split (van Hooydonke t al., 1984). - The probability of approach. This is determined by several factors. It is likely that a shorter distance between reac­ tive sites and the existence of highly flexible strands facilitate the formation of new contacts. This may for example be the case with a higher volume fraction of the building blocks, either locally or for the whole gel. The formation of new contacts is expected to be most prominent during the early stages of gelation, when some dangling strands still occur (Walstra et al., 1985). As gelation proceeds, these strandswil l becomemor eclosel y attached to thecasei nmatrix . Individual strandsals obecom emor e rigid asa result of an increase inth enumbe ro f bondspe rcross - sectionalare adurin ggelation .Thi slimit sthei r flexibility and increasesthei rresistanc eagains tbending . - The breaking of strands.Thi sma y occur under the influence of small local tensile stresses, caused for example by the formation of contacts elsewhere, or by thermal motion. It offers possibilities for the formation of new contacts. At 20 the same time it results in a lowering of the pressure exertedo nth eliqui dphase .Th estrengthenin g ofth estrand s during gelation gradually diminishes the probability of breaking. Walstra et al. (1985)estimate d thebreakin g force of a strand to be of theorde r of 10"11 N. From experiments with mechanically exerted pressure, corresponding values could be calculated (seeSectio n 5.5.3). Breaking of strands presumably depends on the relaxation behaviour of the bonds in the junctions. It should be realized that there is no singlerelaxatio ntime ,bu ta whol espectrum .A sa resul tth e relaxation behaviour in the junctions, the contact area betweenth ebuildin gblocks ,i sdetermine d byth estress ,th e timescal ean dth etype san dth enumbe ro fbond spe rjunctio n (Ferry,1980 ;Roefs , 1986;Zoo ne t al.,t ob epublished) . - Internal rearrangements in the strands. Over longer time scalesth eendogenou s syneresispressur ema y also relax as a result of internal rearrangements in the strands. In such cases a local tensile stress has not or not yet resulted in the breaking of the strand, leaving time for breaking and reforming of bonds between individual protein molecules in the junctions. Also for this process a strong dependence on the relaxation behaviour of the bonds in the junctions must exist.

These processes determine the magnitude of the endogenous syneresis pressure as a function of time after rennet addition in a non-syneresed rennet-induced skimmilkgel .Onl y aqualita ­ tivedescriptio n canb eoffered .Detaile d information about the spatial arrangement and the rheological properties of the individual strands and theireffec to n theendogenou s syneresis pressure is lacking and will be hard toget .Th e lowvalue s of this pressure (0 - 3 Pa),whic h precludes direct measurement with sufficient accuracy, are partly responsible for this. The endogenous syneresis pressure for the whole gel should be envisaged as a kind of system-averaged result of processes in the network. This pressure has a very momentary character. A pressurebalance ,e.g . asgive ni nSectio n5. 5 forth ecas ewit h mechanicalpressure ,doe sno tapply . 21 As a result of the above mentioned processes, rennetmil k gelssho wmicrosyneresi si nth eregion swher eth enetwor kcanno t shrink;thi si sth erearrangemen to f strands,leadin gt odens e andles sdens eregion san do naverag ewide rpores .Thi sresult s ina nincreas eo fth epermeabilit ycoefficien twit htim eafte r rennetadditio n(va nDijk ,1982 ;va nDij k& Walstra ,1986) .Wit h acid caseingel s at pH < 5.2 an increaseo f thepermeabilit y coefficient with time after the onset of gelation was not observed. This was explained by a sharp change in thecasei n particle structure and the relaxation behaviour of theinter - particlebond saroun dp H= 5. 2(Roefs ,1986) .

3.2.1.2Influenc e of time after rennet addition

InFigur e3. 2th echang eo fth eendogenou ssyneresi spressur e withtim eafte rrenne tadditio ni na ge lo fno npreconcentrate d milk (i= 1 )a t3 0 'Ci sshow nfo rthre evalue so fth epH .I n theearl y stageso fth e gelationproces s thebuilt-u po fth e

Po(Pa)

1.0 1.5 2.0 time after rennet addition (h)

Fig. 3.2. Calculated initial endogenous syneresis pressure of rennet skimmilk gels as a function of time after rennet addition. Influence of pH. 500 ppm rennet; 30 'C; pH * 6.68 (•), pH - 6.48 (A),p H - 6.33 (•). ( ) is an assumed extrapolation to the clotting time. 22 pressure by the formation of new contacts is only partly coun­ teracted by relaxation and strengthening processes.Th emaximu m can be explained by the decrease in the number of dangling strands, an increase in the inhomogeneity of the matrix as a result of microsyneresis (so a larger average distance between strands), thestrengthenin go fth estrand san dth erelaxatio no f the stress due to breaking of strands and internal rearrange­ ments. The change of the endogenous syneresis pressure with time after rennet additionwa sclearl y affected bypH .A t a lowerp H the maximum for the endogenous syneresis pressure was found at anearlie r time after rennet addition. Thisma y beexplaine d by assuming that allprocesse s are faster,althoug h toa different degree,resultin g ina highe rmaximu m ata lowe rp Hbetwee np H= 6.7 and pH = 6.3. Thenumbe ro f reactive sitesincrease s faster as a result of a higher rate of the renneting reaction (van Hooydonk et al., 1986b). The reactivity of the paracasein micelles themselves may also be influenced by pH; whether this is due to changes in steric or electrostatic repulsion or both remains to be answered (van Hooydonk & Walstra, 1987). The foregoing is likely to result ina highe r stress inth e strands anda large rpressur eo nth eliqui dphase .Becaus eo fth efaste r increase in the number of reactive site and a possible higher reactivityo fthes esite sals oa highe rrat eo f strengthening of the strands is expected. Moreover, the increased solubilization ofth ecolloida lcalciu mphosphat ea ta lowe rp Hma ypromot eth e fusion of themicelles ,i.e . a faster increasei n thenumbe r of bonds per junction. This may hinder the formation of new con­ tacts and result in an earlier drop of the initial endogenous syneresispressure . The higher maximum values at a lower pH can be explained by assuming that in theearl y stageso f gelation therelaxatio n of thestres s inth e strandsdoe sno toccu r fastenoug ht ocompen ­ sate for the higher reactivity of the particles.Th e relaxation of the stress in the strands may even be slower because of a highernumbe ro fbond spe rjunctio na ta lowe rpH . Intramicellar interactions may also be affected as a result of conformational changes of the casein molecules, influencing 23 the possibilities for internal rearrangement. Results of Theo­ logicalmeasurement s showeda fasterincreas eo fth emodul ia ta lower pH between pH = 6.7 and pH = 6.3. The relativecontribu ­ tionsb yth evariou stype so fbond sdi dhardl ychang ebetwee np H = 6.7 and pH= 6.3 (Zoone t al., tob e published). This points to the number of bonds as a function of time being the more important factor concerning the relaxation behaviour in the strands.

3.2.1.3 Influence of milk preconcentratlon

From experimental results obtained with ultrafiltered milk, highervalue s forth einitia lendogenou ssyneresi spressur ewer e calculated at a further degree of preooncentration, as can be seeni nFigur e3.3 . Thisma yb eexplaine d bya highe rnumbe ro f

1.0 15 2.0 time after rennet addition(h )

Fig. 3.3. Calculated initial endogenous syneresis pressure of rennet skimmilk gels as a function of time after rennet addition. Influence of milk preconcentration. 500 ppm rennet, 30 *C. pH - 6.68 ( ),i = 1.0 (•), i = 0.75 (à),i = 0.6 (•) pH = 6.33 ( ),i = 1.0 (o),i = 0.75 U), i = 0.6 (0) 24 reactivesite san da shorte raverag edistanc ebetwee nthem .Thi s promotesth e formationo f new contacts.I t shouldb enote d that especially at pH = 6.33 the calculated endogenous syneresis pressure was found to vary rather erratically with time after rennet addition. This ma^ be because only single syneresis measurementswer eperforme dwit hgel sfro mpreconcentrate dmilk . Still, an effect of the degree of concentration on the time after rennet addition, at which the maximum value for P^ is reached,appear sabsent .Thi ssuggest stha tth estructur ea tth e levelo fth eindividua lstrand si shardl yaffecte db ypreconcen - tration of themilk , although Theological measurements resulted in a faster increase of the macroscopic moduli with time at a higherdegre eo f preconcentration (vanDijk , 1982;Zoo ne t al., to be published). Apparently, preconcentration of the milk results in a higher average number of strands per cross-sec­ tional area during the early stages of gelation, although a higheraverag ethicknes so fstrand si sexpecte d aftersom etime .

The influence of preconcentration on syneresis will be furtherdiscusse d inSection s 5.4.an d 6.6.

3.2.1.4 Gravitation-Induced syneresis pressure

Besides the endogenous pressure, gravitational forces can contribute to thepressur e on the liquid phaseaccordin g to the experimental setup,althoug hthe yar eo fmino rimportanc edurin g the early stages of shrinkage in a syneresing slab (van Dijk, 1982). Oneca ndeduc e the following relationship for thegravi ­ tationalpressure :

9 P = q>• A P •g • hc (3.6) where

3.2.2 Syneresispressur e ina syneresingge l

Up till now no adequate method has been developed for the directmeasuremen to fth esyneresi spressur ean dchange s therein during syneresis.Fo r theendogenou s syneresis pressure aswel l as for the gravitation-induced pressure only a qualitative descriptionca nb egiven ,resultin g ina se to ftria lfunctions .

3.2.2.1 Endogenous syneresis pressure

The local endogenous syneresis pressure in a syneresing gel maydepen don : - The rheological properties of the strands. For shrinkage to occur, deformation of strands is needed. The deformation is determinedb yth estres san dth etim escal eo fdeformatio n in relation to the rheological properties of the strands. For givenexperimenta lconditions ,thes edepen do nth etim eafte r rennetadditio nan dmayb eals oo nth eloca ldegre eo fconcen ­ tration. - The possibilities for liquid flow. Better possibilities for liquid flow will result in greater shrinkage and a faster dropo fth esyneresi spressure .Th eliqui d flowi sdetermine d by the local permeability of the matrix and the local pres­ sure gradient. Initially, the situation for liquid flow to occuri smos t favourablei nth eoute rlayers . - The local rearranging tendency of the strands.Th e processes at the level of the strands and the building blocks, which are responsible for the endogenous syneresis pressure, were discussed inSectio n3.2.1.1 . Initially,on eha st odea lwit h Pjj.Th eendogenou s syneresispressur e canb e affected by the local condensation of thematrix . For instance,th e distance betweenreactiv e sites is lowered. On theothe r hand shrink­ ageca n locally promote the formation of thicker strands and 26 thus a stiffening of the casein matrix. This retards the formation of new contacts and the exertion of pressure on theliqui dphase .

These factors will make the change in the endogenous syneresis pressure during shrinkage to depend on time and place in the gel. One can try to account for all these factors in mathema­ tical equations, but the lack of detailed quantitative infor­ mationma yinterfer eth esucces so fsuc ha napproach .Expressio n of the resistance of the matrix in a kind of reaction force (Walstra &va nVliet , 1986)ca nremin d oneo f the importanceo f the above mentioned processes, but the very low pressures involved prevent the collection of exact information about the relevant rheological parameters, such as the modulus, the deformation and the time scale in their mutual relation and their dependenceo n concentration. Such an approach may be more appropriatei fmechanica lpressur ei sexerted . For the present situation itwa s not tried to make any kind of subdivision for theprocesse s contributing to the endogenous pressure during syneresis, nor was it tried to account for effectso ftim ean dplace .I nmode lcalculation stria l functions wereused , inwhic honl y theinfluenc eo f thedegre eo fconcen ­ tration was taken into account. These trial functions are presented in Figure 3.4. To simplify matters, it was assumed that the endogenous pressure becomes zero when the relative remainingvolum ei s0.3 ,althoug h furthershrinkag ei scertainl y possible (seeSectio n 3.4). However,thi swil lno tinfluenc eth e calculated shrinkage rate during the first hour after starting syneresis. Van Dijk (1982) hardly found any effect from the shape of his trial functions on the calculated shrinkage rate. From his work trial function (1) was adopted after a small adjustment.Base do nth eresult swit hpreooncentrate dmil ka few other functions were introduced, in which the endogenous syne­ resis pressure at first remained at a higher level as compared to the trial functions used by van Dijk (1982). Some calcula- tional results with these trial functions will be discussed in Section 3.5, after having considered the other parameters relevant forth eshrinkag erate . 27 P0(i)/Po(i=D 0.3 s 1 s 1 1.0

i - 0.3 2) p!?(i) « P?(i-1) • 0.3 s i s1

i - 0.3 3) p!!(i) - Pn(i=D • 0.3 s 1 s 1

4) P=(i) - Pn(l-l) 0.5 < 1 s 1

i - 0.3 P„(i) - Pn(i=l ) • 0.3 s i s 0.5

Fig. 3.4. Trial functions for the endogenous syneresis pressure as a function of the degree of concentration.

3.2.2.2 Gravitation-induced syneresis pressure

The contribution of gravitational forces to the syneresis pressure in a syneresing gel must be approached in a different manner. As with mechanically exerted pressure, the total grav­ itation-induced pressure must be accounted fordurin g thewhol e syneresisprocess .Fo rthi sdescriptio nreferenc eca nb emad et o thepiston-sprin g analogyo fTerzagh i (1965), whichi sdescribe d in Section 5.5.1. During syneresis the visco-elastic character of the stiffening matrix leads to a gradual decrease of the pressureo nth ewhey ,theoreticall yunti lth ecloses tpackin g is achieved. Although the relationship may depend upon time and place in the gel, for model calculations a linear decrease of theinitia lvalu ewit ha lowe ri wa sassume d (vanDijk , 1982):

P9 =P g .1. 5 •(V t - 1/3) (3.7)

where Pg denotes the local gravitation induced pressure before

syneresisha sstarte d andi ^ t representsth erelativ e remaining volume.

28 3.3 Permeability

Theresistanc eagains t flowthroug hth ematri xca nb eexpres ­ sed by the permeability coefficient (Scheidegger, 1960). It is determined by the spatial arrangement of the solid phase.Fo r a given volume fraction of the solid phase, a more inhomogeneous matrix will result in a higher value for the permeability coefficient. At constant conditions the permeability of rennet skimmilk gels was found to increase with time after rennet addition. Deformation of rennet skimmilk gels may further promote this increase (vanDijk , 1982;va nDij k& Walstra , 1986). In this section the attention will be focused on the change ofth epermeabilit ycoefficien twit hth edegre eo f concentration under varying conditions.On eca nobtai n a relationship between the degree of concentration and the permeability coefficient with gels from ultrafiltered milk. However, it was unanswered whetherthes evalue sclearl yreflec tth esituatio ni fth ege li s concentrated by syneresis.Attentio n is paid to this aspect in Section3.3.2 .

3.3.1 Permeability ofgel s frempreconcentrate d milk

The measurements performed by van Dijk (1982)wer e repeated withreconstitute d skimmilka tvariou sp Han dmeasurin gtempera ­ ture. In all cases 500 ppm rennet was used, leading to a some­ what longer clotting time for a higher degree of preconcentra- tiona t3 0 °C(result sno t shown). The initial pressure gradient across the gel was varied, in order to obtain a reasonable flow rate (0.1 - 0.2 mm-min"1) during measuring in each case. The calculated permeability coefficients for the moment at which the gel was pressurized were fitted to power curves.Th e relations with the time after rennet addition and the degree of concentration appeared to be additive. The obtained relationships are given in Table 3.1. A higher initial value for the permeability coefficient and a greater increase with time were found with a lower pH and also witha highertemperature . Thiswil l furtherb edeal t withi n 29 Table3.1 .Calculate d relations forth epermeabilit y coefficient of rennetmil k gelsa sa functiono f timean ddegre eo fconcen ­ tration (by UF). Acidification with HCl at 5 min before rennet addition, 500 ppm rennet added at 30 "C, changed to measuring temperaturea t2 0mi nafte rrenne taddition .

measuring temperature pH <°C) calculated relation r2 6.68 30 Be = (2.1.10-13.i30)+(2.l.lO-i' .i30 «t ) 0.9987 6.33 30 Be = (3.1.10-13-i2-9)+(3.2.10-i' »i2-8-t ) 0.9991 6.68 34 Be = (2.7-10-13.i3-2)+(5.l-lO-i' -I30 -t ) 0.9891

1 =degre eo fconcentratio n (volumeafte rUF/origina lvolume ) t =tim eafte rrenne tadditio n- 180 0 (s)

Chapter 4. For the influence of thedegre eo fconcentratio n the exponent was found to be 3.0, depending only slightly on the experimental conditions. Thisvalu ewa s somewhat lower than the one obtained for acid skimmilk gels (Roefs, 1986) and slightly higher than found before for rennet skimmilk gels (van Dijk, 1982). The exponent for the influence of time was also about 3.0. Thismean s that the relativechang e in thepermeabilit ycoeffi ­ cientwit h time after rennet additiondoe shardl ydepen d on the degreeo fconcentratio n forgive nexperimenta lconditions .

3.3.2 Permeabilityo f syneresed gels

In separate experiments it was also tried to determine the overall permeability coefficient of a syneresed gel in a more directway .Thi swa sachieve db yusin gth eapparatu soutline d in Figure 3.5. Itconsiste d of a beaker (0 = 100mm )wit h a piece of glass filter plate (0 = 26.2 mm),tha t was fixed in the bottom, and a removable funnel part.Durin g theexperiment s the apparatus was kept in a thermostatically controlled water-tank at3 0 °C. Age lwa sprepare db y fillingth eapparatu swit hskimmilk , to whichrenne tha dbee nadded ,throug hth efunne lpart .Th ebeake r was removed at one hour after rennet addition and carefully clanged to another identical funnel part. This was filled with the corresponding prewarmedwhey . Thege lwa s flooded andcu t 30 whey-

beaker- gel

-funnel part glass filter

during gelation during measurement

Fig. 3.5. Principle of the apparatus for measuring the permeability of a syneresed gel. loose from the wall. In case of experiments with mechanical pressurea filterpape rcovere dglas sfilte rplat ewa splace d on topo fth egel .Th ege lwa slef tt osyneres efo r2 5hrs . At first, a measurement was started by raising the level of thewhe y inth e funnelpar t some 60t o 80m m above the levelo f thewhe y inth ebeaker .Th etim eneede d forth edro po fth ewhe y level over a certain distancewa s determined. Repeated measure­ ments showed an increase in the flow-through time and defor­ mation of the gel above the glass filter plate was clearly visible. In later experiments the beaker was completely filled with whey before the measurement was started. The level of the wheyi nth e funnelpar twa sraise d 5t o1 5m m aboveth eleve li n a reference tube. Then the liquid flux through the gel was determined, keeping the level of the whey in the funnel part constant. The liquid flux was measured by making use of an autoburette (Radiometer, type ABU 13) and a cathetometer. With this procedure the liquid flux was almost constant with time. Theinitia lvalu ewa suse d forth ecalculatio no fth epermeabil ­ ity coefficient of the syneresed gel. The level of the milk before clotting and the residual height of the gel were deter­ minedwit hcallipers .

Thesituatio ndurin gmeasurin gca nb econsidere d asa cas eo f two-dimensional stationary flow. To let the calculation of the

31 Overall permeability coefficient not become too unwieldy, the followingassumption swer emade : - theresidua lge li shomogeneou san disotropic ; - theshap eo fth eresidua lge lremain scylindrical ; - thepermeabilit y coefficient of thege l in theglas s filter plate can be calculated from the relation for the change in thepermeabilit ycoefficien twit htim efo rnon-synerese d gels (seeSectio n 3.3.1). Especially, the first assumption ishighl y questionable and the consequenceswil lb epointe dou ta tth een do fthi ssection . For the calculation of the overall permeability coefficient usewa smad eo fa nimplici tfinit edifferenc emethod .A mathema ­ tical grid was introduced to calculate the equilibrium flow potential distribution in the gel, a method often used for the numerical modeling of groundwater flow (Southwell, 1940;Bea r & Verruijt, 1986). The value of the potential directly above the glass filter plate was obtained after correcting for the pres­ suredro pove r thisplate .Fo r thecalculatio no f thispressur e drop, the equations in Table 3.1 were used, taking 1 = 1. The iteration process was accelerated by using a successive over- relaxation factor of 1.4. Examples of the resulting calculated pattern of equipotential surfaces during measuring for a rela­ tively high and a relatively lowresidua lheigh to f thege l are shown in Figure 3.6. The flow lines are perpendicular to this pattern. The liquid flux through the surface of the slab was determined bynumerica l integration,usin gth eequatio no fDarc y with an assumed value for B. The overall permeability coeffi­ cient forth ege lwa s obtained after comparisonwit hth e liquid flux foundexperimentally .

The results forvariou sexperimenta l conditions are shown in Table 3.2. A lower permeability coefficient was found with a higher degree of concentration and with a higher pH, in accor­ dancewit hth eresult sobtaine dwit hnon-synerese d gels. A fewremark sshoul db emad ebefor ecompariso no fth eresult s withthos eobtaine d forgel sfro mpreconcentrate d milk.Extrapo ­ lation by means of the equations in Table 3.2 to over 20 hrs afterrenne tadditio nfo r1= 1 probably leadst oa noverestima - tiono fth e permeabilitycoefficien t forth ege li nth efilte r 32 -glass filter plate

Fig. 3.6. Equipotential lines in syneresed gels with varying thickness during the flow of whey as in Fig. 3.5, calculated for the assumptions of homogeneity and isotropy.

Table3.2 .Compariso no fth epermeabilit ycoefficien twit h various degree of concentration for syneresed gels at 26 hrs after rennet addition and for gels from UF-preconcentrated milk (see Table 3.1). 500pp mrennet ,3 0 °C.Fo rsynerese d gels syneresis was started at 1 hour of rennet addition.

concentrated by syneresis UF initial1> corrected preconc. syneresis residual pressure 1013 1013 pressure height gradient *B *B 1 2 2 pH (Pa) (mm) i (kPa-nr) (m ) (m ) 6.66 0.7 17.3 0.71 4.6 6.2 7.9 0.7 17.3 0.71 1.3 5.5 8.0 8.5 4.7 0.21 20.0 0.8 0.2 8.5 4.7 0.21 10.4 0.6 0.2 27.8 3.7 0.16 24.9 0.7 0.1 27.8 3.7 0.16 12.4 0.7 0.1

6.33 1.6 11.9 0.53 7.4 4.5 5.7 1.6 11.9 0.53 3.8 4.6 5.8 9.4 4.4 0.20 22.8 1.0 0.4 9.4 4.4 0.20 12.4 0.8 0.4 28.7 3.6 0.16 28.2 0.7 0.2 28.7 3.6 0.16 15.1 0.6 0.2

1> initia lendogenou ssyneresi spressur e+ mechanica l pressure plate. The rate of change for the permeability coefficient namely tends to decrease with time after rennet addition (van Dijk, 1982; see also Chapter 4).Fo r instance, halving this 33 value for the first case in Table 3.2 (which is probably more than the maximum deviation possible), the permeability coeffi­ cient forth esynerese dge lwa s foundt ob e9.0-10" 13 m2 instead of 6.2«10" x3 m2. This effect will be less marked for gels further syneresed, because of the relatively smaller pressure dropove rth ege li nth eglas sfilte rplat ei nsuc hcases . Also the inhomogeneity of a syneresed gel causes an under­ estimation of the permeability coefficient at a certain degree of concentration. The effect was estimated for the same case, using two different trial functions for the concentration profile (see Figure 3.7) and B =B(i=0.71)-i 2• 5/0.42 5 for the relation between B and the local degree of concentration. The overall i is the same for both trial functions. The calculated overall permeability coefficients for inhomogeneous gels were 4.8«10~13 m2 whenusin g trial function 1)an d 1.5-10"13 m2 when using trial function 2) So, it can readily be assumed that the permeability coefficient for a givendegre eo f concentration is estimated too low if it is determined in an inhomogeneousgel . Gelswit h a lowresidua l iar eexpecte d tohav ea lesspronoun ­ ced concentration profile. For these gels clearly a higher permeability coefficient was found in thecas eo f concentration bysyneresi sa scompare d togel sfro mpreconcentrate d milk.

1) i = 1 - 0.817 • U/Hn)

J 2) i = 1 - 1.738 • (x/Hn)

x/H0

Fig. 3.7. Trial functions for the concentration profile in a syneresed gel as used for the estimation of their effect on the overall permeability coefficient (see text). 34 Finally, it should be noted that one-dimensional syneresis may cause differences between the permeability coefficient in the horizontal and in the vertical direction. The latter will probablyb eaffecte d toa lesse rexten tdurin gshrinkage . It is concluded that the permeability coefficient of a syneresed gel is underestimated by using the results obtained with preconcentrated milk. In Section 3.5 results of model calculations with exponents of 3.0 and 2.0, the latter being a rough estimate based on the results described in this section, in theequation s for thepermeabilit y coefficienta s a function oftim ean ddegre eo fconcentratio nwil lb eshown .

3.4 Endpoin to f syneresis

Itmus t be assumed that theen dpoin to f syneresis forgive n conditionsa tleas tequal sth eequilibriu mmoistur econtent .Th e latter isdetermine d by anequilibriu m stateo f swelling of the particles in the interstitial moisture. As was discussed by Walstra et al. (1985), only smallquantitie s ofwate rar eboun d to specific groups of the protein molecules (« 0.1 g water/g paracasein). Most water is imbibed in the curd, either between theprotei nmolecule s in theparticle so r between the particles themselves. In someexperiment s during this study gel slabswer e allowed to synerese for 20 to 50 hrs in Petri dishes. The residual heightwa sdetermine dwit ha microscop e (seeSectio n 2.2.3). The results are summarized in Figure 3.8. To give a complete pic­ ture, also the results obtained with mechanical pressure are included. In neither case an end point of syneresis could be established. Theoccurrenc eo fcrack srestraine dcontinuatio no f theexperiment s forlonge rtimes . The equilibrium amount ofwate r per gram paracasein micelles at room temperature and physiological pH was estimated to be about 1.4 (Walstra et al., 1985). When neglecting interstitial moisture,thi swoul dcorrespon dwit ha relativ eremainin g height of at most 0.07. All experimental values were higher, except maybe for pH = 6.33 and P" = 62 Pa. For further shrinkage to occur, itmus t beassume d thatrelaxatio n processes in the 35 60 0 20 30 40 50 60 time after rennet addition (h)

Fig. 3.8. Relative residual height of rennet curd slahs as a function of time after rennet addition. Influence of pH, mechanical pressure, rennet concentration and measuring temperature. Petri dish method except when mechanical pressure was applied (microscope method); renneting temperature = 30 'c. changed to measuring temperature at 20 min after rennet addition, initial height = 5 mm. 500 ppm rennet ( ),25 0 ppm rennet ( ),3 0 'c (e),34 'c (*).

strandsan di nth emicelle sca nlea dt odeformation ,whereb y the amount of interstitial moisture is gradually decreased. Possi­ bly, protein breakdown by rennet enzymes or plasminma y play a part in this. However, halving the amount of rennet did not affectth eshrinkag ei nth eabsenc eo fmechanica lpressure . From Figure 3.8 itca n be seen that a lower pH and a higher temperature resulted in lower values for i. This can partly be ascribed to a higher permeability coefficient at lower pH and higher temperature (see Chapter 4), which speeds up theexpul ­ sion of whey. At the same time the lower voluminosity of the micelles themselves alsoma y contribute toth eobserve d effects (Darling,1982 ;va nHooydon ke t al., 1986a). Inmode l calculations theen d point of syneresis was assumed to be 0.3. According to the experimental results lower values occur, depending upon the experimental conditions and the time after rennet addition. However, changing the value for the end point of syneresis inth e trial functionswil l only havea very limited effect on the calculated syneresis rate in the early

36 stages of the process. Therefore, the assumed value was not adapted.

3.5 Resultso fmode l calculations

3.5.1 The effect of various trial functions forth e endogenous syneresispressur e

Theresult so fmode lcalculation swit hth etria l functionsi n Figure3. 4 arepresente d in Figure3.9 . Thegravitationa l force was kept zero in these calculations in order to make a clear comparison possible. For the initial endogenous syneresis pressure 1 Pa was taken. The first equation in Table 3.1 was used for the calculation of the permeability coefficient as a functiono ftim ean ddegre eo fconcentration .Th einitia lheigh t of the slab was 5 mm. From Figure 3.9a it can be seen that in all cases the initial shrinkage was proportional to t°• 5. The increase of the permeability coefficient with time after rennet addition is responsible for the deviation from the initial proportionality to t°• 5 in Figure 3.9a. The calculated values forQ inEq . (3.5)wer e0.54 ,0.55 , 0.53 and0.5 0 forth e trial functions 1, 2, 3 and 4, respectively. The difference with valueso f0.58-0.62 , foundb yva nDij k (1982), mustb ecause d by the assumed absence of gravitational forces in our case and by thedifferin g relationship forth epermeabilit ycoefficien ta sa function of time and degree of concentration (see also Section 3.5.2). Only small differences in the shrinkage rate were detected for trial functions 1, 2 and 3. With trial function 4 thecalculate d shrinkagerat ewa ssomewha tlower . In Figure 3.9b the calculated concentration profiles after the slabha s shrunk to 0.8 timesth eorigina l height areshown . Itca nb esee ntha tth elowe rshrinkag erat ewit htria l function 4 must be due to the relative high degree of concentration and thus the lower value for the permeability coefficient in the outer layers. If the pressure in the outer layers remains at a high level during concentration, a poorly permeable skin is formed. This results indelaye d shrinkage despite ano n average higherpressure . Forth econsidere d conditionsth epermeabilit y 37 AH t^irn)

800-

600-

400-

200-

O-i ^- 1 1 1 ; 1 1 1 1 1 20 40 60 80 100 •F .05,

Fig. 3.9. Results of model calculations for the one-dimensional shrinkage with time (Fig. 3.9a) and the concentration profile after 20 %shrinkag e (Fig. 3.9b) of rennet skimmilk gels. Influence of different trial functions for the change in the endogenous

syneresis pressure (Bee Fig. 3.4).H Q * 5 mm; B = B(i,t) (see first equation in Table 3.1); P;*= 0. P^ » 1 Pa, trial function: 1) (o),2 ) (Q), 3) (0).4 ) (•).

38 in theoute r layers ist o agreate rexten t rate-determining for thesyneresi srat etha nth eleve lo fth esyneresi spressure . Taking intoaccoun tth elarg eeffec to fth edegre eo fprecon - centration on the initialendogenou s syneresis pressure, it was felttha ttria lfunctio n3 ma yb eth emos trealisti con efo rth e dependenceo fth eendogenou ssyneresi spressur eo nth edegre eo f shrinkage. However, applying other trial functions will only havea limite deffec to nth ecalculate d shrinkagerate .

3.5.2 The effect of different relations for thechang eo f the permeabilitycoefficien t

In Section 3.3.3 it was shown that the permeability coef­ ficient as a function of time after rennet addition and degree of concentration is underestimated by using thevalue s obtained from experiments with preconcentrated milk. Results of calcu­ lations with an exponent of 2.0 instead of 3.0 in the first equation from Table 3.1 for a slabwit ha n initial thicknesso f 5 mm are shown in Figure 3.10. The shrinkage rate is found markedly higher in such a case (see Fig. 3.10a). For Q in Eq. (3.5) a value of 0.75 instead of 0.55 was calculated. Q thus dependso n thechose nexponen t in B{±,t). Theshrinkag e profile afterth esla bha sshrun kt o0. 8 timesit sorigina lthicknes si s not much affected (see Fig. 3.10b), although a somewhat lower degree of shrinkage of the inner layers was found with an exponent of 2.0. This is an indirect effect of the higher shrinkage rate,whic h limitsth econtributio n from the increase of the permeability coefficientwit h timeafte r rennet addition (seeFig .3.10c) .O nth eothe rhand ,on emus tconclud etha tals o in thiscas e syneresis rate ispredominantl y determined by what happens inth eoute rlayer so fth eslab .

3.6 Calculationo fth einitia lendogenou ssyneresi spressur e fromexperimenta l results

Collection of accurate experimental data on the initial shrinkage rate proved to be rather difficult. The change in height couldonl yb efollowe d fromabou t4 5s afte rmoistenin g 39 AH (um) 1200

Fig. 3.10. Results of model calculations for the one-dimensional shrinkage with time (Fig. 3.10a), the concentration profile after 20 % shrinkage (Fig. 3.10b) and the corresponding permeability profile (Fig. 3.10c) of rennet skimmilk gels. Influence of the exponent in the relation for the change in the permeability with degree of concen­

tration (see first equation in Table 3.1).H Q = 5 mm; P^ = 0, P^ = 1 Pa, trial func­ tion 3) for PS (i); B 40 thesurfac eo fth egel .Thereby ,especiall y inth ecas eo fa low shrinkage rate, a considerable error in the values read off at the micrometer knob was to be expected. In Table 3.3 some experimental results are presented for two values of the pH after fitting the shrinkage to power curves of the shapeA H = a-tT, where t is the time after starting syneresis in s. From Table 3.3 it can be seen that a considerable spread in the values fora an d roccurred ,especiall ya tp H =6.68 .

Table3.3 . Least squareestimate so f a and r inA H = a'tT for the shrinkage of rennet milk gels till 5 min after starting syneresis. 500 ppm rennet, microscopemethod ,3 0 °C. estimated 6 pH ta 10 -a r AH(t=300s) (hrs) (m-s"r) (um)

6.68 0.5 3.83-9.91 0.53-0.39 79- 92 0.75 0.93-0.044 0.82-1.26 165- 58 1.0 3.50-3.22 0.63-0.63 127-117 1.5 4.64-5.98 0.61-0.56 151-146 2.0 1.61-0.76 0.74-0.87 110-109 3.0 2.09-6.36 0.72-0.54 127-138

6.33 0.25 1.08-0.93 0.90-0.92 183-177 0.3 1.57-1.17 0.86-0.89 212-187 0.4 4.44-2.05 0.69-0.81 227-208 0.5 3.89-3.08 0.71-0.75 223-222 0.75 3.71-4.60 0.73-0.69 239-235 1.0 2.38-3.65 0.77-0.72 192-196 1.5 2.72-2.05 0.75-0.78 196-175

ta =tim eo f starting syneresisi nhour safte rrenne t addition

Nevertheless, the proportionality of the shrinkage to t°• 5 during the first five minutes after starting the syneresis processappear squestionable .Especiall y theresult sobtaine d at pH =6.33 ,whic har eassume d tob emor ereliabl edu et oa highe r shrinkage rate, suggest a higher value for the exponent. This deviation from the square root proportionality could not be ascribed to the effect of gravitational forces or the increase of the permeability with time, as was established with model calculations (van Dijk, 1982; see also Section 3.5).Further ­ more, r tended to be higher when syneresis was started sooner

41 after rennet addition,whic hpoint st oth ege lpropertie s being responsible. Fora possibl eexplanatio no fth edeviatio n fromth e propor­ tionalityo fth eshrinkag et ot °• 5,th eTheologica lbehaviou ro f thege lshortl yafte rth eonse to fsyneresi smus tb econsidered . An proposed explanation is illustrated in Fig 3.11.I ti s presumed that inth eearl y stageso fth eshrinkag e processth e deformability ofth ematri x isth erate-determinin g factor.A t the onset ofsyneresis , a very small momentary elastic defor­ mation (shrinkage) will occur. Afterwards, thenorma l visco- elastic behaviouro frenne t milk gels (vanDijk , 1982)wil lb e displayed. This maylimi t theshrinkag e rate tovalue s lower than derived fromth emode l calculation; i.e.th eoute r layers cannotb econcentrate d atth erat e calculated according toth e Darcy equation. Theoretically, thelatte r leadst oa ninfinit e shrinkage ast — > 0,henc e infinite viscous stress.A rough calculationbase do nth eTrouto nviscosit yo fth ematri xa tp H= 6.3 mayindicat e whether this argument issound . Therelatio n

between the Trouton viscosity nE andth e loss modulus from dynamic rheological measurementsa ta characteristi c time scale G"(t* )ca nb eapproximate db y(Reiner , 1971):

Tr • G"(t' ) • t* (3.8)

TheTrouto nnumbe rT rrepresent sth erati oo fth eviscosit yi n

AH

Darcy„

-visco-elastic deformation

— elastic deformation

Fig. 3.11.Hypothetica l representation ofth ephenomen a determining theshrinkag e rate in theoute r layers ofa rennet skimmilk geli nth eInitia l stages ofsyneresis .

42 elongationan di nshear .Fo rrennet-induce d milkgel sn oexac t information on Tr is available and the assumed valueo f 3 is thusa minimum . G"i sassume dt ob edu et obreakin gan drefor ­ mingo f bondsrathe r thant oflo wo fwhe ythroug hth ematrix . For G"a tslo wdeformatio n(t *» 1/ u= 15 0s )o fa ge lwit hp H= 6.3a t3 0mi nafte rrenne taddition ,6 N«m~ 2 wastake n(Zoo ne t al.,t ob epublished) .I nth ecas econsidere d (pH= 6.33 ,3 0mi n afterrenne taddition )th eTrouto nviscosit ywoul dthu sb e270 0 Pa«s. Byusin g thevalu e forth e initialendogenou s syneresis pressure as the stress a, thedeformatio n in the outer layer equals t»o/riE =0.18 .I nth ecorrespondin gmode lcalculatio na valueo f0.5 5wa sobtaine dfo rth edeformatio ni nth emos toute r layer at 150 s after starting syneresis. So, visco-elastic effectsma ypla ya par tdurin gsevera lminute safte rth eonse t of syneresis. This affects the calculated values of r when fittingth eshrinkag et opowe rcurves .Th ediscrepanc ybetwee n modelcalculation san dexperimenta lresult so fva nDij k (1982) mayb eexplaine di nth esam emanner .

For thecalculatio n ofP |wit hvaryin g r Eq. (3.5)ca nb e generalized:

1 1, 2 d(AH/H0) t " n• H 0 ( • ?" . (3.9) dt r• Q Be

Thisequatio ncontain stw ounknowns ,Q an d P^. Fo rEq . (3.5)Q wasobtaine d frommode lcalculation s (seeSectio n 3.5).Whe nr deviatesfro m0.5 ,th enumerica lparamete rQ i slikel yt ob ein ­ fluenced.Thi sca neasil yb esee nwhe nEq . (3.9)i swritte ni n anintegrate dform :

2 n • ff0 a = ( )r • (3.10) Be - n ff„ where a represents the slope in AH = a«tr. It was tried to influenceth eresult so fmode lcalculation si nsuc ha wa ytha tr > 0.5 would result,i norde rt ochec kth eeffec to fthi so nQ 43 ando nth edifferenc ebetwee nth evalu etake n forP |an dth eon e calculated. To that end, in the numerical model the maximum valuefo rth eendogenou ssyneresi spressur ewa smad et o increase from 0.27 to0.8 0 Paafte r 1800s afte r starting syneresis (see Section 3.2.1.2). Other conditions for the calculation were as described inSectio n3.5.2 .Fo r ra valu eo f 0.57 wasobtained , whileQ wa s foundt ob e1. 0 (ifusin gth evalue s forth eshrink ­ age till 300 s after starting syneresis). For f*jaccordin g to Eq. (3.9), this resulted in values between 0.22 and 0.26 Pa (depending somewhat on the time after starting syneresis) instead of 0.27 Pa. So,despit e the deviation from thepropor ­ tionality to t°• 5 thecalculate d valueo f P%onl ydiffere d from the given value to a limited extent. Other model calculations resulted invalue s for rbetwee n 0.50 and 0.55 and alsoyielde d a 10 to 15 % difference between the given and the calculated values of Pj|.Value s forr abov e0. 6coul d notb eobtaine d with model calculations, thereby leaving part of thequestio nunans ­ wered.

Based on available information, and taking into account the limited accuracy of the syneresis measurements, it was decided to calculate P%fro m experimental results under the assumption ofa proportionalit y ofth eshrinkag et ot °• 5 between6 0an d 200 s after starting syneresis. Although these values may differ somewhat from the true values, it is expected that the varia­ tionswit hvaryin gcondition swil lno tb egreatl y influenced.

3.7 Conclusions

- Theendogenou s syneresis pressure isdetermine d by processes at the level of the paracasein micelles and the strands. It should be envisaged as a momentary result of the local behaviour of the network and the possibilities for liquid flow. Itscharacte r istherefor e clearly different from, for instance, gravitation-induced pressure or mechanical pres­ sure. - The initial endogenous syneresis pressure was found higher with a higher degree of preconoentration of the milk. This

44 Stresses the importance of the distance between reactive sitesfo rth eformatio no fne wcontacts . Themaximu m for the initialendogenou s syneresis pressure as a function of time after rennet addition depends on pH. At pH =6.3 3 themaximu mwa shighe ran d founda ta nearlie rtim e afterrenne tadditio na scompare d top H =6.68 . By using data obtained from permeability measurements with gelsfro multrafiltere dmilk ,th epermeabilit ycoefficien t as a functiono fth edegre eo fconcentratio ni sunderestimated . For theone-dimensiona l casea n "end point"o f syneresis for rennetski mmil kgel scoul dno tb edetermine d experimentally, but itwa s certainly below thevalu e assumed beforei nmode l calculations.However ,thi sdoe shardl yinfluenc eth eshrink ­ age rate in the initial stages of the process as calculated withth enumerica lmodel . Experimentally, theexponen t inth epowe r curves,whic h were used to describe the shrinkage with time after starting syneresiswa s foundt ob ehighe rtha n0. 5 andmayb et odepen d on the age of the gel. This could not be accounted for in model calculations which were based on the available rela­ tionships forth echang ei nth epressur ean dth epermeabilit y with time anddegre eo f concentration. Inth e initial stages of the shrinkage process thevisco-elasti cpropertie s of the gel matrix may be rate-determining for the shrinkage of the outerlayers ,thereb yinfluencin g thecours eo fth eshrinkag e withtime . Deviations from the proportionality of the experimental shrinkage to t°• 5 do not necessarily lead to strongly dif­ ferentvalue s for the initial endogenous syneresispressure , ascalculate d fromexperimenta lresults . Moreprecis einformatio nabou tth erheologica lpropertie san d the change in the permeability with degree of concentration wouldb eneede d fora mor eaccurat edescriptio no fsyneresis .

45 4 INFLUENCEO FVARIATIO NI NCONDITION SDURIN GRENNETIN GAN D COAGULATIONO NSYNERESI S

4.1Introductio n

Syneresiso f rennet-induced milk gels can be influenced by several factors, as is experienced during practical cheese- making.However ,quantitativ eassessmen to fth einfluenc eo fth e separate processing variables is beset with difficulties. An overview of themethod semploye d and theresult sobtaine d was recentlygive nb yWalstr ae tal . (1985).I twa sconclude dtha t besides mechanically exerted pressure (cutting, stirring and pressing)geometrica l constraints,p H and temperature areth e main variables under processing conditions.A widevariet y of experimentalprocedure sha sbee nused ,showin ga fai ragreemen t ontrend sbu tth ecaus eo fth eobserve deffect sha sno tye tbee n elucidated. In recent years new information about the physico-chemical propertieso fth ebuildin gblocks ,i.e .th eparacasei nmicelles , hasbecom eavailable .Variation si nexperimenta lcondition swer e found to have an effect on the composition and the internal structure of the micellar aggregates (Roefs, 1986;Visse r et al., 1986). This may influence the intermicellar interaction forces.Th einfluenc eo fsevera lvariable swa smainl ystudie db y determining the changes in colloidal properties,suc h as sol­ vationan dzeta-potential ,i ncombinatio nwit hth edeterminatio n of thechange si nth eseru mphase .Attentio nwa sgive nt oth e influenceof : -p H(Tarod od el aFuent e& Alais ,1975 ;Darling ,1981 ;Snoere n etal. , 1984;Heertj ee tal. , 1985;Creamer ,1985 ;Roef se t al.,1985 ;Roefs ,1986 ;Schmid t& Poll ,1986 ;va nHooydon ke t al.,1986a ;Visse re tal. ,1986) ; -temperatur e (Sood et al., 1976; Pearce, 1976; Darling & Dickson,1979 ;Darling ,1981 ;Snoere ne tal. ,1984) ;

-CaCl 2 (Dalgleish,1984 ;Snoere ne tal. ,1984 ;Creamer ,1985 ; vanHooydon ke tal. ,1986c) ; -NaC l (Ovist, 1979b;Dalgleish , 1984;Guffert y & Fox,1985 ; vanHooydon ke tal. ,1986c) . 46 / \10°C 100- "solvation" \ gH 0/g protein 2 30° C 50- \.

30° C 100- % miceUar ^^ casein 501 V"

100 / —---'— % MCP —Ca,—P 50- // 30°C

100 ^ Ç- potential L ^/^ (mV) 50-¥ 5.5 6.5 10 30 10 100 pH temperature +CaCl2 + NaCl (°C) (mM) (mM)

Flg. 4.1. The effect of pH. temperature, added CaClj and added MaCl on some properties of casein and casein micelles. Values shown are relative to those under standard

conditions: pH = 6.7, 30 'c. no CaCl2 or MaCl added. After various sources: see text.

A compilation of the observed effects is given in Figure 4.1. Primaryeffect swer econsidere d tob echange sin : - thestat eo f solubilizationo fmicella rcalciu mphosphate ; - thenegativ echarg eo fth ecasei nmolecules ; -hydrophobi cinteraction sinsid eth emicelles . As a result changes in the solubilization of the casein mol­ ecules and the binding of ions onto the micelles were found. Moreover, alsochange s inth eaverag emicella r diameter and the overall micellar charge were detected. However, it should be noted that differences were noticed between the behaviour of casein and paracasein micelles towards variations in pH (van Hooydonk, 1986a). Therefore, caution is needed in extrapolating the results obtained with casein micelles to aggregated para­ caseinmicelles . For the influence ofvariatio n inexperimenta l conditions on the renneting behaviour one should distinguish between the effect on the enzymatic and the aggregation reaction. The

47 separate influence of rennet concentration, pH, temperature,

protein concentration and CaCl2 addition (constant pH) in milk was summarized by van Hooydonk & vande n Berg (1987). Informa­ tion on the influence of NaCl addition without correction for thep H wasgive nb yva nHooydon k et al. (1986c). These results can for the greater part be explained by the effect of NaCl additiono nth epH . Treating syneresis as an extension of the aggregation proc­ ess, especially the % CMP at the gelation time (as a measure forth eremainin gsteri crepulsion )an dth eaggregatio nrat eca n be considered important parameters. The strengthening of the strandsunde rvariou scondition sca npla ya part ,a soutline d in Section 3.2. For understanding the influence of the mentioned factorso nth esyneresi sbehaviour ,additiona l informationabou t the development of the microstructure and its relation to syneresis behaviour was needed. Thiswa sobtaine d by performing permeability measurements with non-syneresed gels. By combina­ tiono fth eresult swit hinformatio nabou tth esyneresi srat ea t quiescent conditions for the one-dimensional case, the endoge­ noussyneresi spressur ecoul db ecalculated .

4.2Result san d discussion

4.2.1 Treatment ofreconstitute dmil k

Variationsi nth epreparatio nprocedur ewer efoun d tohav ea n effect on the renneting properties of reconstituted skimmilk (Ramete tal. , 1981;Zura we tal. , 1985).Usin g themil k shortly after having dissolved the milk powder resulted in much longer clotting times and delayed gelation. Addition of about 1 mM

CaCl2 or slight acidification led to results, resembling the renneting behaviour of fresh milk. An effect on the syneresis behaviour of the resulting gelswa s already noticed byva nDij k (1982). Part of the spread in some of his results can probably be ascribed to variations in the preparation procedure for the reconstitutedmilk . Some experiments were performed to establish a suitable preparation procedure for thisstudy .Result s forvariation s in 48 temperature history andi non ecas e addition of CaCl2 are shown in Table 4.1.Fo rth epermeabilit y coefficient andth e initial endogenous syneresis pressure the values at 1 hr after rennet addition are given. Before this time it was not possible to start permeability and syneresis measurements with sample C without severely damaging thegel .Furthe r information onth e change in the permeability coefficient with time after rennet addition for samples A,C,Ean dF is given in Figure 4.2. Also some results with uncooled andcoole d fresh skimmilk areinclu ­ ded. For sake of clearness the results for samples B and D, which closely resemble the results of samples A andC respec­ tively, are left out.I n all cases the average change inth e permeability coefficient over five hrs during measurement (and thus including anychang e duet oth eapplie d pressure gradient) is indicated bysoli d lines.

1013.Be(m 2) / m " /y y. y - J" y> y y'y Y y y S -- S' Yy \ 12 3 4 time after rennet addition (h)

Fig. 4.2. Permeability coefficient as a function of time after rennet addition. Influence of variation in the preparation procedure for the reconstituted milk and comparison with fresh milk. Solid lines show the average rate of change during mea­ surement. Dashed lines connect calculated values of Be. pH = 6.7; 500 ppm rennet: 30 "C; initial pressure gradient = 5 kPa-m-1. A (o),C (0).E (A).F (A),se e Table 4.1 for treatment of these samples, fresh milk (•), fresh milk after overnight storage at 4 'C(I) . 49 r^ t^ o rH en c^ vS àôô àà ó

HIOOtOOH < 00 C^ O 00 o co

Cy coe s•< *r ~m in inm « *i ne sc ^ <8 eso so oo oo oe s

•H w oo oo o o OfMBNOO [~~ <0 rH 00 C-* CO O -P a C0C0O»

00 •o H I I I • I •s y,

O O

O O oo oo a£.

o u u o 00 o 00 VD O SÄ < A m in rH O uu O o o

£ £ «tf CS CS CS -o" Oo CJCJ O o O o o o o oo 00 00 o o ££mm ££ «* CO rH CS CS •>* f-t rH

< CQ O D Cd h 50 The results were strongly influenced by the temperature-time history. Samples C and D gave higher values for the clotting time and thepermeabilit y coefficient than sampleA .Wit h these sampleswea kgel swer eobtained .Wit hsampl eC syneresisstarte d only at about 30 min after moistening the surface of the slab (=9 0 min after rennet addition), with sample D syneresis was somewhat delayed. The slow "gelation"o f sampleC canb einfer ­ red from Figure 4.2. At 1 and 2 hrs after rennet addition the solid lines showed a marked deviation from the dashed line, suggesting a relatively small resistance against deformation. The increase indynami cmodul iwit h timeafte r rennet addition, as determined with a den Otter rheometer, was also much slower for sample C (Zoo n et al., 1988) . Although the rates for the enzymatic and the aggregation reaction were not separately determined in this study, the results of van Hooydonk et al. (1986b)poin t to a possible lower reactivity of the paracasein micelles in sample C as compared to sample A. This may retard the built-up of the endogenous syneresis pressure. Moreover, stresses in the network, which are induced by the formation of new contacts, may easily relax for a considerable time after rennetaddition .So ,a nendogenou ssyneresi spressur eca nbecom e manifest only after a kind of minimum resistance is attained. The higher absolute values of thepermeabilit y coefficients for samples C and D are indicative for a somewhat coarser gel with fewerstrength-contributin g junctions.Th erat eo fchang ei nth e permeabilitycoefficien tdu et orearrangemen t ofstrand swa sno t affected. After keeping reconstituted skimmilk for 24hr s at3 0 °C,a n affect of prolonged equilibration on the experimental results could not be established (see sample B).Th e effect of cold storage (sampleF )o nth epermeabilit ycoefficien twa sabou tth e same as in the case of fresh skimmilk. The differences in absolute values could for the greater part be ascribed to a difference in casein content (2.67 w/w % in the reconstituted skimmilk and 2.57 w/w % in the fresh skimmilk). Based on these results, the procedure used for sampleA was taken as a refer­ encedurin g thisstudy .

51 Byaddin g 140pp mCaCl 2 toth emil kafte rcoolin g (sampleE) , the values for the clotting time and the endogenous syneresis pressure corresponded with thevalue s found with sampleA . This must at least be partly due to the slight drop in pH (see Section 4.2.3). The permeability coefficient was still higher than found with sample A, which may be partly related to the lowervoluminosit yo f the particles (Sood et al., 1979;Snoere n et al., 1984;va nHooydon ke t al., 1986c;se eSectio n 4.2.4). Cold storage after equilibration (sample F) resulted in a somewhat longerclottin g timea scompare d tosample sA ,B ,D an d E. Others found a longer rennet coagulation time after cold storage (Ovist, 1979; Schmutz &Puhan , 1980). Van Hooydonk et al. (1984)detecte d a lower rate for theenzymi creactio n after prolonged cold storagean d suggested thist ob epartl ydu et oa

difference in the content of C02. This can not have played a

part in this study, because C02 must have escaped during the processing of the skimmilk powder. Still, a slight increase in pH was observed. The calculated endogenous syneresis pressure wasno taffected .

During storage of milk, especially ß- and as2 -casein are subject todegradatio n by plasmin (E.C.3.4.21.7) . This results in the formation of Y-caseins and proteose-peptones (literature reviewed by Humbert & Alais, 1979; Visser, 1981; Fox, 1981; Kitchen, 1985).A smentione d before,a significan tchang ei nth e measured parameters after prolonged keeping at 30 °C (sample B) was not found. Also Pearse et al. (1986) did not detect an effect of plasmin on the syneresis behaviour of milk gels, althoughconsiderabl ebreakdow no fcasei nha doccurred .However , basedo nresult so flate rexperiment sth epossibilit yo fa small influenceca nno tb erule dou t (seeSectio n 4.2.3.1).

4.2.2 Rennet concentration

Conflicting resultshav ebee nreporte d aboutth einfluenc eo f rennet concentrationo n syneresis (Sammis, 1910;Wurster ,1934 ; Gyr, 1944; Stoll, 1966; Kammerlehner, 1974; Lelievre, 1977; Lelievre & Creamer, 1978; Marshall, 1982). However, a large influence of the rennet concentration on syneresis was never 52 detected, which may be due to themaskin g effect of mechanical treatment (Stoll, 1966). The influence of rennet concentration on syneresis should be considered in relation to the stage of the coagulation process (Kovalenko & Bocharova, 1973;Lelievre , 1977; Weber, 1984). The latter can, for instance, affect the influenceo fmechanicall yexerte dpressur eo nge lproperties . In this study some information on the influence of rennet concentrationwa sneeded ,becaus ei nexperiment swit hvariou sp H the amount of rennet was reduced at lower pH (see Section 4.3.2). Experimentswer eperforme d at pH = 6.3. Theresult s are shown in Fig. 4.3. The maximum percentage of shrinkage after five minutes was somewhat higher in case 500 ppm rennet was used. Moreover, the maximum was found at an earlier time after rennet addition, as can be seen in Figure 4.3a. The effect was mostpronounce d ifacidificatio ntoo kplac ea tconsiderabl e time before rennet addition. Thiswil l be further treated in Section 4.2.3.1. An influence of the rennet concentration on the % shrinkageafte r3 0t o5 0hr swa sno tfoun d (seeFigur e 3.8). The limited effect of the rennet concentration on the permeability coefficients can be seen in Figure 4.3b. Slower renneting apparently did not result in a coarser network. Considering the promotive effect of a higher rennet concentration on the gel firmingrat e (vanHooydon k &va nd eBerg , 1987), itappear stha t this must mainly be due to a higher number of bonds per junc­ tion. The built-up of the endogenous syneresis pressure was markedly retarded with 200 ppm rennet as compared to 500 ppm rennet (seeFi g 4.3c). Thissuggest sa contact-hinderin g ability of intact ic-caseino n notye t fully renneted micelles.Thi sma y lead to a lower rate in the formation of new contacts between existing strands and to a lower number of bonds per junction. Thereby, a somewhat lower attainability of the K-caseinfo r the rennet enzymes in the aggregated particles is conceivable. In thepermeabilit y measurements the rateo f change in thepermea ­ bility coefficient with time was found to be slightly higher with 200 ppm rennet than with 500 ppm rennet. Although this effect is hardly significant, one can think of a shift from junctions with a low number of bonds (due to remaining steric repulsion)t ojunction swit ha highe rnumbe ro fbonds . 53 % shrinkage afte r 5minute s syneresis 7-

ld3*Be(fnz 1.0 1.5 2.0 time after rennet addition (h)

2 3 U time after rennet addition (h)

1.0 1.5 2.0 time after rennet addition (h)

Fig. 4.3. Shrinkage {%) after 5 minutes syneresis (Fig 4.3a). permeability coefficient (Fig. 4.3b) and calculated initial endogenous syneresis pressure (Fig. 4.3c) for rennet skimtnilk gels as a function of time after rennet addition. Influence of acidi­ fication procedure and rennet concentration. pH = 6.34, 30 'c, H_ = 4829 um. Acidifi­ cation and rennet addition: acidification 5 min before rennet addition, 500 ppm rennet (o); acidification 20 h before rennet addition, 500 ppm rennet (•); acidification 5 min before rennet addition, 200 ppm rennet (o): acidification 20 h before rennet addition, 200 ppm rennet (•). 54 4.2.3 Acidification

From practical cheesemaking the stimulating effect ofacidi ­ fication on syneresis is well known. Also in the literature agreementexist so nth ehighe rrat eo f syneresiswit ha lowe rp H between pH = 7.0 and pH = 5.3, although considerable quantita­ tivedifference sca nb enotice d (Walstrae tal. , 1985). Acidificationo fmil k influencesth ephysico-chemica lproper ­ ties of the casein micelles, primarily due to dissociation of OCP and changes in the ionization of the amino acids (van Hooydonk et al., 1986b). An overview of some consequences was giveni nFigur e4.1 . The coagulation rate of the micelles in rennet-treated milk is determined by the rate of the enzymic reaction and the rate of the aggregation reaction. In milk the initial rate for the enzymic reactionwa s at a maximum around pH = 6.0, whereas the rate of the aggregation reaction was found to be higher for a lowerp Hbetwee np H =7. 0 andp H =5. 6 (vanHooydonk , 1986b). It should further be noted that the degree of conversion of K-casein needed to initiate aggregation was found lower with a lower pH. Between pH = 6.7 and pH = 5.6 thisvalu e was roughly halved (Pierre, 1983;va nHooydon k et al., 1986b). The combined influence of pH on the mentioned reactions caused renneting to be faster at a lower pH over the whole pH range considered. Around pH = 5.2 a transitionpoin twa s found forth e properties of thecasei nmicelle s (Heertjee t al., 1985;Reefs , 1986). The consequences for the intermicellar interactions were clearly experienced when the rheological properties of the resulting gels were examined (Roefs, 1986). Thedynami c moduli were at a minimum,wherea so nth eothe rhan d ashar pmaximu m forth evalu e of the loss tangent was observed near pH = 5.2. The latter implieseasie r relaxation of bondsbetwee no r within theparti ­ cles. At pH = 4.6 an increase of the permeability coefficient with time,i.e . microsyneresis,appeare d absent.Base d on these results, Roefs (1986) hypothized that the possibilities for syneresis to occur can be coupled to the height of the loss tangent. The very weak macroscopic syneresis, even after ren-

55 neting, of gels with pH = 4.5 (Sammis, 1910; Emmons et al., 1959)ca nserv ea ssuppor t forthi shypothesis . With experiments on the influence of pH, it was intended to draw more light upon the relation between the interparticle interactions, the development of the microstructure and the syneresisbehaviour .

4.2.3.1 Effect of the acidification procedure

The experimental procedure needed some adjustment, depending on pH. In order to avoid too fast a coagulation, the rennet concentrationwa svarie d from 500pp ma tp H =6.6 7 to10 0pp ma t pH = 5.05.Abov e pH = 6.3 thereconstitute d skimmilkwa sacidi ­ fied at 30 'Ca t 5mi nprio r torenne taddition .Belo w pH = 6.3 this was not possible without causing the formation of floc- cules. Therefore, the milk was acidified after cooling to 2° C on the day before measuring. Because of the variation in the acidification procedure, a few experiments were performed at pH =6.3 3 to establish its influence on the clotting time, the syneresis rate, the permeability coefficients and the initial endogenoussyneresi spressure . For the clotting time 205 s was found if acidification was carried out shortly before rennet addition, while 260 s was found incas e acidification took place at theda ybefor e rennet addition. Some of the results of the syneresis and permeability meas­ urements were already shown in Figure 4.3. Further results are given in Figure 4.4. In Figure 4.4a it can be seen that the initial syneresisrat ewa shigher ,i fHC lha dbee nadde d shortly beforerenne taddition .Thi swa smos tclearl y found inth eearl y stages of the gelation process. The influence of the acidifi­ cation procedure on thepermeabilit y coefficient appeared to be rather limited, as is shown in Figure 4.4b. Only with sample D markedly higher values for the permeability coefficients were obtained. This will be dealt with at the end of this section. For sample A the increase of the permeability coefficient with time was somewhat higher. As a result of the foregoing, the maximum valuesfo rth e initialendogenou s syneresis pressure 56 % shrinkage after 5minute s syneresis

1.0 1.5 2.0 time after rennet addition (h)

Po (Pa) 2 3 time after rennet addition (h )

0.5 10 15 2.0 time after rennet addition (h)

Fig. 4.4. Shrinkage {%) after 5 minutes syneresis (Fig 4.4a), permeability coef­ ficient (Fig. 4.4b) and calculated initial endogenous syneresis pressure (Fig. 4.4c) for rennet skimmilk gels as a function of time after rennet addition. Influ­

ence of acidification procedure (HCl). pH = 6.34, 30 'c, HQ = 4829 um. Acidification: A) at 30 'c 5 min before rennet addition after 23 h 30 'c (o) B) at 2 'c 20 h before rennet addition after 3 h 30 'c ([i ) C) at 30 'c 5 min before rennet addition after 3 h 30 'c (•) D) at 30 °C 20 h before rennet addition after 20 h 30 'c (A) E) at 30 °C 20 h before rennet addition after 20 h 30 "c, aprotinin added (i) 57 were found higher and tended to occur at an earlier time after rennet addition, ifHC l additionha d takenplac e shortly before rennetaddition :Figure s4.3 c and 4.4c. The micelles apparently react with some delay after the environmental conditions have changed. This may be due to the timeneede d forth echang e inth eminera l content of theparti ­ cles, the change in the state of the CCP or the change in the conformation of the proteinmolecules .Th ecalciu m ion activity changed from 0.55 mM to 0.92 mM shortly after HCl addition. After 20hr s0.9 4 mM was found. It should benote d that the low value of the calcium ion activity for standard reconstituted skimmilk must probably be ascribed to the conditions during processing and storage of the skimmilk powder used (Geerts et al., 1983). From these results it appears that the relevant changes are not directly related to a change in calcium ion activity.Change si nth estat eo fth eCC Po rth econformatio no f the protein molecules must be held responsible, awaiting the results of further investigations. Altogether, the longer clottingtim ean dth elowe rvalue sfo rth ecalculate d endogenous syneresis pressure point to a diminishing reactivity of the micelles during their adaptation to a sudden change inpH . From the available information, it can not be said to which extent theobserve deffec ti sdetermine d byth eseparat eeffect so nth e enzymatic and the aggregation reaction during renneting. The strands may possibly have a greater tendency for spontaneous breaking as a result of possible internal rearrangement processes in the building blocks during the adaptation period. This may serve as an explanation for the somewhat greater increase of thepermeabilit y coefficient with time, incas e HCl additionwa sshortl ybefor erenne taddition .

From Figure 4.4b it can be seen that higher values for the permeability coefficient were obtained when the milk was ren- neted 44 hrs after preparation: sample D. If aprotinin, a specific inhibitor for plasmin, had been added during prepara­ tion of the milk (see sample E),th evalue s were closer to the ones found with the other samples. The suspected activity of plasmin was further examined with gel electrophoresis. The 58 resultsindicate d about3 2% breakdow no f ß-caseina tth emomen t ofrenne tadditio n (resultsno tshown) .Th eamoun to fa sl-casein washardl yaffected . Incas eaprotini nha dbee nadded ,breakdow n of ß-casein was reduced to some 9 %. Apparently, the breakdown of ß-casein can have a limited influence on the structural characteristics of thege l atp H = 6.33. Atp H = 6.68 aneffec t ofth ekeepin g timea t3 0 °Co nth epermeabilit ycoefficien twa s not found (seeSectio n 4.2.1). Thisdiscrepanc y may possibly be explained by a pH dependency of the plasmin inhibition by aprotinin. Inconclusio n itca nb e stated thatplasmi nma y have a limitedeffec to ncoagulatio nan d syneresisparameters .

4.2.3.2 pH

Thestimulatin geffec to fa lowe rp Ho nsyneresi srat eca nb e clearly seen from Figure4.5 . The time after rennet addition at which the (maximum) rate occurred, is indicated. In case a limitingvalu e isgiven ,experiment sbeyon d thispoin twer e not performed. Themaximu mvalu ewa shighe ran dreache da ta earlie r time after rennet addition over the whole range between pH = 6.68 andp H= 5.35 .Th ehighe rsyneresi srat ewit ha lowe rp H is qualitatively in agreement with the results from other studies (Walstra et al., 1985). Also over longer time scales more shrinkagewa sfoun da ta lowe rp H (seeSectio n3.4) .

%shrinkag e afte r 5 minutes syneresis

10- ». 02 "s 8- *

0.3 6- 03 — •-. 05 --.,

4- *1.S x- -0.5 • 2-

0- L^T- 5.5 pH Fig. 4.5. Maximum shrinkage (%)afte r 5 minutes syneresis as a function of pH for rennet skimmilk gels. Time after rennet addition at which the maximum was found is indicated. 30 "c, H. 4829 um. Acidification: at 30 *C 5 min before rennet addition after 20 h at 30 *C, 500 ppm rennet ( ); at 2 'c 20 h before rennet addition after 3 h at 30 *C. 200 ppm rennet ( ). 59 Permeability measurements revealed a marked influence of pH on the microstructure and changes therein with time, as can be seen in Figure 4.6a. It was checked whether the high rate of increase of the permeability coefficient below pH = 6.0 was caused by a diminished adherence of the gel to the inner surface

10" » Be ( TÏ) 20 \ ,. ''"'^- ^ pH """"•• 5 05 18 \ y

V 5.35 16 j» 5 75 ji 14 il /'

12 il S 10 • !Î ' ^,« 5.97 8- .' ^v' X \' /" 633 6- _./'' ^^ïï* / „_=*»'' 6X9 4- ~~*~"**~^L—~~~*—!Z~———•• 668 ^ •-—" " 2' K'i:::'-<-~

0- 2 3 4 time after rennet addition (h)

Be W

pH 5.35 5.75 ,,-u s' • m' 5 97 6.33 -• —-• 6 kB ___—•—«-^ . •• 6.68 i^ ^-•'

1) •> —i 1— 2 3 4 time after rennet addition [h)

Fig. 4.6. Permeability coefficient of rennet skimmilk gels as a function of time after rennet addition on a linear scale (Fig. 4.6a) and on a logarithmic scale (4.6b). Influence of pH. 30 *C, internal tube diameter = 4.0 mm unless mentioned otherwise. Clotting time is indicated (x). Acidification: at 30 "c 5 min before rennet addition after 23 h 30 "c, 500 ppm rennet 50( 0 pp• m renne); t (- at 2 'C 20 h before rennet addition after 3 h 30 200 ppm rennet (- at 30 'C 20 h before rennet addition after 20 h 30 200 ppm rennet, internal tube diameter = 2.9 mm ( û ); at 2 'C 20 h before rennet addition after 3 h 30 "C, 100 ppm rennet ( • ).

60 ofth eglas stubes .Repeatin gth eexperimen ta tp H =5.35 ,usin g tubeswit h a smaller internal diameter (3.0instea do f 4.0mm) , did not affect the results obtained. It was felt that the validityo fth eresult swa swarrante dthi sway . With anexperimen t atp H - 5.05 ahig hvalu e forth epermea ­ bility Coefficient was found, but it hardly changed with time. The high values were certainly related to the occurrence of visible aggregates already before rennet was added. Still, a gelcoul db eformed .Th eabsenc eo fa chang eo fth epermeabilit y coefficient with time was found in an earlier study with acid milk gels (Roefs, 1986), which were formed by slow warming to themeasurin g temperature after acidification inth e cold. This underlines the existence of a fairly sharp transition from a rennet-type gel toa nacid-typ ege laroun d pH = 5.2 (Heertjee t al., 1985;Roefs ,1986) . Atp H =5.0 5 andp H =5.3 5 anaddition ­ al increase in thepermeabilit y coefficient with timewa s found during measuring. This indicated a low resistance against deformation for these gels, which was also found in dynamic measurements (Roefs,1986) .

As stated before, the initial permeability coefficient is determined by the spatial arrangement of the strands and the size of the building blocks. Van Hooydonk et al. (1986a) ob­ tained a maximum for the voluminosity of paracasein micelles around pH = 5.6, which was at least partly caused by the swel­ ling of theparticles .A roughestimat e of the initialvalu eo f the permeability coefficient may be obtained by extrapolating the known values to the clotting time (x in Figure 4.6 a) . Althoughth einitia lspatia larrangemen tma yals ob eaffecte db y pH, the extrapolated values of the permeability coefficient coincide with the obtained pH dependency of the voluminosity. The possible contribution of swelling to the change in the permeability coefficientca n beestimate d by using the equation ofKozény-Carma n (Scheidegger, 1960):

3 e •

61 inwhic h ei sth eporosit yan dd TS isth evolume-surfac eaverag e diameter of the particles. Van Dijk (1982) reasoned that for calculations based on geometrical models, such as the Kozény- Carman equation, a value of 0.5 should be taken for e,becaus e of the inhomogeneous character of the system. This conclusion wasals oreache db yRoef s (1986)afte ra mor eextensiv eanalysi s ofhi sresult so nth epermeabilit yo faci dcasei ngels .Assumin g

a valueo f 10u m ford vm , th ecalculate d valueo fB was roughly inagreemen twit hth eexperimentall ydetermine dvalu efo rrenne t milk gels at pH = 6.65 (van Dijk, 1982). According to van Hooydonk et al. (1986a), the drop in voluminosity between pH= 6. 7 andp H= 6. 0 isa tmos t1 0% .Unde rth eassumptio no f a corresponding increase in the value for the effective porosity

and decrease in thevalu e forth eeffectiv ed v m, thecalculate d value of B would change from 2.8-10'13 to 4.3-10"13 m2. Rea ­ soning along this line,abou t 50% o f thedifferenc ebetwee n pH = 6.7 anp H = 6.0 canpossibl yb eascribe d toth edifferenc e in the state of swelling. Also the relatively low value after extrapolation at pH = 5.35 can be partly accounted for in this way. On the other hand, differences in the initial spatial structure to some extent must be held responsible for the observed influenceo fth epH .Th eroughnes so fth e approximation with the Kozény-Carman equation does not allow for a more quantitativeaccoun t forth eseparat econtributions .

Thehighe rvalue s of thepermeabilit y coefficientar e partly responsible for the higher syneresis rate at a lower pH. As a result,th ecalculate d initialendogenou ssyneresi spressur ewa s found higher fora lowerp Hwit ha slightdi paroun d pH = 5.75. The calculated maximum values for each pH are shown in Figure 4.7. Again, the time after rennet addition when the maximum occurred, is indicated. Where no direct information on the permeabilitycoefficien twa savailable ,extrapolate d valueshav e been used for the calculation of the initial endogenous syne­ resispressure .

62 RflPo)

Fig. 4.7. Calculated maximum initial endogenous syneresis pressure as a function of pH for rennet skimmilk gels. Time after rennet addition at which the maximum was

Acidification: at 30 'c 5 min before rennet addition after 20 h at 30 "c, 500 ppm rennet (- at 2 "c 20 h before rennet addition after 3 h at 30 "c, 200 ppm rennet (-

For a hypothetical explanation of the observed effects, concerning the rate of change in the permeability coefficient with time and the values for the maximum initial endogenous syneresis pressurebetwee np H = 6.7 andp H =5.3 5 at 30 °C,th e followingphenomen a shouldb etake nint oaccount : - the rate of the enzymatic reaction, of which the initial valuei sa ta maximu m aroundp H =6.0 ; - the % of CMP removed atth eclottin g time,whic h islowe r at a lowerpH ; - the higher aggregating tendency of fully converted micelles at a lower pH, largely due to diminished electrostatic repulsion; - thehighe r valueo f the losstangen ta t a lower pH,becomin g pronounced aroundp H =5. 7 andver ymarke d atp H =5.2 ,whic h is indicative for bonds with shorter relaxation times (van Vliete tal. , tob epublished) ; - the voluminosity of the building blocks, for which is lower for lower pH between pH = 6.6 and pH = 4.6 with a small maximum around pH =5.6 .

63 The explanation for the behaviour of the initial endogenous syneresis pressure between pH = 6.7 and pH = 6.0 was already- given in Section 3.2.1.2. The higher rate of increase for the permeability coefficient with time is thought to be related to theheigh to fth einitia lvalue .Plottin gth eobtaine dvalue so n a logarithmic scale (see Figure 4.6b), it can be seen that the relative effect of pH on the increase of the permeability coefficient appeared to be rather limited in the considered region. Because of a possible higher number of bonds per junc­ tion with a lower pH, a relatively smaller increase of the permeabilitycoefficien twit htim ewa sexpecte d butno tfound . Between pH = 6.0 and pH = 5.35 the tendency to aggregate stillbecome sgreate rwit ha lowe rpH .However ,dynami cmeasure ­ ments showed the loss tangent to shift to a higher value at every frequencyexamined .Thi scoincide swit ho naverag eshorte r relaxation times for the strength-contributing bonds. A larger potential of possible bonds per junction at a lower pH may be counteracted by a higher proportion of not yet removed CMP in the early stages of gelation with a lower pH. For that, the splitting of theCM P from theparticle s inth eaggregate d state is assumed to be more difficult. The minimum in the curve for the initial endogenous syneresis pressure at pH = 5.75 is thought to be directly correlated to the higher value of the losstangent .Th echang ei nth erelaxatio nbehaviou r sufficest o override the result of the higher aggregating tendency in this region. In this respect, the higher degree of swelling of the building blocks can play a part. This may lead to a change in the contribution of the different types of bonds and thus to a shifti nth evalue s forth elos stangent .

At pH = 5.35 the stronger aggregating tendency of the para­ casein particles would again become dominant, despite the further increase in the value for the loss tangent. The higher aggregationtendency ,whic hi ssuppose dt ob eprimaril ydu et oa sharpdro po fth eelectri ccharg eo fth emicelle s inthi sregio n (Schmidt & Poll, 1986), leads to a high initial endogenous syneresis pressure at pH = 5.35. The low rate of increase for the gel strength and thus for the strength of the strands indicates increased possibilities for bending and formation of 64 new contacts. This contributes to the rapid built-up of a high endogenous syneresis pressure. Thehig h rate of increaseo f the permeability coefficient at this pH can not only be explained from its high initial value. The aggregating tendency itself must contribute significantly to the rate of increase of the permeability coefficient. This may go along with a shift from junctionswit h a high amount of remaining CMP tojunction s with a lowamoun to fremainin gCMP .

4.2.4 CaCl2

During the early stages of cheesemaking a small quantity of

CaCl2 (some 1 or 2 mM) is usually added to standardize the renneting properties of the milk to a certain extent (van

Hooydonk & van de Berg, 1987). Addition of CaCl2 leads to an increase in the calcium ion activity and a slight decrease in pH. The latter is believed to be the primary cause for the shortening of the renneting time (van Hooydonk et al., 1986c). The rate of the enzymatic reaction is not affected by added

CaCl2, if one corrects for pH. So, the higher coagulation rate must be ascribed to a higher rate of the aggregation reaction (van Hooydonk et al., 1986c). Generally, the shielding of charges and the diminished electrostatic repulsion between the paracasein micelles arehel d responsible. Thedirec t binding of Ca2* to negative sites and the increase in the amount of col­ loidal calcium phosphate were considered of primary importance forth epromotiv eeffec to nth eaggregatio nreaction . Walstra et al. (1985) have reviewed the literature with respect to the effect of added CaCl2 on syneresis. Generally, additions up to 10 mM somewhat enhanced syneresis, whereas larger additions appeared to have a retarding effect. In most cases, the drop in pH was not corrected for. This hinders the assessment of the separate influence of added Ca (van der Waarden, 1945). Straatsma & Heijnekamp (1987) concluded from experiments with a cheese processing simulator that the lower moistureconten to fa Goud acheese ,i ncas eu pt o3 m MCaCl 2 was added before renneting, could only for a small partb e ascribed toth eeffec to fpH . 65 Inthi sstud yth einfluenc eo fa limitedadditio no fCaCl 2 on the syneresis parameters was studied for two values of the pH.

Thechang ei npH ,du et oadditio no fCaCl 2,wa scorrecte d forb y adding 1M NaOH.Acidificatio n bymean so fHC lwa sperforme d at

30 °C at 5mi n prior to the addition of CaCl2. Th e results are showni nFigur e4.8 .

4.2.4.1 Time of CaCl2 addition

WhenCaCl 2 wasadde d at2 0hr sinstea do f 5mi nbefor erenne t addition tomil k atp H = 6.67, a somewhat lower initial shrink­ age rate was found (see Fig. 4.8a). Also the rearrangement of the strands apparently was inhibited to some extent, as can be seen in Figure 4.8b. For the initial endogenous syneresis pressure somewhat lower valueswer ecalculated . Ingeneral ,th e

effect of the moment of CaCl2 addition on the syneresis para­ meters considered was limited. Nevertheless, some remarks

should bemade .A t 20mi nafte r theadditio no f 4.5 mM CaCl2 at pH = 6.67 the calcium ion activity was 0.90 mM, whereas twenty hrs thereafter a value of 0.80 mM was obtained. The latter is somewhatcontrar y toth eeffec to fa dro pi np Ho nth echang ei n the calcium ion activity. Roughly speaking, Ca apparently more easily moves out of a casein micelle than goes in. Probably,

otherequilibri a areinvolve d incas eo fCaCl 2 addition. From the results of van Hooydonk et al. (1986c), which pertaint oa somewha tdifferen t situation,i tca nno tb eclearl y established to which extent the renneting and the aggregation reaction may be influenced by the mentioned procedures in our case. However, a change inth e aggregating tendency incas e 4.5

mM CaCl2 had been added shortly before rennet addition was not apparent from the results for thecours eo f the initialendoge ­ nous syneresis pressure with time after rennet addition. The lowervalu e fordBe/d tma y beexplaine d by a faster increase in the number of bonds per junction after aggregation. This may also retard the built-up of the endogenous syneresis pressure andthu scounterac ta possibl ehighe raggregatin g tendency.

66 % shrinkage after 5 minutes syneresis 7

6 /--^ D ^-°"-fc- pH 5 631

D^ 6 67

0 12 3 time after rennet addition (h )

b.

2 3 U time after rennet addition (h) Po I Pa)

2 4- a

d 2.0- X 1.6 PH 6.31 1.2-1 D ° / ^ ^\ Ü.Sn 6.67

04-

n- 7 ^ 1 2 3 time after rennet addition (h)

Fig. 4.8. Shrinkage (%) after 5 minutes syneresis (Fig 4.8a), permeability coefficient (Fig. 4.8b) and calculated initial endogenous syneresis pressure (Fig. 4.8c) for rennet skimmilk gels as a function of time after rennet addition. Influence of CaCl- addition. 500 ppm rennet unless otherwise indicated, 30 "c, HQ = 4829 um. No CaCl,

(o), 1.3 mM CaCl2 (A),4. 5 mM CaCl2 (o). pH = 6.67, CaCl2 added 5 min before rennet addition (- pH = 6.67, CaCl, added 20 h before rennet addition (- PH 6.31. HCl and CaCl- added shortly before rennet addition, 200 ppm renne t ( ).

67 4.2.4.2 Amount of CaCl2 added

Incas eCaCl 2 wasadded ,th einitia lshrinkag erat etende d to bea littl ehigher .Thi swa smos tpronounce d if4. 5 mMwa sadde d shortly before rennet addition. In that case, the maximum for the initial ratewa s also found at anearlie r timeafte r rennet addition. The permeability coefficient and its rate of change with time after rennet addition were not affected (see Fig. 4.8 b) . So, the consequences of a lower voluminosity of the

paracasein micelles due to added CaCl2, as found forexampl e by van Hooydonk et al. (1986c), were not reflected in these re­ sults.Thi sma yb e somewhatsurprising ,becaus eo fth esensitiv ­ ity of the permeability coefficient for a small change in the size of the building blocks (see Section 4.2.3). Presumably,

addition of CaCl2 also,cause s a slightly less irregular network to form. The calculated initial endogenous syneresis pressure

was found higher only when 4.5 mM CaCl2 was added at pH = 6.67 and initiallyals oa tp H =6.3 1 (seeFigur e 4.8c). Again, the results should be explained by a set of counter­ acting processes. Extra Ca enhances the aggregating tendency, but also leads to a faster increase in the rigidity of the strands, so limiting their moving ability. In the early stages after rennet addition the first factor apparently overrides the second one, leading to higher values for the initial syneresis

pressure in case CaCl2 is added. The pressure built-up is also hampered at anearlie r timedu et oth e faster stiffeningo f the strands. This results ina somewhat earlier maximum. The effect of a lower amount of split CMP at the gelation time is assumed to be weak. Van Hooydonk et al. (1986c)onl y detected a small difference in the amount of split CMP at thegelatio n timewit h

varyingCaCl 2 additions.

4.2.5 NaCl

Addition of NaCl increases the ionic strength of milk, thereby influencing existing salt equilibria. Probably, casein- bound Ca2* is exchanged with Na*, resulting in an increased concentration of Ca2* in the serum (Dalgleish & Parker, 1980; 68 Parker & Dalgleish, 1981) . A change in the amount of dissolved phosphate, due to the decreasing effect of a higher ionic strength on the activity coefficients, was not detected in earlierwor k (Gufferty& Fox ,1985 ;va nHooydon ke tal. , 1986c). However,th emethod suse dma yno thav ebee nsensitiv eenough . Added NaClretard s thecleavag e of K-casein by rennet enzymes in milk (van Hooydonk et al., 1986c) as well as in a model substrate (Visser et al., 1980). The effect on the aggregation of paracasein micelles is less clear. With small additions (up to 50 mM) the aggregation rate sometimes increased with the added amount,wherea swit h larger additions theaggregatio n was markedly retarded (Qvist, 1979b; Gufferty & Fox, 1985; van Hooydonk et al., 1986c). It remains to be answered whether the loweraggregatio nrat ewit hhighe raddition si sdu et oshieldin g of charged groups, which impairs ionic bond formation or to a higher stericrepulsio nb yprotrudin g chainso f <-casein.Als oa higher voluminosity of the micelles has been found In case of NaCl addition (Creamer, 1985;va n Hooydonk et al., 1986c) with possibly rearranging and partly dissolving ß-casein (Saito, 1973). Generally,wit hsmal laddition so fNaC lonl ya limite d effect on syneresis rate is experienced (Walstra et al., 1985). With larger additions (over 300mM )syneresi swa smarkedl y retarded. Inmos t cases, changes in the renneting timewer e not properly taken into account, which hindered the clear assessment of the effecto nsyneresis . Inou rcas eNaC lwa sadde d atth etim eo fdissolvin g themil k powder. The pH was corrected to the original value at 30 min before rennet addition by adding KOH. Results are shown in Figure 4.9. Addition of 100 mM NaCl resulted in a prolonged clotting time (1045 s instead of 650s wit hn oNaC l added), but it hardly affected the initial shrinkage rate (Fig. 4.9a), the permeability coefficient of thege l and thechang e therein with time after rennet addition (Fig. 4.9b). Under these conditions the slower renneting reaction is probably offset by a higher aggregation rate. This results in comparable conditions at the time of examination. Therefore, the slightly higher dBe/dt for 100m MNaC l added mayb e realistic, being the result of a 69 % shrinkage after 5minute s syneresis

2 3^5 time after rennet addition (h)

10">Be|m2)

2 3^5 time after rennet addition |h)

3 U 5 time after rennet addition (h )

Fig. 4.9. Shrinkage (%)afte r 5 minutes syneresis (Fig 4.9a), permeability coefficient (Fig. 4.9b) and calculated initial endogenous syneresis pressure (Fig. 4.9c) for rennet skimmilk gels as a function of time after rennet addition. Influence of NaCl addition. NaCl added at time of milk preparation, pH = 6.68, 30 *C, 500 ppra rennet,

Hn = 4829 urn.N o NaCl (o),10 0 mM NaCl (•), 300 mM NaCl (•).

70 slightly higher rearranging tendency. With 300 mM NaCl added, the clotting time was 2625 s. The lower gel firming rate pre­ cluded measurements before 1.5 hrs after rennet addition. The lowge l firmingrat ewa sals oexperience d froma nextr aincreas e in the permeability coefficient during measuring due to the applied pressuregradien t (dB/dt > dBe/dt). Theinitia l shrink­ age rate was lower and the maximum for the initial shrinkage rateoccurre d at a longer timeafte r rennetaddition .Th e lower shrinkage rate resulted partly from a lower permeability coef­ ficient,whic hca nb eascribe d toth ehighe rvoluminosit yo f the building blocks (Gufferty & Fox, 1985; Creamer, 1985; van Hooydonk et al., 1986c). Therat eo f increaseo f thepermeabil ­ ity coefficient with time after rennet addition hardly changed with various amounts of NaCl added. The lower aggregating and rearranging tendency of the building blocks may be compensated for by the lower strength of the junctions, giving rise to disruption and an increase in the permeability of the matrix already at small stresses. Rheological measurements did not reveal changes in the loss tangent with varying NaCl addition (Zoone t al.,t ob epublished) ,whic hdoe sno tindicat ea chang e in the type of bonds between the building blocks with various amounts of NaCl. This points to a lower number of bonds per junctioni ncas e30 0m MNaC lwa sadded .

4.2.6 Temperature

Highertemperature sar eknow nt ohav ea stimulatin geffec to n the syneresis of curd. Walstra et al. (1985) gave a brief overview of the literature about the effect of temperature, showing the same trend as experienced during practical cheese- making. Especially for studies on the microstructural level, one should distinguish between the effect of various renneting temperatures, which are maintained during syneresis, and the effect of a change in temperature after renneting. This is necessary in order to distinguish between the influence on the initial spatial arrangement and subsequent changes therein. For the first case, some results concerning the influence on the 71 syneresis rate, the permeability and the initial endogenous syneresis pressure were reported by van Dijk (1982). Higher valueswer eobtaine dwit ha highe rtemperature . In this study only the effect of a instant change in the temperature of thethermostattin g liquid at 20mi n after rennet additionwa sstudied .Calculations ,base do npenetratio ntheory , showed that the temperature difference between the thermostat- ting liquid and the innermost part of a gel was reduced by a factoro f 10withi nminutes .However ,th ebuildin gblock so fth e gel network were considered to react with a greater delay to a change in temperature. The adaptation may take several hours (vande nBijgaar t &Walstra , 1988). Theresult s forth emaximu m percentage of shrinkage after 5 min are shown in Figure 4.10. The time after rennet addition at which the maximum rate was reached isindicated .A nalmos tlinea rrelatio nwit h temperature was obtained, although this may well be coincidental. The maximum initial shrinkage rate was found at an earlier time after rennet addition with a higher temperature, except at 34 °C. From the results in Figures 4.11a and 4.12, it can be derived that the latter was caused by the dominating effect of dBe/dt. Themaximu m forth einitia l endogenoussyneresi spres -

% shrinkage after 5 minutes syneresis

0 20

Fig. 4.10. Maximum shrinkage (%) after 5 minutes syneresis for rennet skimmilk gels as a function of measuring temperature. pH = 6.68, 500 ppm rennet, renneting temperature 30 'c,change d to measuring temperature at 20 min after rennet addition, H. = 4829 urn. Time (h) after rennet addition at which the maximum was found is indicated.

72 surewa sfoun da t1 h rafte rrenne tadditio n (seeFig .4.12) . At 20 °Cth egel sdi d hardly showan y shrinkage.Thi s isi nagree ­ ment with the results of others (Koestler & Petermann, 1936; Gyr, 1944;Kirchmeier , 1972), although during their experiments mechanicalpressur ewa sinvolved . The permeability coefficients were found to be higher at a higher temperature (seeFig . 4.11). Theusefulnes s of anextra ­ polation to 20 min after rennet addition is questionable, because of the earlier mentioned adaptation processes of the particles after a change in temperature. However, between the extrapolated value at 30 °C and the constant (and therefore unequivocal) value at 20 °C a difference of about 20 % was calculated for thepermeabilit y coefficient.Usin g theapproxi ­ mationwit hth eKozény-Carma nequatio n (seeSectio n4.2.3.2 )a 5 % decrease in the effective porosity and a corresponding in­ crease in the value for dvs would cause a 20 % drop in the permeabilitycoefficient .Th ecalculate d changei nth evalu efo r dvs does not disagreewit h datao n theeffec to f temperature on thevoluminosit yo fth emicelle s (Walstra,1979 ;Darling , 1982). So, theeffec to fa chang ei ntemperatur eo nth epermeabilit yo f the matrix can probably be explained by its effect on the dimensions of the building blocks. This can also serve as a partialexplanatio n forth eeffec to ftemperatur eo nth eshrink ­ ageafte rtw oday s (seeSectio n 3.4), although syneresisha dno t yetcom et oa nend . FromFigur e4.11 a itca nb esee ntha tdBe/d twa s foundhighe r at a higher temperature.A t 20 °Cth e rearrangement of strands appears to be blocked. The logarithmic plot indicates that the effect on dBe/dt can not be explained by the initial level of the permeability coefficient, although the earlier mentioned delay in the adaptation of the micelles may have played a masking role. The faster increase of Be with time at a higher temperature must be caused by easier breaking of bonds and disruption of the junctions. This agrees with the higher loss tangent, which is found at a higher temperature (Zoon et al., 1988) . On the other hand, the rate of increase for the gel strengthwa s foundhighe rwit ha highe rtemperatur e inth eearl y stageso fgelation . Asth eresult so fpermeabilit y measurements 73 1013* Be (m2

2 3 4 time after rennet addition (h )

Beim2)

T(°C) 10*: 34 —•—•— —• .. —•— —• 30 •—" " •— —• 25 m"13- •—-•—*•=*= : •— 2 3 4 time after rennet addition (h)

Pig. 4.11. Permeability coefficient of rennet skimmilk gels as a function of time after rennet addition on a linear scale (Fig. 4.11a) and on a logarithmic scale (4.11b). Influence of measuring temperature. pH • 6.68, 500 ppm rennet, renneting temperature 30 "c, changed to measuring temperature at 20 minutes after rennet addi­ tion. Po (Pal

1.0- .* 1

0.8- / 5

0.6-

0.4- J» 2 ? 0.2-

0- -Ai—t— 20 25 30 35

Fig. 4.12. Calculated maximum initial endogenous syneresis pressure as a function of measuring temperature for rennet skimmilk gels. pH - 6.68, 500 ppm rennet, renneting temperature 30 C, changed to measuring temperature at 20 min after rennet addition,

HQ = 4829 urn.Tim e (h) after rennet addition at which the maximum was found is indica­ ted.

74 pointed to a faster decrease in the amount of stress-contri­ buting junctions at higher temperature, thisma y be interpreted as a relatively quick change tojunction swit h ahig hnumbe r of bonds. This would counteract the formation of new contacts already in an earlier stage of gelation. This may be held responsible for the maximum initial endogenous syneresis pres­ sureoccurrin gearlie ra ta highe rtemperature . In conclusion, it can be stated that the effect of tempera­ ture on the syneresis rate may for a considerable part be attributed toit seffec to nth epermeabilit yo fth ematrix .

4.2.7 Fat

In all the experiments skimmilk was used. Cheesemaking is usually done with whole or only partly skimmed milk. Fat glo­ bules are expected to hinder the shrinkage process and must finally result ina higher curdvolume .Va nDij k (1982)foun d a lower permeability coefficient with a higher fat content, althoughth erat eo fchang ei nth epermeabilit ycoefficien twit h timeafte rrenne tadditio nwa sno taffected .

relative remaining height

1.0- »

0.8-

0.6- 9 X 0.4- O• X O • « 0 O X • • 0.2- X X

n- 1 10 20 30 40 50 60 time after rennet addition ( h )

Fig. 4.13. Relative remaining height of rennet milk gels as a function of time after rennet addition. Influence of fat content. Petri dish method, 500 ppm rennet. 30 "c. pH = 6.37, H0 4807 um. Whole milk with 3.45 * fat (o),whol e milk after correcting for the volume of the fat fraction (•), skimmilk (x).

75 In this study a long-term shrinkage experiment was performed in which a comparison was made between gels from fresh whole milk and skimmed milk. The resultsar e shown inFigur e4.13 . As was expected, a lower shrinkage rate was found for whole milk. Inorde rt odetermin ewhethe rfa tglobule sac ta smor etha njus t fillers, the results were corrected for the volume fraction of fat. For the density of the fat 920 kg.nr3 was used while for the density of skimmilk 1030 kg«m"3 was taken (Walstra & Jen- ness, 1984). The corrected values are also shown in Figure 4.13. It appears that the lower shrinkage rate for whole milk can not solely be explained by the higher particle volume fraction.Th epermeabilit y isals oaffecte d anda neffec to nth e endogenoussyneresi spressur eca nno tb erule dout .Furthermore , the upward pressure exerted on the network by the fat globules should be noted. From the available information one can calcu­ late a maximum value of about 0.2 Pa. This iswithi n the order ofmagnitud eo fth evalue s forth eendogenou ssyneresi spressur e and should therefore be taken in consideration in the case of one-dimensional syneresisunde rquiescen tconditions .

4.3 Conclusions

-Whe n using reconstituted milk for renneting and syneresis experiments,considerabl etim ea televate d temperatures (say, several hours at 30 °C)wa s needed for equilibration of the dispersions. -Col d storage of the milk hardly influenced the one-dimen­ sional syneresis rate. Only a slightly coarser network was obtained. -Whe n studying the effect of rennet concentration on syne­ resis, the consequences of different rennet concentrations for the course of the gelation process must be taken into account.A t a lower rennetconcentratio n thebuilt-u p of the endogenoussyneresi spressur ewa smarkedl yretarded , possibly duet osteri crepulsio ncause db yno tye tremove dCMP . - In case of acidification and to a lesser extent also with

addition of CaCl2, the aggregating tendency of the micelles was enhanced if the additions had been made shortly before 76 rennet addition. Presumably, slow re-equilibration processes after changed environmental conditions (salt composition, conformation of theprotei nmolecules )rendere d themicelle s more stable, leading for example to a longer clotting time andslowe rsyneresis .Th espatia larrangemen to fth estrands , andchange stherei nwit htim e werehardl yaffected . Theobserve deffec to fa lowe rp Ho nth esyneresi srat ecoul d for a considerable part be ascribed to the higher permeab­ ilityo fth egel .Th eknow np Hdependenc yo fth evoluminosit y could serve as a partial explanation for thehighe rpermeab ­ ility between pH = 6.7 and pH = 5.75. At pH = 5.35 the increase in the permeability coefficient with time was possibly partly stress-induced, i.e. promoted by the forma­ tion of new contacts between already aggregated particles. The existence of a transition point around pH = 5.2 for the properties of the casein micelles was also reflected in the results of permeability measurements. Between pH = 6.67 and pH = 5.35 a higher initialendogenou s syneresis pressure was calculated at a lowerp Hwit h a slightdi p around pH = 5.75, which is possibly related to a small maximum in thevolumi ­ nosityo f theparacasei nmicelle saroun dp H =5.6 .

Addition of CaCl2 (0 - 4.5 mM), with correction for the change inpH ,wa s found tohav eonl y a limited effecto n the syneresis rate and the permeability of the gel. The higher aggregating tendency is probably offset by a faster increase in the rigidity of the strands, thereby only showing a limitedeffec to nth ecalculate d initialendogenou s syneresis pressure. A small addition of NaCl (100 mM) did hardly affect the considered syneresis and permeability parameters, although a somewhat longer clotting time was fond. With 300 mM NaCl addedclottin gan dsyneresi swer edelayed .Th ehighe rvolumi ­ nosity of thebuildin g blocksca n behel d partly responsible for a lower permeability of the matrix. The values of the calculated endogenous syneresis pressure point to a lower tendency for the formationo f new contacts between particles anda lowe rnumbe ro fbond spe rjunction .

77 Thepromotin geffec to fa highe rmeasurin gtemperatur eo nth e syneresisrat eo frennet-induce dmil kgel swa sconfirmed . The permeability coefficient was found higher with a higher measuring temperature (20 °C - 34 °C). This can largely be explained by the influence of temperature on the sizeo f the micelles. For the one-dimensional case in the absence of mechanical pressure syneresis did not occur below 20 °C. Highermeasurin g temperaturesresulte d ina highe rendogenou s syneresis pressure up till 30 °C. Presumably, an increased rate of strengthening of the strands and enhanced possibi­ lities for relaxation of theendogenou s pressure (lower loss tangent)cause d a levelling off forth emaximu mvalu eo f the pressureabov e3 0°C . The presence of fat globules retarded the shrinkage of rennet-induced milk gels. Thismus tb e explained by acombi ­ nation of a higher particle volume fraction and a lower permeability of the matrix. In the absence of mechanical pressure also the upward pressure exerted on the matrix by entrapped fatglobule sma ycontribut et oth eobserve deffect .

78 5 INFLUENCEO FMECHANICA L PRESSUREO NSYNERESI S

5.1 Introduction

Duringcheesemakin gmechanica lpressure ,i.e .pressur edu et o mechanical treatment, is exerted on curd or curd particlesby - cutting, stirring, draining and pressing. The importance of mechanically exerted pressure,whic hma y range from a fewP a to several kPa on the rate and the extent of syneresis is easy conceivable, considering the low value of the endogenous syne­ resis pressure (0-3 Pa) (van Dijk et al., 1979; van Dijk, 1982). The prominent role of mechanical treatment, whether exerted by stirring the curd whey mixture (Thomée t al., 1958; Lawrence, 1959; Stoll, 1966)o r by taking the curd out of the whey (Sammis et al., 1910; Gyr, 1944; Thomé et al., 1958; Lawrence, 1959; Berridge & Scurlock, 1959) was recognized but little effort has been spent in quantification of the effect. The aim to mimic more or less thecondition s in thechees e vat has frequently resulted in standardized cutting and stirring procedures during experimental studies on syneresis (Marshall, 1982; Pearsee t al., 1984), butwithou t giving extensiveatten ­ tion to the pressures involved and their contribution to syne­ resis. Several of the measuring methods used involved cutting and stirring as well as drainage of the curd whey mixture. Mechanical forces exerted are clearly different for the subse­ quent treatments. During cutting and stirring, collisions of curd particles and velocity gradients result in irregular local variations of the exerted pressure. Thereby, the rate of mois­ ture release is also indirectly affected due to an increase in the area of the surface exhibiting syneresis (Sammi s et al., 1910; Wurster, 1934;Koestie r & Petermann, 1936; Thomé et al., 1958). The increase in the syneresing surface area is also expected todepen d on the applied cutting and stirring ratean d on the rheological properties and the microstructure of the particles, henceo n experimental conditions.Durin g draining of the curd whey mixture, the rate of whey expulsion is predomi­ nantly governed by the,mor e or less constant, pressure due to the weight of the curd layer, the dimensions of the pores 79 between the curd particles and changes therein due to deforma­ tion of the particles. Also this process is governed by the exerted pressure in relation to the rheological properties of the particles and thus dependent on the composition (water content, pH) and on poorly defined parameters such as size and inhomogeneity (Walstra et al., 1985). Taking into account the foregoing considerations, it isclea r that quantitativeexperi ­ mental information about the effect of mechanical pressure in relation to the microstructure of the curd should be obtained frommode lexperiments . In this chapter the influence of mechanical pressure on the rateo fshrinkag e forth eone-dimensiona lcase ,a smeasure dwit h an extended microscope method (see Section 2.2.2), will be highlighted. First a few remarks will be made on the method itself and on the general characteristics of the results ob­ tained. Attention is given to the influence of mechanical pressure for various pH, measuring temperature and degree of preconcentration. Furthermore, the experimental procedure offered possibilities for the interpretation of the results in light of existing theories for the compression of structured dispersions.Thes ewil lb etreate dbriefl yi nSectio n 5.5.Base d on these theories, some structural characteristics of the gel canb ederive d fromth eavailabl eexperimenta lresults .

5.2 Somegenera lremark so nth eexperimenta l results

5.2.1 Courseo fth eshrinkag ewit htim eafte rloa dapplicatio n

Because some time was needed to moisten the surface of the curd slab, pressure was always applied at 60 s after starting syneresis (see Section 2.2.2). The filter plate was carefully placed on top of the gel by hand. At that time some shrinkage had already occurred in the upper layers, resulting in an inhomogeneous system at the moment pressure was applied. This firstminut eshrinkag edepende do nexperimenta lconditions ,i.e . more shrinkage with a lower pH and a higher measuring tempera­ ture (seeChapte r4) . Someresult sfo rp H =6.6 8 and3 0 "Car eshow ni nFigur e5.1 . 80 AH (

1500

1000-

500-

t" =tim e after load application

Fig. 5.1. Shrinkage of rennet skimmilk gels as a function of time after load appli­ cation, as measured with the microscope method. 500 ppm rennet, pH = 6.68. 30 'c. syneresis started at 30 min after rennet addition, mechanical pressure (Pm) applied 60 s thereafter. The pressure (Pa) applied is indicated.

Thelarg eeffec to fmechanica lpressur eo nsyneresi srate ,whic h ina qualitativ e senseha salread ybee nshow nb yva nDij ke tal . (1979)an d van Dijk (1982), was confirmed by these results.Th e effect was found relatively smaller with higher pressures. The shrinkage in the initial stages after load application (during 10t o1 5min )wa s foundt ob eproportiona l toth esquar eroo to f time after load application (/(t-60)). Generally, the deviation from the initial proportionality after starting syneresis was still small at 15 min after starting syneresis.Upwar d as well as downward deviations werenotice d (seeSectio n 5.5.4). Extra­ polation of the results to the time of load application, in order to calculate the absolute shrinkage, was considered allowable. Although the calculated slopes were somewhat lower thani nth ecas ewher eth epressur ewa sapplie d atth emomen to f starting syneresis, thecalculate d slopesca ncertainl y be used asa measure forth e influenceo f mechanical pressure with 81 slope 1 slope 2

/T6Ö

Fig. 5.2. Schematic representation of slope 1 and slope 2 as used for the description of the effect of mechanical pressure on one-dimensional shrinkage of rennet skirarailk gels, t = time after starting syneresis (s).

various experimental conditions. Nevertheless, it was tried to approximateth evalu eo fth eslop ewhe npressur ewoul dhav ebee n applied at the moment of starting syneresis (see Section5.3) . Totha tend ,superpositio nwa sapplied .Th eshrinkag edurin gth e first 60 s after starting syneresis was subtracted from the totalshrinkag e (=shrinkag edurin gth e firstminut e+ shrinkage after load application). By plotting the resulting shrinkage against /(t-60), theslop ewa srecalculate d andth eslop edu et o endogenous syneresis pressure was added; in this way the slope incas eo f load application at themomen to f starting syneresis is approximated (see Fig. 5.2).Generally , a 3 to 6 % higher valuewa sfoun da scompare d toslop e1 .

5.2.2 Contact surface

During preliminary experiments it appeared that the glass filterplate slef ta clea r "fingerprint"o nto po fth ege lafte r itsremova l atth een do fa nexperiment .A possibl einfluenc eo n theshrinkag erat ewa sexamine d byattachin g filterpape rt oth e surface of the glass filter plate before placing it on top of the gel (see Section 2.2.2.2). The calculated slopes, when the shrinkagewa splotte d against /(t-60), areliste d inTabl e5.1 . 82 Table5.1 . Influence of the glass filter plate contact surface on the initial shrinkage rate of rennet-induced skimmilk gels. 500 ppm rennet, pH = 6.68, measuring temperature 30 °C.Micro ­ scope method, syneresis started at 30 min after rennet addition, mechanical pressure applied 60s thereafter ,H 0 =482 9um .

slope1 slope1 F" without with (Pa) filterpape r filterpape r

27 33.1 (33.9-32.2) 39.1 (37.8-39.6-39.8) 62 44.3 (44.0-44.5) 50.3 (48.1-52.5)

slope1 = (AH-Aff(t=60))//(t-60)),se eals oFig . 5.2 t =tim eafte rstartin g syneresis

By using filter paper, clearly higher shrinkage rates were found. A possible explanation for the observed effect is shown in Fig. 5.3. Without filter paper the distribution of the pressure is considered to be more uneven. The average distance to a pore, through which the whey can drain off, is larger in the absenceo f filter paper.Moreover , local compression in the contact area between the pores can occur in such case. This effect is somewhat counteracted by the limited shrinkage and thus a higherpermeabilit y in the "lifted"part so f thegel .I t was decided to use filter paper to ascertain an even pressure distribution and an unhindered outflow of the whey as much as possible. Cutting the gel from the side-walls before starting the experimentdi dno taffec tth eshrinkag ebehaviour .

glass -filter- ll\ plate \ ^ / t 1 —filter paper

Fig. 5.3. Presumed streamlines of liquid flow from gel after load application. without and with filter paper present.

83 5.3 Resultswit hvariou sp Han d temperature

Since pH and temperature are important variables for the releaseo fmoistur edurin gpractica lcheesemaking , theinfluenc e of mechanical pressurewit h respect tovariou s pH and measuring temperature was studied in more detail. The pH was varied by adding HCl to the reconstituted skimmilk at 5 min prior to rennetaddition .I fth emeasurin g temperaturewa sdifferen t from the renneting temperature, it was changed 20 min after rennet addition. Average values of slope 1 and slope 2 are shown in Table 5.2. Also the shrinkage during the first minute after starting syneresis,estimate d by assuming a square root propor­ tionality also in the absence of mechanical pressure, is indi­ cated. A higher value for the slopes was found for a lower pH and a higher measuring temperature. This is in accordance with results from experiments without mechanical pressure, which were already discussed inChapte r 4 (seeals ova nDijk , 1982). For variation of pH as well as measuring temperature,varia ­ tion in rheological properties with conditions may play a part when mechanical pressure is applied. Especially the change in the concentration profile through the slabdurin g syneresis may be influenced. No direct information is available but some interesting remarks can be made after comparison with results frommode lcalculation sb yusin gEq . (3.5). Forexperiment swit h mechanical pressure the shrinkage rate, Be, P {= P° + P^) an d theviscosit y r\ar eknown .Th eagreemen twit hmode l calculations can be evaluated by determining Q. Results of calculations for Q, using slope 2 from Table 5.2, are shown in Table 5.3. Also the values of the initial permeability coefficient and the viscosityo f thewhe yar eindicated . Depending on the experimental conditions, the calculated values of Q more or less deviate from 0.55, the value which indicates a perfect fit with the experimental results. For values above 0.55 the experimental shrinkage rate was higher than the rate according to model calculations, while the oppo­ site applies to caseswher e Q is found tob e less than 0.55. A deviation of 0.1 inQ coincide swit h adifferenc eo f about 18 % inth evalu eo f thecorrespondin gslope . 84 es oo co co en vO Cv omnacoh co in o cv co co 8, rH Cv CO VO VO Tjl oiOioiCKjm 0 m . vo . cv . Lf) Cv • CO • H CO Cv in Cn• in t-i CO ^ VO Cv m Cv CO — '—' *~* ' w w (N VO ^~s ^•s S—* *^S m co *-*VO C/—\v cv es •H es vO m Cv es cô co in in vo es cv Cv in es cv in co • 1 • 1 • 1 o H vo in TJI co cv cv es cv o cv 0S m • vo • cv . Lf) Cv • co • H CO m es Cv• rH Cv CO •<* vo Cv m CV Cv

: m O vo — ^i m H eo m rH i i •^ co o en rH co es es • co • o vo Cv ^i vo vo oo vo • o -m • i • i • i • i c o •<#•<* m co en co o o es cv m •H g ** i m i m . ^i . vo • VO • vo en m oo en vo co m ^ m vo -H c cri es <*^ œ -H eo in a) Cv u c es m o & ï* en •* m •<* 00 rH Ij ro m en •<-{ i Cv cv có «tf r? " iH vo ro Cv m in cv -s* co m es vo U c • i -i • i •H , en en rs es m in vo o cv -cji eo eo P I o 00 ro m in m «H/ . m . vo • U i eo vo vo es (0 I CO m •rjt m vo p : « •<* co u Cv u-\ eo dm •—-'H VH O (0 (0 vo u (O Q) -U (Ü co en o oo •P ^P eo eo es en 3 -> (0 I I rH en vo rH C -V.H es oo m •<* vo m •H — D i • i • i -i S — U HOCOOOHm C0 rH in CV rH >* O H a 1 o 00 00 CO 00 "3 • co • ** -m • P vO (0 l-l I I o Cv •<* H CO II 0 co u -u CO VO ^ oo •<* m •H en vo VH es 00 5£ co •p < < i es es co co vo vo C Mn m oo oo •H 0 0 i i co cv oo co 00 Cv ^i VO VO 00 Cv 00 00 •<* 00 "<* 3 a) a) I • I • I • I O en vo vo co •<* vo cv es m co en -O 3 3 oo es eo ro oo • oo • ** • •** • H H i i co Q) (0 (0 o vo co m rH Cv D) > > co ^ ^ (0 en C (0 (0 es co •H -H -H o M P P S, -H -H vo ^— m O es CO C C eo O Cv m o CS Cv •H -H m O < VO (U CO en 00 co en 00 H o u X o VO •<* 00 o VO •^ oo SS •9 o. o VO• VO• VO• VOfl vO vO EH eo •0^0 0 0

85 Table5.3 .Influenc eo fvariou sp Han dmeasurin gtempera ­ tureo nth evalu eo fQ a scalculate dwit hEq . (3.5):Q = (slop e2 )• /(n/( P• Be) .Th evalue s for Be, n and P%wer e determined inseparat e experiments. Qfo rvariou s pH 1013*Be 103*n F% mechanicalpressur e(Pa ) (m2) (Pa-s) (Pa) 8 27 62 30° C 6.68 2.17 1.016 0.27 0.75 0.53 0.44 6.49 2.48 1.016 1.88 0.79 0.65 0.55 6.33 2.92 1.013 2.17 0.76 0.63 0.57 34° C 6.68 2.70 0.936 0.89 0.76 0.54 0.44 6.48 3.20 0.935 3.20 0.74 0.60 0.51 6.33 3.73 0.935 3.38 0.77 0.61 0.52

Resultso fmode lcalculation s(se eSectio n3.5.1 )hav eshow n thatdifferin gvalue sfo rQ wit hvariou sexperimenta lcondition s mustb eexplaine d bydifference si nth ecours eo fth epermea ­ bility coefficientwit htim ei nth eoute r layers.A lowvalu e forQ indicate stha tth eoute rlayer sbecom eles spermeabl ea s accordingt omode lcalculations .Thi sca nresul t froma large r decrease of the permeability coefficient with the degreeo f shrinkage than is assumed in the trial function (exponent> 3.0)o r from structural collapse of thematri x inth eoute r layersunde rth eapplie dload .Considerin gth edifferen tvalue s of Q forvariou s pressures,th elatte r certainlycontributes . The effect ismos t pronounced atp H =6.68 ,wher e indeedth e modulio fth ege la tth etim eo fpressur eapplicatio nar elowes t (vanDijk ,1982 ;Zoo ne tal. ,t ob epublished) .

5.4Gel sfro mpreconcentrate dskimmil k

Ingel s fromultrafiltere d skimmilkth einitia lporosit yi s lower. This is likely to correspond with a higher numbero f stress-carryinginterparticl elink spe runi tarea .So ,a large r resistance against structural collapseo fth ematri x undera n applied load is expected. Furthermore, some experiments were 86 performedt oenabl ea furthe revaluatio no fth eresult si nligh t ofexistin gtheorie sfo rth ecompressio no fporou smedia . Usewa smad eo fskimmilk ,whic hwa spreconcentrate d to1.0 , 0.75 and 0.6 times the original volume. Measurements were starteda t3 0mi nafte radditio no f50 0pp mrenne ta t3 0°C .Tw o valueso fth ep Hwer eexamined :p H= 6.6 8an dp H= 6.33 ;acidi ­ ficationwa sdon e5 mi nprio rt oultrafiltratio nb yadditio no f HCl.Al lmeasurement swer ecarrie dou ta t3 0°C .Th ecalculate d slopesar eliste di nTabl e5.4 ,whil eth ecalculate dvalue so fQ areshow ni nTabl e5.5 .Th eabsolut eshrinkag erat ewa sfoun dt o behighe ra tlowe rpH ,highe rpressur ean dhighe ri .Th eshrink ­ agerat ea sa functio no fi wil lb efurthe rtreate di nSectio n 6.5. Calculated values of Q were found higher with lower i. Apparently,th eoute rlayer sremai nmor epermeabl etha nassume d inmode lcalculations ,th emor es owit hlowe rpressur ean dlowe r i.I nrheologica lmeasurement shighe rvalue sfo rth emodul iwer e foundwit hlowe r1 .A sa resul to fth ehighe rnumbe ro fstress - carryinginterparticl elink spe runi tcross-sectiona larea ,th e force perlin k will be lower. This may retard structural

Table5.5 .Influenc eo fvariou sp H anddegre eo fprecon - centrationo nth evalu eo fQ a scalculate dwit h Eq. (3.5): Q =slop e2 •/ (n/( P •Be)) .Th e values for Be,n and P^ were determined in separateexperiment sa t3 0°C . Qfo rvariou s 1 101*B3 e 10*31 Pj; mechanicalpressur e( P a) (m2) (Pa-s) (Pa) 8 27 62 pH= 6.6 8 1.0 2.17 1.016 0.27 0.75 0.53 0.44 0.75 0.89 1.040 0.43 0.81 0.62 - 0.6 0.39 1.063 1.06 1.06 0.66 - pH=6.3 3 1.0 2.92 1.013 2.17 0.76 0.63 0.57 0.75 1.35 1.031 2.73 0.79 0.66 - 0.6 0.70 1.060 3.89 1.02 0.69 -

87 S S i. 1JOÄC es i-l § -H Uo N o moo CO 00. (1) Sfi-rl • i i i i i • f^ 1 1 1 1 " -H to g & .H Cv tO 1 o "" 'S 0 m . [V Ijl a-H - ra H 00 • CO •* m ^ "ri° +J **** f~ N g oo a VO >-in^ t^o^ rH (N m com es c-~ • i i i i i • i i i i i •* co O H afte r e ee d ski n ntratio n peratur e m . es p H] e ta 00 Cv ••^mt'^ •a g S C ^^ •H § ^ 1= •W J&Q fi -ri 0o0 S Q.-P • m oTjl —i-H- — C» tH ^ i-l ra (S •<* O i-H d oô có 1 ? Ö^i (N • O 00 CN CS O to en *tf i-l 00 to 8 • O • 1 • 1 • 1 .1 .1 "H « 1Ü lo a O M< O CO iH C^ co en es es es o 5 p* 0 ^ | 00 -es • in • •** >oo • s^ H V3 en o m es <-< to i ta CS (N m ^i oo •H C um . en ^- w CO -H star t degr é •at e < inet , s (U H S cr> 3 00 U c •Ö G CO CS to t^ g) o to (M CO & ï- • ,—s .^»S CO -H en to o m.— . ,—to . oo-— * ut M a 00 •H 1 CT! TH C-N rH rH efo a (U in in o ^ oo oo -H , u

syneres i •H ... | .|

50 0 pp m ] •p co S

o f p H a m • i • i • i shrinkag e heigh t 4 8 H & CT» CT> CT> CS O en u • a 0 00 CO CS -es • m in• TH* co• coo oo• (0 E 3 a H i en en oo o en •p 3. co a CO CS .H ta (0 ta m •>* es C u [> *••' """ u -Ha ) 00 a) 4-> N»^ •p -^m o n th e initia l M-l O (0 nfluenc e .tration , s metho d ^^ (0 VD o ^ eu +j $ oo co in +j w-p 00 o 3 ^> (0 CS 1 rH CJv •H* ^» '— C ^.H o^iinNOH O •>* to to es es

ressur e lilk . I i iltrafi ] roscop e safter , . i ...... 1 • 1 ai 3 ufi a H O iH es Cn in H en o O en en OH 0 oo oo es • rH . ^f • oo oo es es 4-> vu (0 H 1 .H 00 o w ^ M II Ü ta co es H •* u -u en M-i il! +J

th e m i es befor e ïnica l :e d ski r 0 s th e <^ > 00 ^^ j(qD aij te^ +3 < * en en C^ f* rH H 0 eo es es • •-! • co • es es es es eu (p « H 1 o co C0 w ^ tn > > preconc e HC l a t 5 ur e app l enc e o f measure d to o es H 00 (0 3 e m . V s A H H H e x: tn CTi •— C (0 (0 <« g V o) ö) r—V es •H -H -H c fi -H D) fi O IH -P 4-> M co l> -H -H * £l 00 co •^ W «O 00 ^—..N .. .« tn Ï O rH es < to tO m m v Q) CD i-l •H n O [^ tO II O t> to •u a a ^-0 0 •8 X rH d d X ri Ô Ö EH a oJ

5.5Applicatio no fexistin gtheorie sfo rcompressio n

Separationo fwhe y froma casei nnetwor k underth einfluenc e of mechanical pressureca nb etreate da sa compressio nprocess . The experimental setup formeasurin g one-dimensional syneresis enables evaluation of the results in the light of existing theories.

5.5.1 Terzaghimode l

A compression processma yb ebette r understood byreferenc e toth episton-sprin g analogyo fTerzagh i(e. g Terzaghi,1965 )i n the field ofsoi l mechanics. This isdepicte d inFig . 5.4. In (a) a spring is immersed ina cylinder filled with water. A frictionless piston with a closed stopcock hasbee n placedi n thecylinde ran dloade dwit ha loa dP .Th episto ncanno tdescen d andth eliqui d pressureP 1 isequa lt oth eapplie d pressure.I f thestopcoc ki sopened , waterescape san dth episto nsinks . At

stopcock stopcock /closed f opened no load _S_

load carried , P - 0 P' = P Pl= Pl Pl = 0 by water load carried P = 0 P = 0 P =P-Pl Pe=P by spring

Fig. 5.4.Th e Terzaghi model, explained by a frictionless piston ina cylinder with water, a spring, a stopcock anda load.

89 (c) the liquid pressure isP 1 and the spring carries the load P -P 1. Th e final equilibrium condition is shown in (d),wher e thesprin gi scarryin g thetota lloa dP . The major flaw in this analogy is the even liquid pressure throughout the height; one-dimensional syneresis of rennet skimmilk gels begins at the top of the gel and gradually prog­ ressesinward ,whic hresult si na nuneve npressur edistribution . Nevertheless, the analogy may serve as an illustration for a betterunderstanding . The fundamental equations for the description of the compression of porous media with time after load application according to Terzaghi, e.g. as summarized by Shirato et al. (1986) and Leclerc & Rebouillat (1985), are based on the fol­ lowingassumptions : - thepore sar ecompletel y filledwit hliquid ; -unidirectiona l liquid flowan dcompression ; - thepropertie s of the solid dono t changewit h pressure, i.e. thecoefficient so fcompressibilit y andpermeabilit y remain constant inth eaggregate . Applying this theory to rennetmil kgel spose sth eproble m of a weakan dhighl yporou sthree-dimensiona l network,whic hcertain ­ ly causes the porosity and thus the permeability to depend on the state of stress already in an early stage of the shrinkage process. The compressibility coefficient, obtained in the approach according to Terzaghi, will certainly depend on time and state of strain during syneresis of milk gels. Therefore, this approach was abandoned, although the piston-spring analogy canb euse da sa neas yt ounderstan d illustrationo fth echange s inth eliqui dpressur edurin g shrinkageo fa porou smediu m under anapplie d load.

5.5.2 Amode l forelasti cdispers e systems

Evaluation of the experimental results with a theory for compression of elastic disperse systems, developed by De Jager et al. (1963), seemed worthwhile, although visco-elastic beha­ viour is not accounted for in this theory. It was originally developed todetermin eth especifi csurfac eo f thenetwor k phase 90 basedo nth eresult so fcompressio nexperiment san dcalculatin g thepermeabilit yo fth esyste mwit hth eKozény-Carma nequation . It isconsidere d thatth eliqui d flowrat eou to fth esystem , wheni ti sunilaterall ycompressed ,i sdetermine db yth eperme ­ abilitycoefficien tan dth eelasti cmodulu so fth esoli dphase . A schematic drawing of system and the notations, that are used todescrib eit sstat eo fdeformation ,ar eshow ni nFigur e 5.5.I ti sassume dtha tth enetwor kconsist so fincompressibl e solidparticles ,whil eth epore sar efille dwit ha nincompres ­ sible Newtonian liquid. After load application, the stateo f stress isth e same inever ypoin to fa horizonta l sectionan d theDarc yequatio ni sstrictl yvalid .Th enetwor kdeformatio na s afunctio no ftim ei sdescribe db ya Lagrangia ncoordinat ea . According to De Jager et al. (1963), the behaviour of the systemdurin gshrinkag eca nb edescribe dby : a.th eequatio no fcontinuit yfo rth eliqui d

6 ôv ôa ôe

[— (e.Vi)+E._l]._-_ (5.1) 6a 6a ôx 6t

Vj= mea nvelocit yo fliqui dwit hrespec tt oth enetwor k (m-s-)1 vs =mea nvelocit yo fth esoli dparticle swit hrespec tt o thecylinde r(m^s" )1 e =porosit y b.th eequatio no fcontinuit yfo rth esoli d

6x 1- e 0 (5.2) 6a 1- e

wheree 0 isth einitia lporosity .Afte rdifferentiatio nwit h respectt ot ,resultin gin :

&ve öa 6e (1- e )- .— = (5.3) 6a ôx ôt

91 load liquid flowingou t

X x(a,0) x(a,\

network i Hit)

t = 0

Fig. 5.5.Th enotation s as used inth emode l forth ecompressio n of elastic disperse systems (deJage r et al., 1963).

AddingEq .(5.1 )an dEg .(5.3 )result sin :

(5.4) 6a

Introducingth estrai n X^(a,t) ofth esoli dmatri xyields :

6x E. -E l(a,t) - l- (5.5) 6a 1- E

Differentiation with respect to t yields a relation known fromth etheor yo felasticity :

6^ ^ 6Ç (5.6) ôa ôt

c.th eequatio no fmotio nfo rth eliqui d

B(e) 6P1 (5.7) n• e 6x

Inou rcase ,th erelatio nobtaine dwit hgel sfro mpreconcen - trated skimmilk canb euse d forB(e) .Th echang e ofth e permeability coefficientwit htim eafte rrenne tadditio ni s neglecteda sonl yth einitia l shrinkage ratewil lb eeval ­ uated.S ow ema ywrite :

B(e) = B(ç) = Be- ( 1- ç) 3 (5.8) 92 Transforming to the Lagrangian variable a and using Eq. (5.2),Eq .(5.7 )becomes :

1 1- e 0 B(e) ÔP • v1 = • (5.9) 1- e n• e 6a d.th eequatio nfo rth eequilibriu mo fforce s Referringt oth episton-sprin ganalogy ,w ema ywrite :

ÔP* ÔP1 + = 0 (5.10) ôa 6a

P1 =pressur eo nth eliqui dphas e(Pa ) P*= pressur ecarrie db yth esoli dmatri x(Pa ) e.th estress-strai nrelatio nfo rth enetwor k As a simplification, it is assumed that the solid matrix behavespurel yelastic :

P* = P(Ç) (5.11)

This means that any increase of P6 results in immediate compressionwithou ta tim elag ;th eresistanc eo fth ematri x against deformation is assumed to be not rate-determining. Thesquar eroo tproportionalit yo fth eshrinkag evs ./(t-60) , in case mechanical pressure was applied, supports this assumption. Actually, compression is a slow process duet o thelimite dpossibilitie sfo rliqui dflow .Thi si saccounte d for with the relationship between the permeability and the degreeo fconcentration . Theforegoin gequation sresul ti nth edifferentia lequatio nfo r thestrai n z,(a,t):

ôç 6 B(Ç) 1 <3P* 6C — = — [ • . -— ] (5.12) 6t 6a n i- ç dç 6a thedifferentia lequatio nfo rth epressur eo nth esoli dphase :

93 ÔP6 dP° 6 B(C) 1 ÔP* = •— [ - • ] (5.13) ôt dC ôa n 1- ç ôa

and the differential equation for the pressure on the liquid phase:

ÔP1 (H* 6 B(Ç) 1 ÔP1 = . [ . . ] (5.14) ôt dC 6a n 1- ç 6a

To simplifyth esolutio no f thesenonlinea rdifferentia lequa ­ tions with the unknown factor dP/dC, de Jager et al. (1963) linearized Eq. (5.12), thus assuming that the product of B(C)/(1-Ç)an ddP'/d Çca nb erepresente db ya constant ,whic hi s called A. Asa resul tEq .(5.12 )becomes :

ÔÇ A 62C — = — • (5.15) ôt n 6a2

Aha sth edimensio no fforc eand ,accordin gt oD eJage re tal . (1963),i tdescribe sth ebreakin gstrengt ho fth elink sbetwee n neighbouringparticles .Applicatio no fthi stheor yi srestricte d tocase swher e Ai sconstant ,i.e .compressio nshoul dno tresul t inirreversibl ebreakin go flink sbetwee nth eparticles . Forth estress-strai nrelatio nw ethe nca nwrite :

dP* (5.16) dç Be (1- ç) 2

Integrationwit hth econditio nP"( ç= 0 )= 0 result sin :

A 1 PMC) = . *(ç) with $(ç) = - 1 (5.17) Be 1- ç

Furthermore, for the initial stageso f shrinkage itwa s shown that:

94 AH = 2• V •/ ./ t (5.18)

Two compression tests with different pressures result in two valuesfo rth eslop ewhe nth eshrinkag ei splotte dagains t/t .

Inth euppe rlaye r t>1= ç *fo rP 1an dç 2 =ç *fo r P2.Whil e Ai s considered independento fth epressure ,th efollowin grelation ­ shipapplies :

P, »(CD — = (5.19)

p2 *«;>

Thevalue so f C* and ç*ca nb edetermine dgraphicall y (seeD e Jagere tal .1963 )o rnumericall y(thi sstudy )fo ra give nvalu e of Px/P2 •A ca nthe nb eobtaine dfro mEq .(5.18 )an dB efro mEq . (5.17). Whether A isconstan t canb e determined frommeasure ­ mentswit hdifferen tcombination so f Px andP 2o rb yusin gmil k with varying degree of preoonoentration. The latter can only servei fpreconcentratio ndoe sno taffec tth eaverag ethicknes s of the strands, something which is questionable for rennet skimmilkgels . Inth ecas eo fpreconcentration ,th ecoarsenes s of the network may be affected and this may have a direct influence on the average interparticle breaking strength (see Chapter3) . Thecalculate dvalue so f Beca nb ecompare dwit hth eexperi ­ mentally determined values.Thi sca nprovid einformatio nabou t the validity of the calculated characteristics, despite the neglecto fth evisco-elasti ccharacte ro frenne tskimmil kgels .

5.5.3Result so fcalculation swit hexperimenta ldat a

A and Be were calculated from the available experimental results,a soutline d inth epreviou ssection .Calculation shav e beenmad ewit hslop e2 (se eTable s5. 2an d 5.4). Px andP 2wer e obtained by summating the applied mechanical pressure and the initialendogenou ssyneresi spressure ,despit eth edifferenc ei n wastake n0.9 ,0.8 5a n0. 8 fori = 1.0 ,i = 0.7 5an d 95 i = 0.60, respectively. The results forvariou spH , temperature anddegre eo fpreconcentratio nar eliste d inTabl e5.6 . From these results it appears that the direct applicability ofth eoutline d theory islimited : - A isno tal lth esam efo rvariou spressur ecombinations ; - the strainç * for a certainpressur e dependso n the pressure combination considered, even resulting in clearly erroneous values. Thismus tb eascribe d toth evisco-elasti cpropertie so frennet - induced milk gels. Moreover, changes in the properties of the network may occur as a result of compression, for example thickening of strands.S o itmus tb e concluded that theassump ­ tions for the linearization procedure, which resulted in the parameter A, dono thol d forrenne tskimmil k gels. Nevertheless, some interesting remarks may be made on the results inTabl e 5.6. Calculated values for A range from 1.7 to 8.9-10"12 N for non-preconcentrated skimmilk, depending on experimental conditions. By a rough calculation based on the macroscopic breaking stress,Walstr a et al. (1985)arrive d at a valueo fabou t 10"11 N.Thi si sno ta tvarianc ewit hth epresen t results.A highe rvalu eo f Awa sobtaine dwit ha lowe rp Hi nal l cases. InSectio n 3.2.1.2, itwa s argued that thisca n probably be ascribed to a higher number bonds per junction. A was found lower with a lower i, the opposite of what was expected. An explanation can not begiven , although invalid assumptions must at leastb e held partly responsible. Thecalculate d valueso f A were hardly influenced by the measuring temperature. As the temperature had beenchange d only 10mi nbefor epressur eappli ­ cation,th ehighe rgelatio nrat ean dthereb yth eincreas ei nth e strength of the interparticle links, can still be offset by shrinkageo f thebuildin gblock s (Zoone tal. , 1988). The values of Be (model) agreed with the experimentally determined values within an order of magnitude. Be (model)wa s generally found lower with higher pH, lower measuring tempera­ ture and lower i. All this is qualitatively in agreement with results from direct determinations of the permeability coef­ ficient.

96 •O Ol tl)

CTI cv o co CTi CTl •«# co o in CSr H *J"o voV O co cv f in tv rH cv m vo co •<*r v !DH vOr H

cvc v o\ a\ COC O ese si no OO oo COC O rHr HC OC O CTCTl! C OC v CvC v CSC S CvC v CSC So o CSC S CSC Sr H O CS CS CO CO

coo ONc o CTlV O "^C CT»v CT ! CO CO CO CO inv o ^1C Sr HC S •^C O inc oe s ^ inc s inc o CV ^

CvTjirH'tf CTiCTi COCTi^O COC O rH O CT>V O rHcorHO com •* CO CS rH CSC O •«*t v ^fc ö

co es o «3 IT) •* !OiniO!C CTiC O rHC S CO CS in m m co VOV O VOV O VOV O IÛ IDVD m in vo d dd d o o odd d O O O O O O i i i i i i i i co en co ir> CO CTi CT! CO CO ^ CO CS rHC O CS CO CO Cv CO CTi tv Cv Cv Cv Cv CTi CO co COC v CO Cv oooo oo oooo oo oo oo

rHCOCSCTi CvO O CO in Cv Cvo oc o COCvCOCO CTiCTi OCTiOrH COc o rH O rH O CS CS rH CS rH rH CS rHC O rHC O I I I I II I I I I cocaines o m cococoo Oi n COC O ino Cv i-H Cv CO a\ co CTlC CT!O rH coc s O ^ O in CS VO CS CN CS VO CS CO cov o coV O CS VO CSV O

ino in CT!CT! C OC O CTi CTi CTtCTt C OC O CTi CTi CTi CTi CTi CTi oooo O O oooo oo oo o O 3 m UU ro a a co a i ai it) O O 4-> m g co kj> w to ino ino o O Cvv o Oo oo oo Oo O O Cvv o VO H rH O O rHr HO O in

97 Basedo nth eresults ,a compressio nmodulu sK (orbette ra negative elongation modulus) could be calculated. This was possibleb yusin gEq . (10)i nKamphui se tal .(1984) ,omittin g therelaxatio nter m

dP 1 — = (P• ( 1- Ç )+ 2 • K• ( 1- e 0))- (5.20) d£ (1- t, ) 2

Forsmal lvariation saroun d c= e 0 thefirs tter mo nth erigh t handsid eca nb eneglected .Combinatio nwit hEq .(5.16 )ultima ­ telyresult si n

(5.21) Be. ( 1- e 0)

Values of K, as obtained with Be (calculated), are also listedi nTabl e5.6 .Th eresult sfo r Kca nb ecompare dwit hth e values for the shear modulus as obtained from dynamic Theo­ logical measurements. Reserve with respect to this comparison should be exercised for several reasons. The frequency to be considered isuncertain ;deformatio n outsideth elinea rregio n may occur during syneresis measurements; the relation between the elongation and the shear modulus is uncertain; and the applicabilityo fth etheor yo fd eJage re tal .(1963 )i sques ­ tionable, as discussed above. When measured at 30 °Cwit h a frequencyo f1 rad•s~ x,th evalue so fth estorag emodul irange d from3 0N-nr 2 ati = 1 an dp H= 6.6 8 to30 0N-nr 2 ati = 0.6 andp H= 6.3 3 (Zoon,persona lcommunication) .Th eeffec to fp H on the calculated K is in line with the effect on the shear modulus.Fo rth einfluenc eo fth edegre eo fpreconcentratio nan d measuring temperature such agreement is not found; especially thelowe r Kfo ra lowe ri i shighl ysuspicious .

5.5.4Sam eremark sconcernin gvisco-elasti csystem s

In case of a visco-elastic system, such as a milk gel, breaking of strands (i.e. stress relaxation) may affect the 98 liquid outflow during compression under an applied load.Refer ­ ring to the piston-spring analogy (see Section 5.5.1), stress relaxationi nth esoli dmatri xwil lresul ti na partia l "return" of thepressur e on the liquid phase.On ema y try toaccoun t for the visco-elastic character of a system by adjustments in the outlined theory. Thiswoul d need detailed information about the relaxation spectruman dcomplicate d adjustmento f theequations . A simplified approach was introduced by Kamphuis et al. (1984), who used a Maxwell-typemodel .Theoretically , itwa s shown that the consequences of stress relaxation for the square root proportionality between the shrinkage and time depend on the applied pressure in relation to the rheological parameters of the network, i.e. the compression modulus and a single relaxa­ tiontime .Th emai nfacto ro finteres twa s found tobe :

L =P / ((1- e 0) .JO (5.22)

For L << 1 an upward deviation of the initial slope of the shrinkagevs ./ tma yb eexpected ,wherea s for L >>1 a ndownwar d deviation should be found (see Figure 5.6).Th e extent of the deviation depends on the value for the single relaxation time that is used in the Maxwel1-model . At relatively low pressures the liquid-pressure gradient will overwhelm thedecreas e of the permeability inth e outerlayers , whichresult s ina nupwar d

Fig. 5.6. Illustration of the theoretical effect of stress relaxation in the solid matrix on the course of shrinkage with time for a structured dispersion in compres­ sion. After Kamphuis et al. (1984). See text for further explanation.

99 deviation from the initial proportionality with Jt. For the opposite situation, stress relaxation results in the formation of a poorly permeable skin, resulting in a downward deviation. 2 For a rennet skimmilk gel with K = 100 N-nr and e0 =0.9 the criticalvalu efo rP ,i.e .th evalu ea twhic h L =1 ,i s1 0Pa . The results of syneresis measurements were reevaluated with respectt oth eforegoing .A novervie wo fth erelativ edeviation s from the initial proportionality at 15 min after load applica­ tion is given in Table 5.7. As stated before, deviations from the initial proportionality were small.However , network-stress relaxation may have occurred unnoticed. Measurements only start at about oneminut eafte r load application and at that timeth e change in liquid flow rate possibly has already taken place. Furthermore, in reality one has to deal with a spectrum of relaxation times, instead of a single value aswa s the case in the considered Maxwell approach. Consolidation of the upper layer may flatten the response of the shrinkage rate and can result in a new region, where approximately a square root proportionality is found. This explanation was given for the behaviour of a dispersion of glyceryltristéarate crystals in paraffinoi l (Kamphuise t al., 1984). Based on the results in Section 5.3 and Section 5.4, it was already concluded that structural collapse of the outer layers possibly influences the one-dimensional syneresis rate of rennet-induced skimmilkgel sunde ra napplie d load.Th eobserve d differences between the measured and the calculated initial shrinkage rates are qualitatively in agreement with theconse ­ quenceso fstres srelaxatio ni nth eoute rlayers .Hig hvalue so f Q (seeTabl e 5.3 and Table 5.5)ma y be partly explained by the appliedpressur ebein g stillbelo wth ecritica lvalue .

During practical cheesemaking the above mentioned phenomena areexpecte d topla ya par tdurin gcuttin gan d stirring,bu tth e exertion of pressure then is generally discontinuous. During pressingth eeffec tma yb emor eprominen ta spressur e isexerte d continuously. Recently, attention has been paid to this by Straatsma& Heijnekamp (1987). They studied theeffec t of pres­ singcondition so nth emoistur ereleas efro mcur d blocks.I twa s 100 i?£"S:

CO co co 00 t co co <* rH• C5 CS I es i t>. i ION'• * 1 'S* 1 • \ • ^ • ~v. • v> v^ vO CS . CO VO CO rH

to CU a\ co Cl) CM -H CO CN H u I . . I . co Cv in r-\ CS 3 v^ CO O v^ N • CO f-

O +J oo co r> co ^ i i "* Cv (V) v^. ^i CO v^ . -"N >-l CTirHiH 1 Cv 1 MJ>0 H WCJl CV Cv . . .\ .^ . . .\ . . CO I (OrlOO •«*C0CMOlD00C0OrHVOCMCM cö 00 O r-i CS ID rH Cv S O CS -v O v^ I . VO v^ 1 in —' ^ CO H CO iß 1 i

in o cv ID O O Cv VD r-î O O iH iH d

Cv m O O CO oo co o co 00 00 X oo \o

101 'shown that the height of the pressure had hardly any effect on themoistur e content of theresultin g cheese. The highest mois­ ture loss was even found in case no pressure was applied. The effect of a higher pressure was offset by the formation of a more condensed skin.Speakin g inphysica l terms,stres srelaxa ­ tion in the curd matrix is probably the explanation for the observedeffect .

5.6 Conclusions

-Accurat e and reproducible measurements on the influence of a constantmechanica lpressur eo nth esyneresi srat eo fskimmil k gels could be carried out for the one-dimensional case. Pressurewa sapplie dwit hhighl ypermeabl eglas s filterplate s at 60 s after starting syneresis. The shrinkage was propor­ tional to the square root of time after load application. For theexamine d regions of pH (6.3 - 6.7) andmeasurin g tempera­ ture (30 - 34 °C),th e shrinkage rate was higher with higher pressure (0- 6 2 Pa), lowerp Han dhighe rtemperature . - Theexperimentall y obtained shrinkage ratesdeviate d from the results of calculations with the model of van Dijk (1982). Obtained resultsindicate d thatth ecours eo fth eshrinkag ei n the outer layers during the initial stages of syneresis was dependent on experimental conditions. This was supported by results with preconcentrated skimmilk. Observed differences betweenmeasure d and calculated shrinkage rates werequalita ­ tively in agreement with the consequences of stress relaxa­ tion, i.e. structural collapse of thematri x under an applied load. - Experimental results were evaluated with an existing theory for the compression of an elastic disperse system, which allowed calculation of the permeability coefficient from results of two syneresis measurements with different pres­ sures. Although the applicability of the theory was limited, thecalculate dvalue sfo rth einitia lpermeabilit y coefficient were reasonably in line with experimentally obtained values. Thebreakin g strengtho f interparticlelink swa scalculate d to be2 t o9 •10" 12 N,dependin go nexperimenta lconditions . 102 6 SOMEASPECT SRELEVAN T TOPRACTICA L CHEESEMAKING

6.1 Introduction

Several processes are involved with the curdling of milk during practical cheesemaking, i.e. souring of the milk and addition of CaCl2, renneting, cutting and stirring. Variations in the mentioned steps have up till now been studied under closely controlled conditions, as partly reported in the pre­ vious chapters. This may increase the understanding of what happens during syneresis and can be helpful in explaining observed effects of varying conditions. Still,usefu lquantita ­ tive predictions about the syneresis rate under practical conditions can hardly be given. The complexity of the changes during syneresis under practical conditions hampers adequate modelling. One lacks, for instance, information about the mechanicalan dhydrodynami ceffect sdurin gcuttin gan dstirring , especially forparticle swhic hchang ei nsize ,shap ean dmechan ­ ical properties. In this chapter the latter will be considered in more detail. In addition, some attention is given to one- dimensional syneresis with intermittent pressure and decreasing pH as calculated by the numerical model of van Dijk (1982). Also, some remarks are made with respect to syneresis of gels from preconcentrated milk. A further step to evaluation of syneresis under practical conditions is made by introducing a preliminarythree-dimensiona lmodel .

6.2 Mechanicalpressur ean dhydrodynami ceffect sdurin gcuttin g and stirring

Cutting and stirring during cheesemaking impliesexertio n of mechanical pressure on curd or curd particles. This strongly promotes the release of the whey as was discussed in the pre­ viouschapter . Itals o results ina n increaseo f thesurfac e to volume ratio of the curd particles and the hindering of their sedimentation. The latter assures an even shrinkage of the particlesa smuc ha spossible .So ,wit hcuttin gan d stirring the contact area of the curd particles is strongly influenced. At 103 "thesam etime ,generatio no fcur d finesmus tb eavoide d inorde r to minimize losses with the whey at the time of separation. Cuttingan d stirringmus ttherefor eb eexhibite d undercarefull y controlledconditions . An estimation of the exerted pressures asks for a detailed analysiso fth einvolve deffects .Th econdition san dth econtri ­ bution of the various effects are constantly changing during cutting and stirring. This makes a complete description not feasible.Therefore , itwil l be tried togiv ea descriptio n for twosituations : (1)th eearl ystage so fcuttin g and (2)stirrin g atth emomen twhe nconsiderabl eshrinkag eha salread yoccurred .

6.2.1 Cutting

When moving the knives after renneting, the curd is locally deformed at first, the strongest right in front of theknives . Very soon,sa ywithi npart so f a second, thecur dwil l start to fracture. This means that the breaking strength is exceeded. Very little useful experimental information is available on the breaking strength of rennet milk gels and the influence of the time scale and the height of the applied stress on the defor­ mation at fracture. Creep measurements with rennet skimmilk gels (pH = 6.68, 30 °C, stress applied at 1 hour after rennet addition) showed that a constant shear stresso f more than 150 Pa was needed to cause irreversible breaking within 1 s (van Dijk, 1982). Over longer time scales lower stresses were suf­ ficient to cause fracture. It is to be expected that a higher stressi sneede d fora ge lwit ha highe rmodulus ,a sfo rexampl e with a higher temperature, a lower pH or at a later time after rennetaddition . A slit may progress rather easily after it has been formed, because fracture results in the release of energy. Part of the strain energy is released in the material close to the slit. This compensates at least partly the energy needed for the increase in the fracture surface (Luyten, 1988). So, the force needed forfurthe rdevelopmen to fa sli twil lgraduall ydecreas e aftera sli tha sbee nformed .

104 So, theexerte dpressur eo nth ecur d inducing immediatefrac ­ turewil lb erathe rhigh ,i na chees eva tprobabl ymor etha n20 0 Pa right in front of the knives and gradually less at larger distances from the cutting edge. Higher pressures may be involved in cutting curd strips that have already been subject tosyneresis ,becaus eo fa mor econcentrate d outerlayer . During the initial stages of cutting shear forces may also play a part. A passing knife causes a stress on the gel along­ sideit sedges .A smallliqui d layerwil lb e formed immediately. Assuming a liquid layer of 10"5 m and a cutting speed of 0.32 m«s" 1 (see Table 6.1) the velocity gradient amounts to 32000 s"1, which corresponds (n = 10"3 Pa•s ) to a momentary local shearstres so fabou t3 2Pa . In addition, deformation of the curd results in changes in the microstructure. It was shown by van Dijk (1982) that this mayresul ti na nincreas eo fth eloca lpermeabilit ycoefficient . This is an indirect promoting effect on local whey expulsion, which probably more than compensates for the effect of a pos­ sible structural collapse inth eoute r layersunde r the exerted pressure (seeChapte r 5).

6.2.2. Stirring

Mechanical stress can manifest itself in many ways during stirring of a liquid/particle mixture (Cherry & Papoutsakis, 1986). One can distinguish between bulk liquid turbulence effects, boundary layer shear forces and collisions, either between particles or between particles and solid objects. The magnitude of the various effects will be roughly estimated by referring to a standard set of modelled conditions, which bear some relevance to a curd/whey mixture in a closed horizontal cylindrical curd-making tank (de Vries & van Ginkel, 1984). Thesedat aar eliste d inTabl e 6.1.

Bulk liquid turbulence One of the purposes of stirring is to keep the particles floating. Turbulence provides the necessary vertical drag and lift forces. Forth e exertiono f mechanical stresses oncur d 105 Table6.1 . Standard seto f conditions ina curd-making tankdurin g stirringo fa curd/whe ymixture .

tank: volume 10m 3 radius 1.4m knives distance 40m m stirringrat e 2.2-1 0rp m 0.32- 1.4 6 m-s"1 max.powe rconsumptio n 3k W

curdparticles : shape equallysized , fairly smoothsphere s radius 2.5m m volumefractio n 0.5 density 1035kg»m~ 3

whey: density 1025kg-nr 3 viscosity 1mPa> s

particles, bulk liquid turbulence isconsidered . When consid­ ering thegeometr y of a curd-making tank, wall turbulence effects,whic hresul t froma viscou sdra galon gth ewall so fth e tank, arebelieve dt ob eo fmino r importancefo rth e heighto f theexerte dpressures .Velocit y fluctuationsi nth eliqui d phase will resulti nth e occurrenceo feddies .Th e kinetic energyi s passedo nfro m largert oeve r smaller ones, wherei ti sultima ­ tely dissipated ashea tb yfrictio n forces near stationaryo r moving surfaces (Hinze, 1971). Thesiz e andrat e of these energy-dissipating eddiesi sdetermine db yth eloca l dissipation peruni to fmas san dth eviscosit yo fth eliqui dphase . Beforefurthe rdiscussin gbul k liquid turbulencean d possible effects of differently sized eddies, the surface-to-surface spacingi scalculate d forth e datai nTabl e 6.1.B yassumin ga tetrahedral arrangement (seeFigur e 6.1)an dfollowin gCherr y& Papoutsakis (1986)on earrive sat :

'particle,to t "• " = (6.1) V 3 cube 12-(D//2)

where the quotient of Varticle tot and Vcube represents the

106 Fig. 6.1. Curd particle spacing in suspension. volume fraction of the particles. A volume fraction of 0.5 results in 0.14«d for the surface-to-surface spacing. So, the structureo fturbulenc ei scertainl y influenced by therelative ­ ly large volume fraction of the particles. As a result, the exerted forces due to bulk turbulencewil l probably be somewhat overestimated. Nevertheless,the yma y servea sa nindicatio nfo r themaximu m forcesinvolved . If the size of the smallest eddies is larger than the curd particles, the latter will follow the local flow pattern (see Figure 6.2a). Only small velocity gradients will occur, which resultsi nrathe rsmal lshea r forces (atmos ta fewPa )ont oth e particles. Thedensit y difference will cause a slight deviation fromth estreamlines . Whena cur dparticl e comest oth eborde r

OtoO o o

Q.

Fig. 6,2. Interaction between curd particles and eddies, a) eddies much larger than curd particles, b) multiple eddies of about the same size as a curd particle, c) eddy size about the same as interparticle spacing. 107 of a flow area,a larger velocity gradient and thus a higher shearstres swil lb eexperienced . Higher shear stresses will exist when interactions occur betweencur d particles and turbulent eddieso f roughly the same sizea sth eparticle s (Figure6.2b) .A singleedd ycanno tengul f the particle and can only act on part of the surface, possibly causing the particle to rotate and thus the exposure to larger velocity gradients. The velocity of these eddies can be esti­ mated fromKolmogorov' stheor y (seee.g . Sinnar &Church , 1960), thus assuming local isotropy. Taking forth eenerg y inputE 0.2 kW-nr3 and for theviscosit y of the curd/whey mixture a rather high value of 1•10"5 m2•s" x, one can show that foreddie s with the sizeo f a curd particle inertial forcesdominate .B y apply­

ingdimensio nanalysi s (vs ~ f(e,d))on eobtains :

1/3 1 3 ve = e •d ' (6.2)

with e as the energy density in m2•s" 3 and d as the particle diameter in m. This results in 0.1 m-s"1 for the averagevelo ­

city ve ofthos eeddies .Suc ha velocit yapplie dcontinuousl y to a point on the surface of a curd particle would cause it to rotate at about 6 revolutions per second. This appears to be rather high. It must be kept in mind that several eddies will interact simultaneously, resulting in a variable rotational movement. During stirring of a curd/whey mixture, the energy- density varies with location and is highest around theknives . As a result, higher eddy velocities may locally occur. Taking the local energy density ten times as high as the averageone , thecalculate d eddyvelocit yamount st o0.2 1 m-s"1. Whe nsuc ha n eddymeet sa knif ea tit shighes tstirrin grat e (seeTabl e6.1) , this results in a maximum shear force of some 160 Pa, when assuming a boundary layer of 10"5 m. This value will fall immediately, because thevelocit y differencewil l influence the rotationalmovement . While a particle rotates,th e locally exerted pressures will rapidlychange .Still ,th eexerte d forcesappea r insufficient to cause disruption of the intact particles. But where a crack already exists in a particle, further disruption due to turbu- 108 lence-induced shear forces cannot be ruled out. Without going into any detail, eddies significantly smaller than the curd particles will also cause shear stresses, but these will be smaller compared to those induced byeddie so f the same size as the particles (seeEq . (6.1)). An indirect effectca nb eexpec ­ ted through coupling between particle motion and large-scale eddy flows (Hinze, 1971).

Boundary layer shear forces Relatively large areas of high shear forces are expected in the boundary layers around the solid objects. The high volume fraction of the particles and the construction of the knives will certainly result in turbulent flow around them. Therefore, the effect on the exerted pressure is considered to be of the same magnitude as the turbulence-induced shear effects, which werediscusse dabove .

Collisions Collisions between particles can be caused by changes in liquid flow direction and velocity differences over small distances. The small surface-to-surface spacing between curd particles will give rise to a high collision frequency during stirring. According to Cherry and Papoutsakis (1986) the col­ lision frequencype runi tvolum e N iso fth eorde r

N ^ — (6.3) d4

where v r is the root mean square relative velocity between neighbouring particlesan d< pi sth eparticl evolum e fraction.A s

indicated before, v r will vary with local energy density and eddy size. It will become significant when the eddy size is about equal or smaller than the particle diameter. Taking for 1 v r themaximu m eddyvelocit y of 0.1 m-s" , onearrive s forN c at4.10 7 nr3-s" .1 The effect of a collision depends on the type of collision and 2 the involved collision energy (~m-v1 ). A head-on collision

109 inducesa compressio nforc ea tth epoin to fcontact .Th esurfac e is flattened andth ecompressio n force onth ewhe y gradually decreases duet o a larger contact area and because ofth e resistanceagains tdeformatio no fth eparticles .Collision smor e off-centerca nresul t ina shear force component. Inaddition , rotationca nprovid ea nextr a shear component toth ecollisio n force.Al lthi s complicatesth eanalysi so fth eeffec tan d the exerted pressuresfo rcollidin gparticles . It canreadil y be assumed that themovin g knives areo f predominant importance forcollision s with solid objects ina curd-making tank.Collision soccu rwhe na particl ei swithi nth e so-calledcollisio nwindo w (seeFigur e6.3) .Particle sfollowin g a streamline that passes within one particle radius ofth e surface will collide with it.Thi s iscalle d interception.I t must be noted that thedensit y difference between the curd particles andth ewhe y results ina small deviation ofth e particle movement from theflui d streamlines. Furthermore,th e passage ofth eparticle s between theknive s maypromot e par­ ticle-particle collisions asth estreamline s areben d during passage ofa knife andsho w a more condensed pattern between them. Considerable compression forces mayb eexerte d whentw o particles simultaneously have topas s through thega pbetwee n the knives eveni ncas eo fa napparen t off-center charactero f theircollision . Itma ysafel yb eassume dtha tth elarges tpressur ei sexerte d on a particle when ahead-o n collision withon eo fth eknive s occurs. Whenth e "escaperoute " ofa particlei sblocke db y

T Collision i- window

boundary layer

Fig. 6.3.Collisio n between curd particles anda knif e ina curd-making tank. Partic­ les inside thecollisio n window arewil l hitth eknife . no Fig. 6.4. Notation used to describe the collision of two curd particles. other particles, the maximum pressure will become as high as thebreakin g strength.Thi sma y amount to severalhundred so f Pa, becauseo f amor econcentrate d outer layera scompare d to thesituatio ndurin gcutting . Itwil lb etrie dt oestimat eth eexerte deffectiv epressur e witha head-o ncollisio no ftw oparticle s(se eFigur e6.4) .Th e energyneede dfo rsqueezin gou tth eliqui dbetwee nth eparticle s will not be taken intoconsideration .Whe n two particles col­ lide, their contact area changes as a result of compression. Thecontac tare aZ ca nb eestimate db y(se eFigur e6.4) :

(r2 - h2 ) (6.4)

Assumingelasti cdeformatio non eca nwrite :

(r-A) F =Z • e . K = Z ' • K = n • K • (r-h)2 (6.5)

where F is the total exerted force (N), e is the relative deformation (r-h)/r and K is theelasti cmodulu s (N-nr2). The work W neededt oobtai nth eultimat edeformatio n h*is :

(r-h*)

2 Wn = J Ti• (r-h) •d(r-h )

111 1 = - • it • K • r3 • (r - h* )3 3

1 = - • it • K • r3 • e3 (6.6) 3

The available kinetic energy Wa per particle amounts to:

1 4 W = - 'AP • - • rt • r3 • (Av)2 (6.7) 4 3

AsW n =W a thefollowin g relationship fore i sobtained :

AP • (Av)2 e3 = (6.8) K

At anytim e during compression theeffectivel y exerted initial pressurei sgive nby :

F Z h F" = -. = K -( 1 )2 (6.9) Z rt• r 2 r

Forth emaximu m initial syneresis pressure,th epressur ea tth e firstmomen to fcontac t (h=0), oneca nwrite :

PS.n.ax = AP2/3 •Av^ 3 •K 1'3 (6.10)

Deformationdu et ocollisio noccur sove ra shor ttim escale .Th e outer layer is concentrated already to some extent. Moreover, during deformation the resistance against deformation will increase. Based on results of creep measurements byva n Dijk (1982) , and taking into account that compression rather than shear occurs, the effective K is assumed to be 104 Pa.Th e maximum initial pressureo nth eliqui d phasei na particl e then amounts to4. 6P afo rA v= 0. 1m-s" 1, while forA v= 1. 0m-s" 1 a valueo f10 0P ai sobtained . Thedeformatio n andth eexerte d

112 maximum initial pressure thus are strongly influenced by the prevailingvelocit ydifference .

6.3 Intermittent pressure

As was mentioned before, large pressure variations occur during cutting and stirring. This may result in a momentary increase of the local syneresis rate or in rupture of the particlesi fth eexerte dpressure sexcee d thebreakin gstrength . After a short while of cutting and stirring the curd particles varywidel yi nsiz ean d form.Takin ga singl ecur dparticle ,th e cornerarea swil lhav ebee nsubjec tt oth egreates tvariatio n in external forces. This results in a rounding of the particles withloca lvariation si nth eunevennes so fth eoute rlayers . Despiteit slimitations ,th eintegrate d formo f Eq. (3.5)ca n be used for calculation of the shrinkage with varying pressure combinations:

Be r AH = Q •/ ( — • J P(t) •dt ) (6.11) 1 t =0

Eq. (6.11) implies that for a given shrinkage and pressure combinationth ecorrespondin g constantpressure ,whic hi sneede d to obtain the same shrinkage rate, can easily be determined by calculating the time averaged E(P«t). An example is shown in Figure 6.5 for a combination of 1 and 9 Pa, which were alter­ nated every 40 seconds. The change of the pressure was assumed tooccu r instantaneously. For thechang eo f the pressure during shrinkage trial function 3 from Figure 3.4 was taken,whil e for the change in the permeability coefficient with the degree of concentration the equation for pH = 6.68 and 30 °C from Table 3.1 was used. Taking the shrinkage after every 40 seconds, the shrinkage rate was equal to the shrinkage rate in case a con­ stant pressure of 5 Pa was applied, whether the pressure was alternated every 5,10 ,2 0o r4 0seconds .

113 AH(pm )

500- J S uoo- „/ 300 - ,/ 200- 7 100-

oJ 1 j 1 1 10 15 20 25 30 vT (s0-5)

Fig. 6.5. Calculated shrinkage of a rennet curd slab with time after starting syne- resis. Influence of intermittent pressure (1 Pa - 9 Pa, alternating every 40 s) as

calculated with the numerical model of van Dijk (1982). P = P(PQ.i) (trial function 3 from Figure 3.4.), B = B(i,t) (first equation from Table 3.1).P ^ = 0. Shrinkage with constant pressure of 5 Pa is given for comparison ( ).

6.4 Modelcalculation swit hchangin gp H

A lower pH results in a more permeable casein matrix and a higher endogenous syneresis pressure between pH = 6.7 and pH =6. 0 (se e Chapter 5) . During cheesemaking the pH gradually decreases fromth etim eo faci do rstarte raddition ,a situation which is different from that in the performed experiments. A decreasing pH may result in a continuous change in rheological properties, rearrangement of strandsan d changes inth e primary casein particles, and thus in constantly changing syneresis properties of the gel. In this section some calculational results aregive n of a first attempt to quantify the effect of anongoin gdecreas ei np Ho nth eone-dimensiona l syneresisrate . Only simple linear relationships wereused , whichwer ebase d on available experimental results.Mor e detailed information about

114 therelationship s ina dynami csituatio ni slacking .Th enumeri ­ calmode lo fva nDij k (1982)wa suse d forsom ecalculation swit h andwithou tmechanica lpressure .

6.4.1 Resultso fcalculation swithou tmechanica l pressure

The assumed relationships for the change in pH, the endoge­ nous syneresis pressure and thepermeabilit y coefficient of the gel are listed inTabl e 6.2 and areals opartl y shown in Figure 6.6. The maximum Pjj and the time after rennet addition, when thismaximu mwoul db eobtaine d (tamax)fo rth emomentar ypH ,ar e recalculated with Eqs. (2) and (3) in every time step. The momentary value for Ps(i) is thenobtaine d from Eqs. (4)- (6) . The fictitious permeability coefficient at the time of rennet addition, (timewhe nth ep Hstarte ddecreasing )wa sestimate d by extrapolation. The change in the permeability coefficient with time and degree of concentration is accounted for in Eqs. (8) and (9). Gravitational-induced pressure was neglected. The shrinkage and the concentration profiles for a decreasing pH werecompare dwit hthos eobtaine d forconstan tpH :p H =6.6 8 and pH = 6.34, the latter being the pH when AH = 500 urn in an acidifyinggel .

Table6.2 .Th e relations for thechang e in pH, endogenous syneresis pressure and permeability, as used for the calculation of the one-dimensional syneresisrat eo fa rennet-induce d skimmilkge l withvaryin gp Hb yusin gth enumerica lmode lo f vanDij k (1982).

5 (1) pH = 6.68 - 8-10" • ta (2) Fornax = 29.6 - 4.30 • pH (3) tamax = 7634 - 2887 • P^max (4) Pj; = ta/tsmax ' P^max (ta < tamax) a 4 (5) PI = 2 .64 - 3.46-10" • ta (t > t max) (6) P(i) = P^ - (i - 0.3)/(0.7 - 1)

13 (7) Be(ta=0) = 1.78-10- (8) dBe/dt = 33.5-10"17 - 4.75-10"17 • pH (9) B(i) = Be • i3

115 Po (Pa)

J.U

pH=6.34

2.0- /^\ / \. pH=6.49

1.0- 5 / / P0 max pH=6.68-^iv / / ^ \ >^ / / ^^ tQ max pH=6.68^v

o-\l^ \ 10 n-3 10~J*ta(s) to =tim e after rennet addition

Fig. 6.6. The change in the initial endogenous syneresis pressure with decreasing pH as used in model calculations with decreasing pH. t = time after rennet addition. p ,az = O maximum PQ for a given pH. ta max - time after rennet addition when P^m is obtained.

AH l^im) 500

a.

Fig. 6.7. Calculated shrinkage (Fig. 6.7a) and shrinkage profile after 10 * shrinkage (Fig. 6.7b) of rennet skimmilk gels at constant pH (pH = 6.66 and pH = 6.34) and for decreasing pH (pH - 6.68 -> pH - 6.35) in the absence of mechanical pressure. Rela­ tionships for the change in endogenous syneresis pressure and permeability are listed in Table 6.2. H_ = 5 mm. P? = 0 Pa. 116 The calculated shrinkage rates and concentration profiles at 10 % shrinkage are shown in Figure 6.7. The pH-history is seen to have a marked effect on the rate of shrinkage. This was lowest for constant pH = 6.68 and highest for constant pH = 6.34. At the time of starting syneresis, the pH in the acidi­ fying gel was 6.54, which causes the initial rate to be higher than at pH = 6.68. The slight undulation in the curve for the acidifying gel is caused by the still increasing P^niax in the first stage of the shrinkage process and the decrease there­ after. The shrinkage profile at a relative remaining volume of 0.9 is given in Figure 6.7b. An effect of pH-history on the shrinkage profilecoul dno tb eestablished .

6.4.2 Resultso fcalculation swit hmechanica l pressure

Results of calculations in which the endogenous syneresis pressure was replaced by a mechanical pressure of 27 Pa are shown in Figure6.8 . From the relations for the syneresis

AH( Hm) i i it t II 1.0 pH= 6.68— / /LpH=6.48 800- / 0.8 / / /LpH=668 600- /' 1// / 0.6 i II t II i II I II 400- i i IIII 0.4 i f i i 200- if if 0.2

0- 1 1 ! 0 10 20 30 40 \/T (s05)

Fig. 6.8. Calculated shrinkage (Fig. 6.8a) and shrinkage profile after 20 % shrinkage (Fig. 6.8b) of rennet skimmilk gels at constant pH (pH = 6.68 and pH = 6.48) and for decreasing pH (pH = 6.68 ->-p H = 6.48). with a mechanical pressure of 27 Pa. Relation­ ships for the change in endogenous syneresis pressure and permeability are listed in Table 6.2. The contribution of mechanical pressure with varying degree of concen­

tration was calculated with trial function 3 from Figure 3.4. HQ = 5 mm, P^ = 0 Pa, B „ i3 ( ), B~ i2 ( ) 117 pressure in Table 6.2 only Eq. (6)wa s used. Thecours eo f the shrinkagewit htim efo ra nacidifyin gge lcoincide swit hth eon e for constant pH = 6.48 (pH whenA H = 1000 um in an acidifying gel). Moreover, thedifferenc ewit h thecurv e forconstan t pH= 6.68 appears tob e rather small.Th e strongpromotin g effect of the mechanical pressure masks the effect of a decreasing pH on the rate of shrinkage. Based on experimental results, Stoll (1966)arrive d atth esam econclusion . The concentration profile with a relative remaining volume of 0.8 washardl yaffecte db ypH-history .Whe na differen t relation for thepermeabilit y coefficient asa functiono f thedegre eo f concentration was used (B =Be«! 2), a higher syneresis ratean d a slightlydifferen tconcentratio nprofil ewer e found.Probably , muchmor ecomplicate d relationships shouldb euse d forpicturin g the changes during syneresis.Especiall y thedevelopmen t of the concentration profile with varying conditions needsmor eatten ­ tion, as thepermeabilit y of theoute r layerst oa considerable extentdetermine sth ecours eo f furthershrinkag ewit htime .

6.5 Milkpreconcentratio n and syneresis

Forsevera l typeso fcheese ,manufactur e from preconcentrated milk, as achieved by ultrafiltration, has been found possible without causing quality defects in the resulting cheese. Ren- neting of preconcentrated milk leads to a faster increase of the dynamic moduli after the onset of aggregation, themor e so for a larger degree of preconcentration. The higher casein concentration is held responsible for this (Culioli & Sherman, 1978; Zoon et al., 1988). Preconcentration results in a denser casein network with a higher amount of stress-carrying strands percross-sectiona l area (vanDijk ,1982 ;Gamot , 1986;se eals o Chapter 3). The consequences for the endogenous syneresis pressure were already discussed in Section 3.2 and for the permeability in Section 3.3. When considering the influence on syneresis rate, differences in solids concentration should be taken into account. According to Green (1987), it appeared probable that the initial rateo f syneresis is relatively lower with a larger degree of preconcentration, even when correcting 118 for the actual amount of aqueous phase in the curd. By fitting experimental results to an exponential equation representing first-order kinetics. Peri et al. (1985) did not detect an effecto fth edegre eo fpreconcentratio no nth erelativ eshrink ­ agerate . During thisstud yone-dimensiona l syneresismeasurement shav e been performed with preconcentrated milk (pH = 6.68 and pH = 6.33), which have been partly discussed in Chapters 3 and 5. Experimental results forth etota lshrinkag ea t3 0minute safte r starting syneresis are shown in Figure 6.9. For the determina­ tion of the relative shrinkage,th evolum eo fth eaqueou sphas e ati =1. 0wa sestimate d bytakin g2. 7ml/ g forth evoluminosit y of the paracasein fraction and neglecting the contribution of other constituents (Green, 1987). The relative shrinkage for other i was obtained by dividing the absolute shrinkage by (i-0.07/1). As can be seen in Fig. 6.9, the absolute shrinkage was less for further preconcentration (lower i) for both pH values andwhethe rmechanica l pressurewa s applied ornot . This

% shrinkage after 30minute s syneresis 30-

Pm= 27 Pa 60-

A

40- o 2^^Co ^* .»• >r ^^^"^

••"•

20-

n r A 0.6 0.7 0.8 0.9 1.0

Fig. 6.9. Shrinkage {%) of rennet skimmilk gels after 30 min of syneresis as a func­ tion of the degree of concentration. Shown are results for the absolute ( ) and relative ( ) shrinkage. Mechanical pressure none and 27 Pa (note the difference in scale). 500 ppm rennet, 30 *C, syneresis started at 30 min after rennet addition, mechanical pressure applied 60 s thereafter. pH = 6.68 (•), pH = 6.33 (*).

119 canb e explained by thedominatin g effecto f a lowerpermeabil ­ ity in case of a higher casein concentration. The relative shrinkage after 30 minutes of syneresis was found higher for lower i, the effect being somewhat more pronounced in the absence of mechanical pressure. The effect on the relative shrinkage can be ascribed to a higher endogenous syneresis pressurea ta highe rcasei nconcentratio n (seeSectio n 3.2.1.3). Thiswa smaske d incas emechanica lpressur ewa sapplied . In our experiments, syneresis was started at a fixed time, i.e. at3 0minute safte rrenne taddition . Ifth efirmnes so fth e gel is determent for the time of starting syneresis,th eperme ­ ability (andprobabl yals oth eendogenou ssyneresi spressure )o f the gels from preconcentrated milk will be slightly lower as compared to the present situation. The differences in the relative shrinkage rate after the onset of syneresis will then become smaller. This can not explain the discrepancy with the results of Peri et al. (1985) and Green (1987), because in their experiments syneresis was also started at a fixed time after rennet addition. However, the use of whole milk may have resulted in a relatively stronger drop of the permeability coefficient with degree of concentration as compared to the present study. Also effectso f differing methods of determining syneresisca nno tb erule dout . The results given above pertain to gels where the initial i is the sameeverywher e in thematrix .A s a resulto fsyneresis , a gel with an equal value for the average i may be obtained. However, the gel which has been subject to syneresis then will exhibit a lower shrinkage rate,becaus e of theexistin g shrink­ age profile. This implies a lower permeability of the outer layers. Inaddition ,th eendogenou ssyneresi spressur ewil lhav e partlyrelaxe db ythe ni nsuc ha gel . In conclusion, it is to be expected that during practical cheesemaking processing conditions must be adapted in order to avoid excessive moisture loss during cutting and stirring of curd frompreconcentrate d milk.

120 6.6Preliminar y calculationsfo rthree-dimensiona l syneresis

Up till now only one-dimensional syneresis has been con­ sidered. Calculations forth e three-dimensional case aremor e complicated asthe y should, forexample , account foranisotrop y of the shrinking matrix. Collection of experimental results under controlled conditions that canb euse d forcompariso ni s hardly feasible in our opinion. Nevertheless, the one-dimen­ sional numerical model ofva nDij k (1982)ha sbee n adapted to makesom eroug hcalculation sfo ra sphere . For the model a sphere is considered to be divided into spherical layers,whereb yth eorigina l thicknesso fth elayer s

Fig. 6.10. Notation used for the numerical description of the shrinkage of a sphere. varieswit hposition ,ever ysubsequen t layer fromth eoutsid et o the inside being 1.2 times as thick asth e former one.Whe n numbering the layers and using the notations as pictured in Figure6.1 0w eca nwrit efo rth eloca ldegre eo fconcentration :

Vt = <6-12) (r3 -r 3 )

For theflu x density from kt ok+ 1 oneca nwrit e analogoust o Eq. (3.3):

121 str p *.t = (Pk.t - K.i.t) •( 2 /n) r -r r -r 1 ( k.t k-l.t + k• 1 , t k.t r (6.13) Bk. t Bk • i. t

3 -1 Theliqui d flux Vstrk t fromk t ok+ 1i nm »s equals:

2 strk t .4 . n •r k, (6.14)

whereby rk, represents theradiu s halfway thedistanc e between

rk andr kt: . Thevolum e changeo f spherical layer ki srepre ­ sentedby :

A^k.t = (Vstrk-i.t" Vstrk.t)* A t (6.15)

Forever y time stepth espher ewa s"rebuilt "b ycalculatin gth e radiuso fth einnermos tlaye rfro mth eremainin gvolume ,joinin g the volumes of the subsequent layers and calculation ofth e

radii rt tto r t.Th ereliabilit yo fth eapproac hwa schecke d by varying the number of layers, the time interval andth e spacingo fth elayer swithi nlimits .A swit hth eone-dimensiona l numerical model (vanDij k et al., 1984), only small deviations werefound . Calculationswit hthi smode lhav ebee nperforme di nwhic hth e radiusan dth einitia l pressure were varied. Forth echang eo f the pressure with degree of concentration trial function 3i n Figure3. 4wa sused .Th efirs tequatio ni nTabl e3. 1wa sintro ­ duced to describe the change of the permeability coefficient with timean dth edegre eo fconcentration .N oattemp tha sbee n made to account foranisotropy , as relevant information about theexten to fthi seffec twa sno tavailable .However ,i tappear s that theshrinkag e rateo fa syneresing sphereca nb eapproxi ­ mated by this approach, ifw e restrict ourselves toth eearl y stageso fsyneresis . The calculated values forth erelativ e shrinkage between 0 and 300second s were fitted topowe r curves. Thecoefficient s werecompare dt othos eobtaine dfo rth eone-dimensiona lcase .

122 Table6.3 .Calculate d leastsquar eestimate so f a/V0 and r r inAV /V 0 =( a/V0 ) - t forth erat eo fshrink ­ age of rennet skimmilk gels till 5 minutes after starting syneresis.Compariso n of calcu­ lations for a slab and a sphere with varying initialheigh to rradiu san dpressure .Se etex t forfurthe rexplanation .

height/ sphere slab pressure radius 103* r 103* r (Pa) (mm) a/V0 a/V0 1 1.25 21.4 0.43 5.8 0.51 2.5 11.1 0.44 2.9 0.51 5.0 3.1 0.55 1.5 0.51 10.0 1.5 0.56 0.7 0.51

8 1.25 55.8 0.37 17.7 0.50 2.5 29.7 0.42 8.9 0.50 5.0 15.2 0.44 4.4 0.50 10.0 8.0 0.45 2.2 0.50 27 5.0 27.6 0.42 8.1 0.50 62 5.0 41.6 0.39 12.3 0.50

V0 :initia lvolum eo fth espher eo rheigh to fth esla b

The results are listed in Table 6.3. In all cases a/V0 for a spherei shighe rtha nfo ra slab ,du et oa highe r surface/volume ratio.Thi shighe rrati oals ocause sth erelativ eshrinkag erat e to be higher with a smaller height or radius. However, after some time the shrinkage of a sphere will be hindered more and more due to extra condensation of the matrix phase as a result of "inward shrinking". Also less whey is available in deeper layersa scompare d toth eone-dimensiona l case.A highe r initial pressure or a lower initial radius will result in an earlier expression of this effect.A s a result, for a sphere theexpo ­ nent r (seeTabl e 6.3) isdependen t on the initial pressure and the initial radius/height. This can also be seen from Figure 6.11, in which the calculated relative shrinkage after 30 minutes of syneresis for some combinations of the initial pressurean dth einitia lradiu si sshow nfo rbot hcases .Fo rth e relative shrinkage of a sphere a relatively smaller increase witha lowe rradiu sand/o ra highe rpressur ei sobtained .

123 % shrinkage after 30minute s syneresis

0 2 4 6

R0,H0(mm]

Fig. 6.11. Calculated shrinkage (%)afte r 30 minutes syneresis for a slab and a sphere for various pressure. Influence of height/radius. Slab (•). sphere (o).

Itwa s tried before toquantitativel y describe the syneresis rate under conditions resembling the situation during practical cheesemaking (Kirchmeier, 1972; Marshall, 1982; Weber, 1984). This resulted in widely differing relationships, which were discussed by Walstra et al. (1985). For the proportionality of the initial relative shrinkage rate of a individual cubic curd particle (5x5mm )wit ht r oneca ndeduc efro mTabl e6. 3 thatfo r a constant pressure of about 10 Pa r is at most 0.4. An even lowervalu ei sexpecte d tob e foundwit hsmalle rradi io rhighe r pressures. Duringpractica lcheesemakin g onewil lcertainl yexperienc ea higher value of r during the initial stages of syneresis. This must partly be attributed to the increase in the surface/volume ratioo f theparticle s as aresul to fcutting . Alsoth epartic ­ les will be damaged due to mechanical treatment, which may result in the occurrence of cracks and local increases in the permeability of the casein matrix (van Dijk, 1982). Thus, the value of rwil l be strongly dependent onprocessin g conditions. Adequate modelling of theprocesse s involved and calculation of

124 the expected r during cheesemaking seem at present not practi­ cable.

6.7 Conclusions

- During cheesemaking the syneresis rate is influenced by a number of variables (mechanical pressure, hydrodynamic effects, souring, changing temperature), of which the separateeffect sar edifficul tt oquantify . - Cutting of the curd may result in pressures of over 100 Pa at the point where the curd fractures. Shear effects may also play a part.Moreover , thepermeabilit y coefficient of the gel can be affected by the forces applied during cut­ ting. - An analysis of the hydrodynamic effects during stirring revealsbul kliqui d turbulenceeffects ,boundar y layershea r forces and collisions as the pressure-inducing phenomena. For the effects of bulk liquid turbulence, the size of the eddies is considered to be of predominant importance. Collisions can give rise to pressures of some hundreds of Pa,dependin go nloca lcondition si nth ecurd/whe ymixture . - The one-dimensional shrinkage rate in case of intermittent pressureca neasil yb eestimated . - Results of model calculations for the one-dimensional case show that mechanical pressure partly masks any effect of changingp Hdurin gsyneresis . - Preconcentration of skimmilk caused the absolute shrinkage tob elowe ran dth erelativ eshrinkag erat et ob ehighe r for gelso fa mor econcentrate dmilk . Resultso fcalculation swit hth eone-dimensiona lmode lan d a preliminary three-dimensionalmode l stressed therol eo fth e surface/volume ratio for the syneresis rate of a curd particle. For the threedimensiona l case this and the extra condensation due to inward shrinking resulted in a propor­ tionalityo fth eshrinkag et ot r,wher er <0.5 .

125 LITERATUREREFERENCE S

AmericanDr yMil kInstitute ,inc .(1971) .Standard sfo rgrade so f drymilk .Bull .916 . Bear,J .& Verruijt ,A .(1986) .In :" Modelle nvoo r grondwatertransport".Orgaa nvoo rpostacademisc honderwij si nd e technischewetenschappen ,intersecti eGezondheidstechniek , Delft. Berridge,N.J .& Scurlock ,P.G . (1970).J .Dair yRes .37 ,417-429 . Bohlin,L. ,Hegg ,P.-0 .& Ljusberg-Wahren ,H .(1984) .J .Dair ySei . 67,729-734 . Buma,T.J . (1965).Neth .Mil kDair yJ .19 ,249-265 . Chaplin,B .& Green ,M.L .(1980) .J .Dair yRes .47 ,351-358 . Cherry,A.S .& Papoutsakis ,E.T .(1986) .Bioproces sEng .1 ,29-41 . Crank,J . (1975)."Th emathematic so fdiffusion" .Oxfor d Univ. Press,London . Creamer,L.K . (1985).Milchwissenschaf t40 ,589-591 . Culioli,J .& Sherman ,P . (1978).J .Textur eStudie s9 ,257-281 . Dalgleish,D.G . (1979).J .Dair yRes .46 ,653-661 . Dalgleish,D.G . (1980).J .Dair yRes .47 ,231-235 . Dalgleish,D.G .& Parker ,T.G . (1980).J .Dair yRes .47 ,113-122 . Dalgleish,D.G . (1981).J .Dair yRes .48 ,65-69 . Dalgleish,D.G . (1982).In :P.F .Fo x(ed.) , "Developmentsi ndair y chemistry",vol .1. ,pl57-187 ,Appl .Sei .Publ. ,London . Dalgleish,D.G . (1983).J .Dair yRes .50 ,331-340 . Dalgleish,D.G . (1984).J .Dair yRes .51 ,425-438 . Dalgleish,D.G . (1987).In :P.F .Fo x (ed.),"Cheese :chemistry , physicsan dmicrobiology" ,vol .1 ,p63-96 ,Elsevie rAppl .Sei . Publ.,Londo nan dNe wYork . Darling,D.F .& Dickson ,J .(1979) .J .Dair yRes .46 ,441-451 . Darling,D.F .& va nHooydonk ,A.CM . (1981).J .Dair yRe s48 , 189-200. Darling,D.F . (1982).in :Th .F .Tadro s (ed.), "Theeffec to f polymerso ndispersio nproperties" ,p285-300 ,Acad .Press , London. Dejmek,P . (1987).J .Dair ySei .70 ,1325-1330 . Dijk,H.J.M ,van ,Walstra ,P .& Geurts ,T.J . (1979).Neth .Mil k DairyJ .33 ,60-61 . Dijk,H.J.M ,va n(1982) .Doctora lthesis ,Agricultura lUniversit y Wageningen,th eNetherlands . Dijk,H.J.M ,van ,Walstra ,P .& Schenk ,J . (1984).Chem .Eng .J . 28,B43-B50 . Dijk,H.J.M ,va n& Walstra ,P .(1986) .Neth .Mil kDair yJ .40 , 3-30. Emmons,D.B. ,Price ,W.V .& Swanson ,A.M . (1959).J .Dair ySei . 42,866-869 . Ferry,J.D . (1980)."Viscoelasti cpropertie so fpolymers" ,3 rd ed., JohnWile y& Sons ,Ne wYor kan dLondon . FIL-IDF,internationa lstandard .21 :1962 . FIL-IDF,internationa lstandard .33A :1971 . Fox,P.F . (1981).Neth .Mil kDair yJ .35 ,233-253 . Garnot,P . (1986).ID FAnn .Bulleti nB-Do c123 . Geerts,J.P. ,Bekhof ,J.J .& Scherjon ,J.W . (1983).Neth .Mil k DairyJ .37 ,197-211 .

126 Glaser,J. ,Carroad ,P.A .& Dunkley ,W.L .(1979) .J .Dair ySei .62 , 1058-1068. Glaser,J. ,Carroad ,P.A .& Dunkley ,W.L .(1980) .J .Dair ySei .63 , 37-48. Green,M.L .& Crutchfield ,G . (1971).J .Dair yRes .38 ,151-164 . Green,M.L. ,Hobbs ,D.G. ,Morant ,S.V .& Hill ,V.A . (1978).J . DairyRes .45 ,413-422 . Green,M.L. ,Turvey ,A .& Hobbs ,D.G . (1981).J .Dair yRes .48 , 343-355. Green,M.L .(1987) .J .Dair yRes .54 ,303-313 . Gufferty,M.B .& Fox ,P.F . (1985).Ir .J .Foo dSei .Techn .9 ,1-9 . Gyr,A . (1944).Ann .Paediatric i163 ,314-327 . Have,A.J .va nde r (1954).Neth .Mil kDair yJ .8 ,157-162 . Heertje,I.J. , Visser,J .& Smits ,P . (1985).Foo d Microstructure 4,267-277 . Henstra,S .& Schmidt ,D.G . (1970).In :P .Favar d (ed.)"Proc .7 th Int.Congr .Electro nMicr .1 ,p389-390 . Hinze,J.O . (1971).Prog .Hea tMas sTransfe r6 ,381-385 . Holt,C . (1975).Proc .Int .Conf .Surf .Sei. ,Akademia iKiado , Budapest,p641 . Holt,C .& Dalgleish ,D.G . (1986).J .Colloi d Interf.Sei .114 , 513-524. Hooydonk,A.CM .van ,Olieman ,C .& Hagedoom ,H.G . (1984).Neth . MilkDair yJ .38 ,207-222 . Hooydonk,A.CM .van ,Hagedoom , H.G.& Boerrigter ,I.J . (1986a). Neth.Mil kDair yJ .40 ,281-296 . Hooydonk,A.CM .van ,Boerrigter , I.J.& Hagedoom , H.G.(1986b) . Neth.Mil kDair yJ .40 ,297-313 . Hooydonk,A.CM .van ,Hagedoom ,H.G .& Boerrigter ,I.J . (1986c). Neth.Mil kDair yJ .40 ,369-390 . Hooydonk,A.CM .van ,d eKoster ,P.G .& Boerrigter ,I.J . (1987). Neth.Mil kDair yJ .41 ,3-18 . Hooydonk,A.CM .va n& Walstra ,P . (1987).Neth .Mil kDair yJ . 41,19-47 . Hooydonk,A.CM .va n& va nde nBerg ,G . (1988),IDF-documen t(i n press). Home,D.S . (1984).Biopolymer s23 ,989-993 . Home,D.S . (1986).J .Colloi d Interf.Sei .Ill ,250-260 . Humbert,G .& Alais ,C (1979).J .Dair yRes .46 ,559-571 . Jager,E.M .de ,va nde nTempel ,M .& d eBruyne ,P . (1963).Proc . KoninklijkeAcademi eva nWetenschappe n B-66,17-35 . Jolies,J. ,Alais ,C .& Jolies ,P . (1968).Biochim .Biophys .Act a 168,591-593 . Jong,L .d e (1975).Neth .Mil kDair yJ .29 ,162-168 . Kammerlehner,J . (1974).Deutsch eMolkerei-Ztg .95 ,342-350 . Kamphuis,H. ,Jongschaap ,R.J.J .& Bouter ,P . (1984).J .Coll . Interf.Sei .98 ,459-466 . Karman,A.H .& va nBoekel ,M.A.J.S .(1986) .Neth .Mil kDair yJ .40 , 315-336. Kirchmeier,0 . (1972).Milchwissenschaf t 27,99-102 . Kitchen,B.J . (1985).In :P.F .Fo x(ed. )"Development s inDair y Chemistry";vol .3. ,p239-279 ,Elsevie rAppl .Sei .Publ. ,Londo n andNe wYork . Knoop,A.M .& Peters ,K-H . (1975).Kieler .Milchw .Forsch .Ber .44 , 227-248. Koestler,G .& Petermann ,R . (1936).Landw .Jahrb .Schwei z103-120 . 127 Lang, K. &Miethke ,M . (1933).Milchw .Forsch . 14,195-199 . Lawrence, A.J. (1959).Austr . J. Dairy Techn. 14,169-172 . Leclerc,D . &Rebouillat ,S . (1985). In:A .Rushto n (ed.), "Mathematical models and designmethod s in solid-liquid separation", p355-391,Martinu s NijhoffPubl. , Dordrecht. Lelievre, J. (1977). J. Dairy Res.44 ,611-614 . Lucisano, M., Peri,C . &Donati ,E . (1985).Milchwissenschaf t 40, 600-604. Luyten, J.M.J.G. (1988).Doctora l Thesis,Agricultura l University Wageningen, The Netherlands. Marshall, R.J. (1982). J. Dairy Res.49 ,329-336 . McGann, T.C.A.,Donnelly , W.J., Kearney, R.D. &Buchheim , W. (1980).Biochim . Biophys.Act a 630,261-270 . Mulder, H., deGraaf , J.J. &Walstra , P. (1966). In:Proc .17 th Int.Dair y Congr., D, p413-420. Parker, T.G. &Dalgleish , D.G. (1981). J. Dairy Res.48 ,71-76 . Pearce, K.N. (1976). J. Dairy Res.43 ,27-36 . Pearse, M.J., MacKinley, A.G., Hall,R.J . &Linklater , P.M. (1984). J. Dairy Res.51 ,131-139 . Pearse, M.J., Linklater, P.M., Hall,R.J . &MacKinlay , A.G. (1986). J. Dairy Res. 53,477-480 . Peri,C. , Lucisano, M. &Donati , E. (1985).Milchwissenschaf t 40, 650-652. Pierre,A . (1983). LeLai t 63,217-229 . Ovist,K.B . (1979a).Milchwissenschaf t 34,467-470 . Ovist,K.B . (1979b).Milchwissenschaf t 34,600-603 . Ramet,J.P. ,E lMayda , E.& Weber , F. (1981).Milchwissenschaf t 36, 143-146. Reiner,M . (1971). "Advanced Rheology", H.K. Lewis, London. Roefs,S.B.F.M . (1986).Doctora l thesis,Agricultura l University Wageningen, the Netherlands. Saito, Z. (1973).Neth .Mil k Dairy J. 27,143-162 . Sammis,J.L. , Suzuki, S.K. &Laabs ,F.W . (1910).Bur . Animal Industry, USDA, Bull.122 . Scheidegger, A.E. (1960). "Thephysic s of flow through porous media". Oxford Univ. Press, London. Schmidt, D.G. &Poll , J.K. (1986).Neth .Mil k Dairy J. 40,269-280 . Schmidt, D.G. (1982). In:P.F . Fox (ed.), "Developments in dairy chemistry", vol.1. ,p61-86 ,Appl . Sei. Publ., London. Schmutz,M . &Puhan , Z. (1980).Schweiz .Milchw . Forschung 9, 39-48. Schmutz,M . &Puhan , Z. (1978). Schweiz.Milchw . Forschung 7, 69-72. Shinnar, R. &Church , J.M. (1960). Ind. Eng.Chem . 52,253-256 . Shirato, M., Murase, T & Iwata,M . (1986). In:R.J . Wakeman (ed.), "Progress in filtration and separation 4",pl81-288 , Elsevier, Amsterdam. Snoeren, T.H.M.,Klok , H.J., vanHooydonk , A.C.M. &Damman , A.J. (1984).Milchwissenschaf t 39,461-463 . Sood, S.M., Sidhu, K.S. &Dewan , R.K. (1976).Milchwissenschaf t 31, 470-474. Sood, S.M., Gaind, D.K. &Dewan , R.K. (1979).N.Z .J . Dairy Sei. Techn. 14,32-34 . Southwell, R.V. (1940). "Relaxation methods in engineering science", Oxford Univ. Press, London.

128 Spaargaren,A . (1987).M .Sei .Thesis ,Agricultura lUniversit y Wageningen,Th eNetherlands . Stoll,W.F .(1966) .Ph.D .Thesis ,Univ .o fMinneapolis ,Minnesota . Storry,J.E .& Ford ,G.D . (1982).J .Dair yRes .49 ,469-477 . Straatsma,J .& Heijnekamp ,A .(1987) .Zuivelzich t79 ,24-28 . Swaisgood,H.E .(1982) .In :P.F .Fo x (ed.),"Development si ndair y chemistry",vol .1. ,pl-59 ,Appl .Sei .Publ. ,London . Tarodod el aFuente ,B .& Alais ,C .(1975) .J .Dair ySei .58 , 293-300. Terzaghi,K . (1965)."Theoretica lSoi lMechanics" ,2 nd ed.,Wiley , NewYork . Thome,K.E. ,Axelsson ,I .& Liljegren ,G . (1958).Mil kDair yRes . Rep.Alnar p53 . Tokita,M. ,Futakuchi ,M. ,Niki ,R. ,Arima ,S .& Hikichi ,K . (1983).Biorheolog y 20,1-10 . Visser,S. ,va nRooijen ,P.J .& Slangen ,C.J . (1980).Eur .J . Biochem.108 ,415-421 . Visser,J. ,Minihan ,A. ,Smits ,P. ,Tjan ,S.B .& Heertje ,I.J . (1986).Neth .Mil kDair yJ .40 ,351-368 . Visser,S .(1981) .Neth .Mil kDair yJ .35 ,65-88 . Vliet,T .van. ,Zoon ,P. ,va nDijk ,H.J.M .& Walstra ,P .t ob e published Vries,E .d e& va nGinkel ,W .(1984) .NIZ ORappor tR 120 . Waarden,M .va nder .(1947) .Onderzoe knaa rd efactore ndi einvloe d hebbeno pd ewei-uittredin gui tgestremd emelk .Alg .Ned . ZuivelbondFNZ ,Th eHagu e(unpublished) . Walstra,P . (1979).J .Dair yRes .46 ,317-323 . Walstra,P. ,Bloomfield ,V.A. ,Wei ,G.J .& Jenness ,R .(1981) . Biochim.Biophys .Act a669 ,258-259 . Walstra,P .& Jenness ,R . (1984)."Dair yChemistr yan dPhysics" , JohnWile y& Sons ,Ne wYork . Walstra,P. ,va nDijk ,H.J.M .& Geurts ,T.J . (1985).Neth .Mil k DairyJ .39 ,209-246 . Walstra,P .& va nVliet ,T .(1986) .Neth .Mil kDair yJ .40 , 241-259. Waugh,D.F . (1971).I n: H.A .McKenzi e (ed.),"Mil kproteins" , vol.2 ,p3-85 ,Ac .Press ,Ne wYor kan dLondon . Weber,F . (1984).In :A .Ec k(ed. )"L eFromage" ,p22-36 ,Lavoisier , Paris. Wurster,K . (1934).Milchw .Forschunge n16 ,200-214 . Zoon,P . (1988).Persona lcommunication . Zoon,P. ,va nVliet ,T .& Walstra ,P .(1988) .t ob epublishe di n Neth.Mil kDair yJ . Zuraw,J. ,Qvist ,K.B. ,Madsen ,J.S. ,Braun ,H .& Kjaergaar d Jensen,G . (1985).Repor tno .265 ,Staten sMejerifors0g , Hiller0d.

129 LISTO FSYMBOL S

a Lagrangianvariabl e a slope(se eTabl e3.3 ) (m«s~)r

ai ionsiz eparamete r A constanti nEq .(2.3 ) A breakingstrengt h(constan ti nEq .5.15 ) (N) constanti nEq .(2.3 ) B permeabilitycoefficien t (m2) permeabilitycoefficien ta tth emomen t Be thege li spressurize d m2) diameter m) d volumesurfac eaverag ediamete r m) interparticlecenter-to-cente rspacin g m) D energydensit y m2• s~ 3) e energyinpu t kW.m'3) E force N) F gravitationalacceleratio n m«s"2) g lossmodulu sfo ra characteristi c timescal et * N«nr ) height m) h reducedradiu si nEq .(6.4 ) m) distancet oth esurfac eo fa sla b m) H lengtho fa ge li na ge ltub e m)

»o initialheigh to fa ge lsla b m) AH shrinkageo fa ge lsla b m) ± relativeremainin gvolum e I ionicstrengt h mMol•dm")3 K compressionmodulu s N-nr2) I distance m) L evaluativeparamete ri nEq .(5.22 ) 3 1 Wc collisionfrequenc y nr »s" ) VP pressuregradien t( =dP/dx ) Pa-nr1) initialendogenou ssyneresi spressur e Pa) initialgravitationa lpressur e Pa) syneresispressur ei nslic ek Pa) P"; P ,; appliedmechanica lpressur e Pa) pressurecarrie db yth ematri x Pa) 130 pressureo nth eliqui d Pa) F" maximum initial syneresispressur eo n O, m a x colliding curd particles Pa)

O, m a x maximum initial endogenous syneresis pressure (seeTabl e6.2 ) Pa) Q evaluative parameteri nEq . (3.5) r radius m) summated radiifo rspherica l layer1 t ok m) Re Reynolds number str,. liquid volume fluxi nEq .(3.3 ) m«s")1 time, mostly after starting syneresis s) timefo rwhic hP ? „„ isobtaine d Pa) O, m a x °C) T temperature Tr Trouton number m»s"x) v velocity m'S"1) v superficial liquid velocity m3-s" 1 : ) critical liquid flux m-s" 1 ) average eddy velocity mean liquid velocitywit h respectt o m-s"1) solid particles meanvelocit yo fth esoli d particles m-s"1) with respectt oth ecylinde r m«s"x) rootmea n square relative velocity 3 V m) n initial volume m3) volumeo fa cub e m3) tot totalvolum eo fparticle s m3-s 1) Vstrk liquidvolum eflu x N«m) K kineticenerg y N«m) K work x m) distancet oth ebotto mo fa ge lsla b zi valency Z m2) contactsurfac e freeio nactivit ycoefficien t r exponent(se eTabl e3.3 ) 6 m) porediamete r A difference e porosity e relativedeformatio n(se eEq .(6.5) ) 131 e0 initialporosit y H dynamicviscosit y (Pa«s)

nE elongationalviscosit y (Pa-s) X,; Z,*; ç* straini nth esoli dmatri x p massdensit y (kg«nr3) a stress (Pa)

132 SUMMARY

This study deals with the influence of cheesemaking para­ meters, i.e. variations in milk properties and process condi­ tions asma y commonly occur,o n the syneresis of rennet-induced milk gels. The shrinkage of theprotei nmatri x isconsidere d in relationt oth epressur e (whetherendogenou so rexternal )o nth e liquid phase and theresistanc e against flow through thepores , asexpresse d inth epermeabilit ycoefficient . After a brief introduction in Chapter 1, the experimental approach is described in Chapter 2. Generally, reconstituted skimmilk was used for the determination of the shrinkage rate andth epermeabilit ycoefficien to frennet-induce d skimmilk gels underclosel ycontrolle dconditions . In Chapter 3 the concept of one-dimensional syneresis is considered in more detail. This approach offers possibilities for comparison of experimental results and calculations with a numerical model, which is based on the equation of Darcy. The origin of the matrix-induced pressure on the whey (theendoge ­ nous syneresis pressure) is treated in terms of the reactivity of thebuildin g blocks,thei r probability tocontac t oneanoth ­ er, the breaking of formed strands, and the occurrence of internal rearrangements in the strands. The change with time after rennet addition canb e explained by relaxationphenomena . Experimental results for various pH and degree of preconcen- tration underline the importance of thementione d processes. It can be concluded that theendogenou s syneresispressur emus t be related to the Theological properties in a very subtle way. Quantitatively, theendogenou ssyneresi spressur eappeare d tob e a poorly definable parameter with aver y temporary and history- dependentcharacter .However ,evaluatio no fsom etria l functions for the change in the endogenous syneresis pressure during macroscopic shrinkage revealed a rather limited effect on the shrinkage rate. According to model calculations the latter is largely determined by the change of the permeability in the outer layers.Furthermore ,i ti sshow ntha tpermeabilit ydat aa s obtained with gels from preconcentrated milk result in an

133 underestimationo fth epermeabilit ycoefficien ta sa functiono f thedegre eo f shrinkagedurin gsyneresis . Experimentally, deviations from the expected proportionality of the shrinkage to the square root of time after starting syneresiswer e found. Thiscoul d notb e easily accounted for in modelcalculations ,bu tma yb eexplaine d byth e rate-determining visoo-elastic behaviour of the casein network in the outer layers. The deviation from the square root proportionality does not necessarily lead to strongly different values for the initial endogenous syneresis pressure, as calculated from experimental resultsb yassumin ga squar eroo tproportionality . In Chapter 4 the effect of variation in experimental condi­ tionsdurin g renneting andcoagulatio no nsyneresi si sdiscusse d in relation to the structure and the properties of the casein network. After having selected a proper preparation procedure for reconstituting skimmilk, attention is given to the effects of cold storage, rennet concentration, pH and acidification

procedure, CaCl2 addition, NaCl addition and temperature. Cold storage caused only minor effects.Mos t striking is the higher permeability coefficient for a higher temperature and a lower pH,whic hca npartl yb eaccounte d forb yth esmalle rsiz eo f the building blocks. Furthermore, higher values of the endogenous syneresis pressure are derived for such conditions, especially for a lower pH. Thecours eo f theendogenou s syneresis pressure with time is clearly related to the stage of gelation. The maximumvalu emus tb elargel ydependen to nth ereactivit yo fth e particles and the rate of strengthening of the strands. The aggregating tendency of paracasein micelles is markedly higher

shortly after a change in pH or addition of CaCl2, as compared tochange sinduce da tlonge rtime sbefor erenne taddition . The effect of fat content on the shrinkage rate can not simply be explained with a higher volume fraction of the dry matter. Apparently, also the permeability of the matrix is affected. The important effect of mechanical pressure is treated in Chapter 5. The use of glass filter plates allows accurate and reproducible determination of the shrinkage rate at various pressure for the one-dimensional case. The shrinkage was found 134 tob e proportional toth e squareroo to f timeafte r loadappli ­ cation. The strong promoting effect of mechanical pressure on syneresisrat eca nb eclearl yestablishe d inthi sway . The precisely known value of the now dominating mechanical pressure considerably facilitates the analysis of the shrinkage process. Differences between model calculations and theexperi ­ mental resultswer e fairlysmal lan dca nfo rth egreate rpar tb e attributed to an inadequate description of the shrinkage in the outer layers with various conditions. Observed differences can beexplaine d by stress relaxation phenomena,whic h may occur in the protein matrix under the influence of an applied pressure. Broadly speaking, high pressures may lead to a fairly imper­ meable outer curd layer, which strongly impedes further syne­ resis. The experimental results are evaluated in the light of an existing theory for the release of liquid from two-phase elas­ tic,dispers e systems,du et ocompression .Althoug h somereserv e must be exerted with respect to the applicability of such a theory,th ecalculate d permeability coefficients arei nth esam e range as the experimentally obtained values. Moreover, the average breaking strength of individual strands can be calcu­ lated and is found to be 10"11 to 10"12 N, which is not at variancewit ha nestimat ebase do nrheologica lparameters . InChapte r 6 some aspects relevant topractica l cheesemaking aretreated .A roughanalysi so fth emechanica l and hydrcdynamic effectsdurin g theinitia l stageso fcuttin gan ddurin g stirring in a curd-tank is given. Bulk liquid turbulence effects,boun ­ dary layer shear forces and collisions are considered to be of predominant importance.Involve dpressure sar eestimated . It is shown that for the one-dimensional case the effect of intermittent pressure can easily be calculated. Furthermore, results of model calculations for decreasing pH are given. Mechanical pressure can mask the effect of a decreasing pH considerably. Therelativ e shrinkagerat eo fgel sfro mpreconcentrate dmil k ishighe rwit ha large rdegre eo fpreoonoentration ,althoug h for the absolute shrinkage the opposite is found. It is concluded taht close attention should be paid to the effects of differ- 135 encesi nth eshrinkag eprofil ewhe ncomparin g theshrinkag erat e of a partly syneresed gel with the one for a corresponding gel frompreconcentrate dmilk . Results of calculations with a preliminary three-dimensional model stress the importanceo f the surface/volume ratio for the syneresisrat eo fa cur dparticle .

136 SAMENVATTING

Deze studie behandelt de synerese van melkgelen die door stremselinwerkingzij ngevormd .D eaandach tword tdaarbi j inhe t bijzonder gericht op de invloed van een aantal procesvariabelen zoals we die kennen uit de kaasbereiding. Het krimpen van de eiwitmatrix wordt beschouwd in termen van de druk (zowel endo­ geenal sextern )o pd evloeisto f inhe tge le nva nd estromings ­ weerstand ind e poriënva n het gel,uitgedruk t inee npermeabi - liteitscoefficient. Na eenkort e inleiding op het onderwerp inhoofdstu k 1volg t in hoofdstuk 2 een beschrijving van het experimentele gedeelte van dit onderzoek. Voor de bepaling van de krimpsnelheid en de permeabiliteitscoefficient onderzorgvuldi gbeheerst eomstandig ­ heden werd in het algemeen gereconstitueerde magere melk ge­ bruikt. In hoofdstuk 3 wordt ééndimensionale synerese nader bespro­ ken. Deze benadering biedt mogelijkheden voor de vergelijking van experimentele resultaten met die van berekeningen aan de hand van een numeriek model dat isgebaseer d op de vergelijking vanDarcy . Voor een beter begrip van het verschijnsel endogene druk wordend ereactivitei tva nd ebouwstenen ,d ekan so pvormin g van bindingen tussen deze bouwstenen, het breken van gevormde strengen en het optreden van interne herrangschikking in de strengennade rbelicht .He tverloo pva nd eendogen e syneresedruk met de tijd na stremseltoevoegin g is mede afhankelijk van relaxatieprocessen. Het belang van de genoemde factoren wordt gesteund door experimentele resultaten bij verschillende pH en indikkingsgraad vand egebruikt emelk .Ee ne nande rwijs to pee n subtiele relatie met de reologische eigenschappen. De endogene syneresedruk blijkt in kwantitatieve zin een moeilijk te be­ schrijven parameter te zijn met een tijdelijk karakter, die bovendien nogal afhankelijk is van de voorgeschiedenis. Hoe de endogene syneresedruk precies verloopt met voortschrijdende synerese heeft volgens resultaten van modelberekeningen overi­ gensmaa ree nbeperkt e invloedo pd ekrimpsnelheid . Depermeabi - liteit ind ebuitenst e lagen isva n overheersend belang voor de 137 krimpsnelheid. Er kon worden vastgesteld dat bij gebruik van gegevens uitmetinge n aangele n uit door ultrafiltratievoorge - concentreerdemel kd epermeabiliteitscoefficien tal sfuncti eva n deindikkin g tijdenssyneres eword tonderschat . Experimenteel werden afwijkingen van de verwachte evenredig­ heid tussen krimp en de wortel uit de tijd na aanvang synerese gevonden.Di tka nnie tworde nverklaar d aand ehan dva nmodelbe ­ rekeningen. Kort na aanvang van het synereseproces isd evisco - elastische vervorming van de eiwitmatrix in de buitenste lagen mogelijk bepalend voor de krimpsnelheid. De afwijking van de evenredigheid tussen krimp en wortel tijd hoeft bij de bere­ kening van de initiële endogene syneresedruk nietnoodzakelij ­ kerwijs te leiden tot een grote afwijking van de waarde die wordtbereken donde raannam eva nzo' nevenredigheid . De invloed van variaties in de experimentele omstandigheden tijdens stremmen en coagulatie op de structuur en de synerese- eigenschappen van het gel worden besproken in hoofdstuk 4. Aandacht is geschonken aan de invloed van het koud bewaren van de melk, de stremselconcentratie, pH en aanzuurprocedure,

toevoeging van CaCl2, toevoeging van NaCl en de temperatuur. Koud bewaren van de melk had maar een zeer beperkte invloed op een en ander. Met name bij hogere temperatuur en lagere pH was de niet-gesynereerde matrix beter doorstroombaar. Dit kan gedeeltelijk worden toegeschreven aan de zwellingstoestand van de bouwstenen. Berekeningen lieten eveneens hogerewaarde n zien voor de initiële endogene syneresedruk, vooral bij lagere pH. Het verloop van de endogene syneresedruk is sterk gerelateerd aan het verloop van de gelering. De maximale waarde wordt voornamelijk bepaald door de reactiviteit van de paracaseine- micellen en de snelheid waarmee de weerstand tegen vervorming van de strengen toeneemt. De neiging tot aggregeren van para- caseinemicellen is groter naarmate een verandering in pH of

toevoegingva nCaCl 2 korterva ntevore nheef tplaatsgevonden . Vet verhoogt de volumefractie van de droge stof. Dit kan echter maar als een gedeeltelijke verklaring dienen voor de lagere syneresesnelheid bij aanwezigheid van vet. Ook de door- stroombaarheid vand eeiwitmatri xword tbeïnvloed .

138 Debelangrijk e invloed vanmechanisch edru k komt aand e orde inhoofdstu k 5.Nauwkeurig ee nreproduceerbar emetinge nvoo rhe t ééndimensionale geval zijn mogelijk door gebruik te maken van glasfilters.D eexperimentee lbepaald ekrim pblijk t indi tgeva l welevenredi gt ezij nme tworte ltij dn adrukaanlegging . Doordat de waarde van deze overheersende drukcomponent precies bekend is,word td eanalys eva nhe tkrimpproce svergemakkelijkt . Geconstateerde verschillen met resultaten van modelberekeningen zijnvri jgerin ge nzij nwaarschijnlij khe tgevol gva nontoerei ­ kendeinformati eove rhe tverloo pva nd eindikkin g ind ebuiten ­ ste lagen tijdens synereren onder uiteenlopende omstandigheden. Deverschille nhange nmogelij k samenme the trelaxatiegedra g van de eiwitmatrix in relatie tot de aangelegde druk. Globaal mag worden gesteld, dat hoge drukken kunnen zorgen voor de vorming van een weinig doorstroombare buitenste laag, waardoor verdere syneresester kword tgehinderd . Evaluatie van de resultaten aan de hand van een bestaande theorie voor het uitpersen van vloeistof uit elastische, dis­ perse systemen bleek zinvol. De op basis van syneresemetingen berekende permeabiliteitscoefficient komt redelijk overeen met een anderszins gevonden waarde.Bovendie nworde nvoo rd e gemid­ delde kracht die nodig is voor breuk van een streng waarden tussen 10"ll en 10"12 N gevonden, wat ruwweg overeenkomt met informatieui treologisch emetingen . Enkele meer direct relevante aspecten voor de kaasbereiding in de praktijk worden behandeld in hoofdstuk 6. Er wordt een overzicht gegeven van enkele mechanische en hydrodynamische effecten tijdens het snijden en het roeren in een wrongelbe- reider. Turbulentie op vrij grote schaal, afschuifkrachten in grenslagen en botsingen zijn van aanwijsbaar belang voor de drukuitoefening opwrongeldeeltjes .D eoptredend edrukke nkunne n globaalworde ngeschat . De invloed van intermitterende druk kan met het beschikbare ééndimensionale model eenvoudig worden berekend. Modelbereke­ ningen in geval van een dalende pH laten de maskerende invloed vanmechanisch edru kduidelij kzien . Voor gelen uit door ultrafiltratie voorgeconcentreerde melk was de relatieve krimp hoger bij een verdere indikkingsgraad. 139 Wanneer nietword t gecorrigeerd voor de indikking, geldt echter hetomgekeerde .Aandach tvoo rhe tconcentratieprofie l isnoodza ­ kelijkbi jbeschouwin gva ndi tsoor tresultaten . Voorlopige berekeningen met een driedimensionaal model onderstrepen het belang van de verhouding oppervlakte/volume voord ekrimpsnelhei d vanee nwrongeldeeltje .

140 CURRICULUMVITA E

De auteur werd geboren op 27 juli 1958 te Oisterwijk. Hij behaalde in 1976 het Atheneum-B diploma en begon in hetzelfde jaar met de studie aan de Landbouwhogeschool te Wageningen. In juni 1980 werd het kandidaatsexamen Levensmiddelentechnologie afgelegd, terwijl in november 1983 het doctoraaldiploma werd behaald met als hoofdvakken Zuiveltechnologie en levensmid- delenchemiee nal sbijva kMarktkund ee nmarktonderzoek . Tussen 1maar t 1984 en 1maar t 1987 werkte hij alsonderzoeks ­ assistent bij de sectie Zuivele n Levensmiddelennatuurkundeva n deLandbouwhogeschoo l (laterLandbouwuniversiteit) .Sind s1 jun i 1987 is hij werkzaam bij de Coöperatieve Vereniging voor Melk- onderzoek "Zuid-Nederland"W.A . teVeldhoven .

141