Structure and Function of Food Products: A Review

Abstract A proper understanding of the behavior of food products requires knowledge of its structure, i.e., the spatial arrangement of the various structural elements and their interactions. The structure can properly be studied by visual observation techniques. In products such as fat spreads, creams, dressings, cheese, bread, milk, yoghurt, whipped cream, and ice cream, different structural elements can be distinguished. A number of those elements are discussed, viz., water droplets, oil droplets, gas cells, particles, fat crystals and strands. In addition examples of interactions between structural elements are presented, viz., oil droplets/matrix, protein/protein, protein carbohydrate, and fat crystal/fat crystal interactions. Finally, it is indicated how these elements cooperate in the formation of structure and contribute to function and macroscopic behavior of food products. Particular attention is given to fat spreads, processed cheese, protein gelation, and examples of the mutual interaction of milk proteins and of carbohydrates with milk proteins. It is expected that a proper understanding of the relation between structure and function will help us to design new ways of structuring in our continuing efforts to manufacture high quality, healthy and tasty food products.

Paper received December 2 1, 1992 Manuscript received Apri l, 7 1993 D irect inquiries to I. Hec rtje Telephone number: 3 1-10-460-5 51 3 Fax number : 3 1-10-460-5800 343 Introdu c ti on Most food products a re composed of rathe r a limited variety of structu ral elemen ts, such as droplets, ai r cells , granules, crystals, strands, mice lles and inte rfaces. A prope r unde rstand ing o f the behavio r of a food product requi res know ledge of its st r uc ture , i.e., the spatial arra ngement of th e st ruc tural e leme nts a nd th e ir in teract ions. Inte rac ti o n forces de termin e th e co nsiste ncy an d th e ph ys ica l stabilit y of food produ c ts.
T he spat ia l a rra ngement of struc tura l e le me nt s can be studi ed by visua l obse rva ti o n tec hniqu es, suc h as light -a nd e lec tron mi croscopy (E M ). Al so, th e res ult of the inte rac tion be tween th e va rious compo ne nt s in a system can be studied in this manne r . Visua l observation techn iques are therefore an important aid in the analys is of food structure.
A variety of techniques are availab le for the determination of functional properties. Rheological measu rements give insight into mechanical prope rt ies and consistency. Consume r panels, and rheo logica l characte rizations are used to measure sen sory prope rties. The stability of a foam can be followed by visual inspection.
It is the aim of this review to illustrate how va rious structural elements cooperate in th e formation of structure and contribute to funct iona l prope rt ies and macroscopic behavio r o f food p roducts. To th is end, a number of struc t ura l e leme nt s, as obse rved by mic rostru c tura l tec hniques, wi ll be descr ibed as we ll as exa mp les o f the i r int e ract io n . F ina ll y, a number of food systems wi ll be discussed .

St ru ct ur·a l E le me nts
Wa t er dro p lets Water droplets are important structural elements in emulsion type food products such as margarine , butte r and low-fat spreads. The drop lets are stabilized by soluble emulsifiers, such as monoacylglycerols and lecithins (Madsen , 1987) and /or by solid particles, such as fa t crys tals (Lucassen-Reyndc rs and van de n Tempe l, 1963).
An example of wate r drop lets stabilized by a sol ub le emulsifier in a wate r-in-oil emu lsion , obse rved by I. Heertje F igu r e I. Water droplets in a water in oil emulsion stabilized by DKF 10 , observed by freeze-fracture EM. Arrow points to multilayaed shells.        freeze fracture EM is given in Fig. 1. The emulsifier used is DK -FlO , which is a mixtu re of di -, tri -, and polyesters of hydrogenated tallow fatty acids with sucrose (VanVoorst Vader and Groeneweg , \989).
Thin -sectioning EM of a wate r drop let in a margarine reveals the presence of emulsifier at the inte rface between oil and wate r (Fig. 2).
Freeze-fractu re EM of an 80 % fa t spread shows (Fig. 3) tha t fractu re occ urs either over th e su r face of th e wa te r d roplets or th rou g h th e dropl et. Fa t crystals o n the surface o f the d ropl e ts are cle arl y perceptib le, ind icati ng stabilization by fa t c rys ta ls. Simi la r obse rvations have been repo rt ed (Prech t and Buch heim, 1980b).
Wate r droplets can a lso be properly obse rved by light microscop ical techniques. Fig. 4 reveals the structure for a 25% fat spread. In this case, the droplets are stabilized by protein, which is located at the oil / water interface .
Oil dr opl cl s Oil or fat drop lets occur as str uctura l e lements in such diverse produc ts as milk , c rea m, c heese, mayonnaise , dre ss ings, mea t produc ts a nd ice c ream. Oi l droplets can be stabilized by high-molecula r mass su rfa c tants, suc h as proteins or by low-molec ular mass su rfa ctant s such as Tweens and lecithins o r combinations thereof. Crystals in the oil phase may induce aggregation of the oil drop lets, resulting in partial coalescence (Boode, 1992 ;Boode and Walstra, 1989). The stabilization of emulsions has been amply discussed (Larsson and Friberg, 1991).
Examp les of oil droplets covered with a smoot h interfacial protein la ye r (Larsson and Friberg,199 1,p. 216) , observed by freeze fracture EM, are given in Fig.  5. Fig. 6 shows milk fat droplets covered with casein micelles in a homoge nized milk, obse rved by thin -sectioning EM.
Bridge formation between oil drop lets in an oi l-inwater emulsion , d ue to fa t c rysta ll iza ti o n , is in d ica ted in F ig. 7 ( Hecrtje, 1993 ) . Such cr ys talliza ti o n phenome na a re also importan t in the produ ction of b utt er by chu rning Buchhc im, 1979, 1980a). Mi lk fat globules of the original cream may surv ive the churning process. A butte r globule, observed by scanning EM (SEM, Fig 8) after deoiling (Heertjc, et a/., 1987), shows an outer crystalline layer composed of highmelting fat , surrounding liquid oil inside.
Mayonna ise is an oil-in-water emulsion containing a hi gh percentage of oil (80% or mo re). As observed with confocal scanning laser mic roscopy (CSLM) (Fig.  9), this hi gh volume of oil causes the fo rmation of a honeycomb structu re of closely packed and often distorted oil droplets. In this case, emu lsion stab ilization is achieved main ly by egg yo lk lec ith in. 346 Figure 10 . Air cells in a shortening . C ryo SEM. Figure II . Ai r cell in a churned product. The ai r interface is cove red with fat globu les. (g) fat globules; (w) wate r. C ryo SEM.
Figu1·e 12 . Air cells in a whipped cream at low (a ) an d (b ) high magnification, showi ng inte rface stabi lized by fa t globu les an d protein fi lm. C ryo SEM .
Fi gure 13. Air ce ll in wh ip ped c rea m. T he a ir cell is stabil ized by fat globules (fg) adsorbed at the interface an d re mnants of prote in fil m (i). T hin -sectioni ng E M . (Cou rtesy B. Brooke r). An important aspect of oil droplets is their size and the homogeneity of the size distribution. It may affect both rheological and sensorial properties.

Cas ce lls
In many food products , such as bread , cake, whipped cream, icc cream, bee r , and c hocolate mousse, gas cel ls play an important role . Foam stability depends on many factors. Gas ce ll s can be immobi lized in a so lid mat ri x, whe reas in ot her cases, stabilization is ac hi eved by p rotei n film s aro und th e a ir cell s. In th e la tt er case, so lid particles can easily rup t ure th e frag il e foam lamellae. Also interactions at the inte rface (e.g., between protein and fat) may enhance foam stabil ity (Buchheim, 1991;Brooker, 1993).
An example of air cells immobi lized in a solid matrix is given in Fig. 10. Air was dosed at an early stage of margarine processing at low solid levels, giving rise to smooth round air bubbles. Deformed air bubbles arc found when high solid conten ts are present at the moment of air introduction , e .g., during c hurning (Fig.  II).
Air ce11s in a whipped cream , observed by cryo SEM, are presented in Figs. 12a ,b. Stabilization in this case is achieved by the combined action of protein and fat globules (Brookeret a/., 1986). Thin sect io nin g EM (F ig. 13) c lea rl y demonstrates the local izat io n of fa t globu les a t th e inte r face ( Brooker , et a/., 1986).
Gas cell s (Fig. 14) play a n i mpo rtan t ro le al so in doug h to obtain proper bread vo lume. In thi s case, th e lamellae between the gas cells consist of a solid mat ri x of starch and protein.
A foam prepared from saturated cl6 and CIS mo noacylglycerols shows a polyhedric structure, indicative of a close packing of air cells (Fig . 15).
1t is evident that there is an enormous difference in shape , size and phase volume of gas cells as well as in the composition of foams. Such diffe rences expla in the wide variety of properties of foamed food products, such as beer and bread. Recently , a new method to determine bubble size distributions in foamed food pro ducts has been described (Bisperink e1 a/. , 1992; Akke r man eta/., 1992).
Starch granu les, case in micelles and ot her partic les Di ffe ren t varie ties of unmodified starch ex hibit wide va ri atio ns in granule appearance (Fi tt and Snyder , 1984). Large differences in shape and size arc observed in sta rch g ranul es from corn, maize , wheat, ri ce and po tato. Fig. 16 g ives an example of a nat ive wheat starch. · Large len ti cular g ranules and small spherical granu les are observed. However , in most finished food products, such as bread, soups, sauces and dre ssin gs, starch is not prese nt in its native fo rm , but gelatinized by heating in th e presence of water . Upon heating above the so-call ed gelat ini zation temperature, irreversib le swelling takes places.
Even after extensive swelling, the sta rch g ranules can still be recognized.
Microstr uctural changes in whea t sta rch dispersions during heating and cooling have been desc ribed (Langton and Hcrmansson , 1989). An example of a 5 % maize starch suspension, heated to 95 oc an d cooled to 20°C is shown in Fig. 17. The more com pac t re gion s 348 represent the irreversibly swollen g ranules which have been show n to be largely amylo pectin , whe reas the less dense , leached out amy lose su rrounds th e compact regions .
Th e re a re a number of serious disadvantages related to th e use of native sta rche s as ingredi ents for food products. In recent yea rs, this has Jed to commercial produc tion of a w hole ran ge of partic ulated modifi ed starches. Maltodex trin s fro m National Starch (N-oil ), Grain Processin g Co rporation (Maltrin), Ceres tar (Snowflakes) and Av ebe (Paselli SA2), should be menti o ned in this rega rd . It has been claimed that in th e latter case, a hig h yield of unifo rm pa rti cles ( 1-2 1-trn) is re spon sib le fo r the fa t mimetic properties of the resulting gel.
In other, non-s tarch , ca rbohyd rate-based products, pa rt icles have also been suggested to be responsible for fat-like properties. For example, in Slendid (Copenhagen Pectin Company) soft particles of Ca-pectinate are approximately 40 1-Lm in diameter whereas particles of microcrystalline cellulose (Avice! , FMC corporation) are approximately 0.2 I'm long .
Al so protein pa rti c les play an important role in food st ru cturing.
Casein mice lles are an important component of milk. In fresh milk , the micelles are p resent in the fo rm of coll oidal particles of about 0 . 1 I'm in diam e ter (Fig . 18 , Kalab , 1993). Their aggregation behavior is respo nsible for such important prod uc ts as c heese and yog hurt .
Other protein -based particles have been developed, the most noteworthy example being Simplesse (NutraSwcct Company ; Singer and Dunn , 1990), p repared from egg white and whey o r milk proteins in a size of about 2 ~-'"' (Fig. 19) . It is claimed, th at particles of this size roll easily ove r one another yielding a Figure I S. Air cells in a foam, prepared from monoglycerides, showing a polyhedric str ucture . Light microscopy. Figure 16 . Large lenticular and small spherica l granules of wheat starch. Cryo SEM. Figure 17. Swell ing of sta rc h g ranules in a 5% maize starch suspension on heatin g to 95°C . (P) dense a mylopectin-rich zo ne; (A) leac hed out amy lose. Cryo SEM . Note: the appa re nt network st ructure is induced by the freezing p rocess during sample preparation an d is not a representation of the true structu re. Figure 18 . Casein micelles from co ws ' milk.
Replica T EM.
(Courtesy M . Kalab). 349 c ream y perception and a ri ch texture. This is in contrast with particles equal o r smaller than 0.1 I'm (e.g., casein micelles) which are perceived as gelati nou s and slippery .
On the othe r hand, particles larger tha n 3.0 I'm are perceived as powdery or g ritt y (Singer and Dunn, 1990).

Fat crysta ls
The formation of texture in fat spreads is the result of c ry stall ization of triacyl g lyce rol s wit h hi g h meltin g points. In general, the crys tals in food products do not occ ur as s in gle crystals, but show diffe re nt types of agg regation. T ri acylg lycero ls c rystalli ze in four different modifications: sub-a, a, {3' and {3 (Larsson, 1986). The sub-a and a modifi cations a re unstable and th e refore do not exist in sp read s. The {3' modification is stable but its c rystal lattice is less well ordered than th at of the {1 modification. The {3 modification has the highest ordering and consequenlly the highest melting point. As a consequence of the rather compli cated triacylglycerol composition, the {1' modification is the predominant one in commercial fat blends. Fat crystals in the {1' modification can be either needles or platelets (Fig. 20).
Sometimes, a defect called sandiness o r graininess is observed in margarines and shortenings (Heertje, 1993).
This phenomenon appears to be caused by th e fo rmatio n of large , and often sphe ruliti c, c rysta l agg regates in the {3 mod ifi cation (Fig. 21 ).

Str·ands
Biopolymeres such as proteins and po lysacc harides are present in many food sys tems, often in the form of gels. The gels are formed by networks of polymer strands. Several types of strand formation can be distinguished (Clark, 1987). Strands may be formed by aggregation of macromolecules or macromolecular assemblies in globular form (particularly proteins). Such strands usually produce particle gels (Dickinson , 1980).

I. Heertje
F igure 20. Fat crystals (pl atelets and need les) in th e .B'-modification. Rep lica TEM.    Alternatively, st ran ds may consist of extended random coil mol ec ule s. These latter st rand s form association networks with entanglements or junction zones.
Num e rous publications deal with the formation of st rands fro m proteins and carbohydrates. Strand s from bovine myosi n show different morphologies, depending on ioni c st rengt h (Hermansson eta/., 1986 ;Herma nsson and Langton, 1988).
Fine strands (Fig. 22a) were formed at low ionic strength (0.25 M KC I), whereas coarse agg rega ted particle strands (Fig. 22b) were form ed at hi gh ionic strength (0.6 M KC I). Similar differenc es are reported for othe r , heated , proteins as a function of pH. Agg regated particle strands are found close to the isoelectric point (IEP) of the protein, whe reas fin e strands are foun d both at low and hi gh pH , awa y from the IEP (Stading and Her mansson, 199 1; Hee rtje and van Kl ee f, 1986 ; Harwalkar and Kal ab , 1985) . Fig.  23 g ives an example for ovalbumin, close to its IEP an d Fig. 24 an example for /l-lactoglobulin at pH 8.0.
Also the morphology of strands of pol ysacc harid es has been amp ly in vest igated (Stokke and Elgsae te r , 1987;Stokke et al., 1989;Hermansson, 1989;Ha rada et a/., 1990). St rands largely varying in size, shap e , flexibility an d branching ca n be observed . An example take n from th e work of Harada et af.
Pi gs. 26a, b present micrographs of heated (80°C) solution s of K-ca rragee nan in 0.1 M KCI and 0.0 1 M NaC l, res pec tiv e ly. A strong influence of th e e lec trolyte type a nd it s concentration on the size, thi c kness a nd flexibi lity (persistence len gth) of the st ra nd s is indi ca ted (H e rmansson eta/., 199 1). Thi s is in agreement with th e well -known in fl uence of potassium ions on th e gelati on behavior of K-carrageenan (Morris et a f. , 1980). Firm cohesive gels of ~~:-carrageenan are obtained in the presen ce of potassiUm, whereas under id e nt ica l conditions of concentration and ionic strength no gels a re formed in the presence of sodium.
In tera ctions By combining the va rious st ru c tural elem ents discussed in th e preceding section, a large numb e r of possi· bl e int eraction types ca n be distinguished , such as drop· le t/ matri x, droplet/air cell , strand/strand , strand/panicle, parti cle/droplet, crystal/crystal, crys tal/drop let, and dropl et/dropl e t int eractions. In this sec tion a numb e r of example s will be prese nted.
Oi l drop lets/ matrix in t e r ac ti on Various authors ha ve discussed the importance of the int erac tion between the dispersed phase and th e continu ous phase in relation to rheologi cal and senso rial properties (Van Vliet and Dentene r-Kikk er t, 1982; Masson and Jost, 1986;Jost e1 at., 1986;Lang ley and Green , 1989 membrane consists mainly of specific proteins and lipids which do not interact with the conti nuous network of casei n st rands in the acid milk gels. Wh e n the milk is homogenized fat globules are tran s formed into markedly smaller globules, cove red wit h milk protein s such as whey proteins and casein micelles. In th a t case, a strong interaction occurs between th e species (i n particular casein micelles) located at th e interface between oil and water and the continuous mat rix as is indicated in Fig .  27 (Kalab , 1993). This is also renec ted in a strong differe nce in rheological behaviou r of th e two types of ac id milk gels (Fig. 28).
Similar observations a re reported for oi l-i n-wa ter emuls ions stabilized by heaHreated w hey protein (Masson an d l ost, 1986; Jost e / a Langley and Green, 1989). Th e size of the di spe rsed oil droplets a ppears to be a very importa nt pa ram e ter for proper gel formation. Emul sion s wi th a hi gh ge lation capacity are cha racterized by a sing le dropl e t family of relatively narrow size distribution and a mean droplet diameter ranging from 0.3 -0.7 J.Lill . Ex tensive coat in g of the oil droplets with coagulated protein le ad to a co ntinuous gel structure in which the oil droplets a re strong ly embedded ( Fi g. 29, Jost e1 a/ . , 1989) . Thi s coating does not occur when the oil droplets are first stabili zed by lecithin. In the latter case , ve ry weak gels are ob tained with smoo th oi l droplets, devoid of any adsorbed p rotei n (Jost eta/., 1989).
Also, gels filled with fat g lobules in a mixed matrix of casein and whey protein have been studied (Agui lera and Kessler, 1989 ; Xiong a nd Kin sella , 1991; Agu ilera and Kin sel la, 1991). By co mbined microstructural and rheol ogica l studi es, be obtained. Stabilization of the globules by whey protein resulted in the strongest gels , followed by stabilization by sodium caseinate and skim milk powder. Stabilization by Tween 80 gave relatively weak gels (Xiong and Kinsella, 1991). These res ults are once more explained considering th e interaction forces between th e inte rface of the fat g lobules and the continuous protein mat ri x. These exa mpl es show the important effect of particle/mat rix interactions on the behavi or of food systems.
Prolc in / p1· otci n int erac ti on As a typical example of protein / protein interaction, the heat induced association between ,8-lactoglobulin and K-casein an d casein micelles will be discussed.
The properties of dairy products are influenced by heat treatment of the milk. The influence of pre-heating at 85-90°C on texture, firmness and syneresis of yoghurt and cheese is well known (Kalab, 1993). Many studies attribute this effect to associations between ,8-lactoglob- of th e interaction has been desc ribed (Haque, 1987) . Minor variations in pH were show n to have severe effects o n the adhe re nce of protein aggregates to casei n mi cell es (Creamer eta!. , 1978). In thi s context, model expe rim ents with I)-lactoglobulin and casein micelles ha.ve been performed as a function of pH. Figs. 30 and 31 show micrographs of the heat -induced association at pH 6.5 and 7.0, respectively: -At pH 7.0, separate micelles and thread -like protein st ructures are visible. At pH 6.5 , ,8-lactog lobulin is concentrated on the micelles; -at pH 7.0, the granular micelle structure is clearly visible. At the lower pH, the micelles have a woolly , fuzzy appearance, apparently due to a reac tion of the outside protein with the micelle surface; -the separate protein panicles at pH 7.0 are in the form of threads with a limited length and sometimes a few cross-l ink s (Fig . 32).
Apparently, a minor change in pH has a severe effect on the react ions taking place at th e micell a r su rface. This may be related to th e properties of th e ,8-lac toglobulin. It has been show n (deWit, 198 1) th a t both the heat stability in milk and the thermal behavior of ,B-Iactog lobulin change quickly in a very narrow pH range between pH 6.7-7.0.
To get a bette r insight into the nature of the free protein particles observed at pH 7.0, micrographs of heated solutions of K-casein, ,8-Jactoglobulin and of a I: I mixture of both proteins have been prepared. Heated Kcasein, often exhibits spherical particles (Fig. 33). Heated ,8-lactoglobulin shows long , irregular thread-like particles , which appear to be connected into a network (Fig. 34). The tested mixture of both proteins exhibits noodle -like truncated particles (Fig. 35), sim ilar to those obse rved for the interaction between (J -lac tog lobulin an d 354 Structure and function of food produc ts F igure 30 . Hea t-induced associatio n of ,6-lactoglobu lin and casein micelles at p H 6.5. Strong inte raction. Nega tive staining.  casein micelles at pH 7.0 (Fig. 31). Apparently , the free protein particles represent the produc t of interaction between ,B-lactoglobulin and K-casein. This has been further substantiated by ultracentrifugat ion and electroph oretic analysis (Smits and van Brouwershaven , 1980). These resu lt s indi ca te that K-c asein exerts an aggregation -l imiting influence, simi lar to its rol e as a protec tive co ll o id in li miting the size o f casein mi celles.

Prolein /ca r·bo hydr a tc inl c r·acl io n
Protein/carbohydrate interaction can be ill ust rated by the reaction between !(-carrageenan and milk prote in s. Carrageenans are widely used as thickening agents and stabilizers in the food industry, in pa rticular, in neutral dairy produc ts such as cocoa milk, puddings , creams, ice-c reams and mou sses. Carrageenans are particul a rl y suited in neutral dairy applications due to their milk reactivity. The milk reactivity in the pH range 6-7 is ascribed to an electrostatic interaction between the negative sulphate groups in carrageenan and positive charges in the K-cascin (Snocrcn, 1976). In this context, the influence of hea t treatment on mixtu re s of K-carrageenan and K-casein as well as mixtu res of !(~carr agee n an an d casein mi celles was studied. K-ca n agcc n:1 n/ K-ca sci rr Mi crographs of K-car rageena n , K-casei n and a mixture of both components, heated at 60oC fo r 10 minutes arc presented in Fig. 36 . It shows the very fi ne car rageenan thr eads with a diameter of about 3 nm (Fig.  36a), sphe ri ca l K-casein partirles with a diamete r of about 15 nrn (Fig. 36b) and the structu re of the mixtu re (Fig. 36c). Some particles of !(-casein appear to be adherin g to the carrageenan th reads, sometimes accompanied by chain formation. Similar observations have been reported by Snoercn eta/. ( 1976) and Snoeren (1976) who ascribe the react ivity of K-ca rrageenan to an e lectrostatic int erac tion between the negat ive sulphate groups in K-carragee nan and positively cha rged areas in the K-casei n.
Replica. Figure 37. A mixture of K-carrageenan and casein micell es hea ted to 80°C for 10 minutes. Note th e pa rti a l brea k-up of case in micelles and adherence of th e ir re mnants to ca rrageena n threads accompani ed by chain fo rmation. Repli ca. figure 38. A mixture of K-carrageena n a nd casei n micell es heated to 90°C for 15 minutes. Note the adherence of casein pa rtic les o f varying size to th e car rageenan threads. Replica.

K-carragec na n/ cascin micelles
Mixtures of K-ca rrageenan and casein mi cel les a t p H 6.6 he a ted for 10 minutes to temperatures of 60, 80 and 90°C have been studied. Th e structu re of i so la ted micell es has been amp ly described a nd discu ssed (th is pape r , Fig . 18 figure 39 . Schematic drawing of the inte rac tions bet ween K-carragccnan and K-cascin a nd casein mi celles, respec ti ve ly . Int e raction be twee n K-ca rragee na n and Kca se in at 60 oC. At low tempera tures (20 and 60 oC) no appreciable disinte g ration o f casein mi ce ll es occu rs wh e reas at high tempe ra tures (90°C) a l mo st complete disint eg ration occurs, acc ompani e d by the appearance of loose casein pa rti cles and the aggregation of casein panic les in threads.
inte rac tion , obtained afte r heating a mixture fo r 10 minutes at 80°C, is presented in Fig . 37. It shows that under th e influence of ca rrageenan , the micelles are partly disintegrated into sma ll e r units th a t partly adhe re to the ca rra geenan threads and fo rm cha in s. It further appears that the ex te nt of disintegration of the casei n micel les depends on the heating regimen. At low temperatures (20 and 60°C), no appreciable disin teg ration of casein mi ce lles occurs and co nseq uently, on ly a very limited degree of protein pa rti c le aggregation into thread s is obse rved . At hi g h temperature (90°C) , almost com pl e te disi nteg ra tion of casei n micelles occurs, accompanied by th e ap pea ra nce of loose casein particles as well as th e agg rega tion of casei n particles into th reads. An e xam ple showing the irreg ular build-u p of di ffe rent size casein panicles on the carrageenan strands is prese nt ed in Fi g. 38. A sc hematic drawing of these obse rvations is shown in Fig. 39.
The mechanism of interaction between ca rrageenan and protein s (mi lk a nd pl a nt protein s) has been the subject of a numbe r of investigations ( Hansen , 1968;Lin and Hansen , 1970;Ch a krabo rty and Randolph, 1972;Ha nsen, 1982;Dea et a/ ., 199 1). crs-a nd ,8-casein reac t with ca rrageen an only in the prese nce of Ca-io ns , whereas th e reaction with K-casein does no t require Ca and is shown to be electrostatic in nature (Snoeren eta/. , 1976;Snoeren, 1976). In this view , the stabilization of a 5and ,6-casein and also of plant proteins against precipitation by Ca-ions is achieved by entrapping small calcium aggregated-protein bodies in the carrageenan structure before suc h bodies can furthe r agglomerate into la rge , colloidally unstable particles . The protein aggregates are considered to interact to a greater extent in some areas of the polysaccharide stra nd a nd not in o th er a reas, but in no case is th e p rotei n di stribution unifor m and continuous over the entire ca rrageenan structure. T his behavior is asc rib ed to the presence of a double he lix structure in th e carrageenan, which is considered to be a region of low protein reactiv ity . The doubl e helix junction zones may provide effective separation of the protein particles from each other, thus imparting physica l stabili ty to the system. This alternating structure of protein-free zo nes (whe re ca rra geenan strands are seen) and of protein -ri ch zones (Hansen,19 82) is presented in Fig. 37 , and in particular, in Fig . 38.
Stronger gels are obtained by heating carrageenan with skim milk or case in micelles than by heating ca rrageenan alone. It is tempting to assume that this phenomenon is related to the present microstructural observations. Under the influence of heat, calcium -se nsiti ve proteins such as as-casei n in casein micelles become exposed (P. Smits, unpublish ed resu lts). Subsequently, a reaction occurs between these calci um-sensit ive proteins and !(-carrageenan accompanied by pa rtial disruption of th e casein mi celles. Protein bodies, formed by aggregation under the influence of ca lcium become part of the carrageenan network. The stronger gel is th e resu lt of the more homoge neo us di stribution of proteinaceous material along the polysaccharide chains and the interaction between the carrageena n network and the protein bodies. This effect on rheol ogica l properties is compa rab le to the influence of panicle/ matrix interactions on rheolog ical properties discussed before .

P ro tei n/ lip id surfacta n t interaction
Interactions of surfac tants with proteins are of importance in a wide variety of systems, such as biomembranes and pharmaceutical preparations. Also, in th e foods area , this type of inte raction plays an important role (Larsson, 1986). Reactions between polar lipids (monoacy lglycerols a nd phospholipids) and p roteins are important in the stabilization of emulsions such as margarine and ice-c rea m (Barford, 1991).
Al so , in bakery doughs, the interaction with the gluten proteins is the major function of lipidic su rfactant molecules (Krog , 1977).
The interaction between milk proteins and naturally occurring fatly acids is the basis of a new product called Lipro (Lipophilized protein) (Haque, 1992). Lipro is claimed to be a fat-like perception enhance r and function s by decreasing particle/ particle interactions thus causing a slippery o r oily feeling.
Fa t crysta l/ra t crysta l in tc l'ac ti o n Fat spreads derive their consistency from 357 interactions between fat crystals which may form a three dimen sional network (Figs. 40a,b). The nature of the interactions between the fat c rystals determines the netwo rk st ructure and the rheology of the fina l product. Two types of bonds are assumed for crystal-crystal interactions (Haighton, 1963): -primary bonds, which result from crys tals growing together at some points. These bonds are regarded as "i rreversib le'' , i.e . , they do not refo rm afte r rup ture; -secondary bonds, whic h are weak London -va n der Waals forces, are "reversible", i.e., they do reform after rupture .
According to others (Shama and Sherman, 1970), such a distinction between primary and secondary bonds is considered to be arbitrary and it was suggested that a true characterization should be based on th e concept of a spectrum of bond strengths.
Many aspects are rel ated to the amount of fat crystals and the nature of the interaction between fat c rystals (Haighton, 1976;deMan et at., 1990). Among these a re : -the hardness of a spread depends on the amount of fat crystals ; -blend composition innue nces the molecular a rrangement and modification of fat crys tal s and consequently th e strength of interactions between crystals; e.g., the prese nce of (1-crystals prevent s the formation of a con tinuous fat crystalline network of ,6' -crys tals (Jurittttnse ttnd Hee rtje , unpubli shed res ult s); slowly c rystallizin g blends co ntinu e to crystalli ze afte r packaging, which favors the formation of a strong network; -hi gh crystallization speeds give rise to many small crystals and soft and overworked products.
An example of the influence of processing on the microstructure and rheology of a shortening is presented in Figs. 4la,b and 42 (Heertje et at., 1988). In product (a), fully crystallized in the processi ng lin e, strong (p rimary) bonds form the major part of the bond strength spectrum. ln product (b) , partially c ry sta lli zed during storage at rest, weak (secondary) bonds dominate. The rh eo logy (Fig. 42) is in agreement with the observed mi c ro st ru c ture.
Exa mpl es of Stru cture a nd Fun c t io n l)roccssed cheese Processed cheese is prepared by mixing natura l cheese and melting salts (sodium citrate or a mixture of sodium. phosphates) for a period of time, the so called c reaming time, at a temperature of90°C. The influence of c reaming time on the microstructure and rheological properties of a pasteurized processed cheese has been investigated.
The torque exerted on the stirrer , which can be considered as a measure of the viscosity of the c heese mass, was recorded as a function of c reamin g time (Fig.  43). The structu re as a function of creaming time was followed by thin sectioning EM. F igu re 44 . Processed cheese after 7 minutes creaming time (sample 1 in F ig. 43). Thin-sec tioning TEM.
F igure 45 . Proce~sed cheese after 110 minutes creaming time (sample 2 in Fig . 43). Note beginning strand formation of the protein phase. Thin-sectioning TEM.
F igure 46. Processed cheese after 160 minutes c reaming time (sample 3 in Fig . 43). Note: strong structuring of the protein phase with parallel alignment of st rand s. Large fat fields (F) are observed . Thin-sectioning EM.
F ig ure 47. Processed cheese after 530 minutes creaming time (sample 4 in Fig. 43). Note the formation of a coagulated, strongly aggregated, protein phase. Thin-sectioning TE M .  I. Heertj e Afte r 7 minutes c ream in g tim e (sa mple 1) , stru cturin g of the protein ph ase is scarcely observab le (Fig.  44). The well -dispersed oil droplets a re about I J.Lm. After 110 minutes (sa mple 2), a limited structuring in the form of protein st ra nd s is observed (Fig. 45) . T he fat phase is not yet affected. After 160 minutes (sample 3 , the max i mum of the torque curve) , a strong stru c turing of the protein phase with distin ct and aligned protein strands is observed, indicating a st rong protein-protein inte rac ti on (Fig. 46). Large fat field s are observed , indicatin g a n expulsion of th e fat phase from the protein matrix . A discon tinuous p rotei n phase can be di scern ed (Fig. 47) in the st ron gly overcreamed case (sample 4), after 530 minutes. The protein shows a n aggregated structure, a type of inte rnal coagulation, and fat and protein are present as se parate phases.
These obse rva tion s are in good agree ment with macroscopic behavior. Sample I (Fi g. 44) shows typical liq uid -like beha vior , sa mpl es 2 (Fig. 45) an d in particular 3 (Fi g. 46) have a gel-like charac te r. Moreo ve r , the latter sam pl e is brittle. Sample 4 (Fi g. 47) is softe r a nd even more bri ttl e than sa mple 3.
As is generally assu med , protein gelation involves th e th e rm a l denaturat io n o f prote in mo lec ules foll owed b y agg regation into a network ( Hermansso n , 1979). The optimal condi tion s fo r gel formati on are a delicate balance bet wee n chain -c hain a nd c hain -so lvent interactions. When th e c hain -chain interaction is too st rong (sa mpl e 4) , the phases may separate and und es irable produc t prope rties may develop; w he n thi s inte ract ion is too weak, no gel will be obta in ed a t all (sample 1).

Protein gels
Understanding th e ge lation be hav ior of protein s is of paramount importance in o rder to manipulate the properties of many food syste ms in the co ntext of produc t or process impro ve ment. Ma ny studies on protein gelat io n have recently been published (Hermansson, 1988;C lark, 1987) . Two di stinctly different gel types can be distingu ished: (i) partic le ge ls com posed of more or Jess spherical protein prec ip itates whic h form an irregular fractal structu re, a nd (ii ) fine st randed ge ls com posed of extended po lyme r mo lec ul es which fo rm entanglements and junction zones. The occurren ce of both ge l types criti ca lly depends on external cond itions, suc h as p H and ion co nce ntration .
Figs. 48 a nd 49 show the microst ru c ture of 20% o valb umin ge ls a t pH 5 a nd pH 10 respectively (Hcertje and van Kl eef , 1986). A homoge neou s di str ibution of fine protein strands is observed at pH 10 , whereas an inhom ogeneous distrib ution of strongly aggrega ted pro tei n particles is found at pH 5 (see also Fig. 23). Consequ e ntl y , the ge l at p H 5 has a much mo re open st ructure than the gel at pH 10. Similar observations have been reported by others (C la rk er a/., 1981; Stading an d He rmansson, 1990, 199 1 ). At the same time , the rheologi cal properties of th e two gel types show grea t differences. The fine stranded ge ls appea r to be muc h more ex tensible than the aggregated pa rticl e gels (Heertje a nd va n Kleef, 1986; Stading 360 and He rm a nsson , 199 1) . However , th e brea kin g stress appea rs to depe nd on the protein conce ntration in a co mpl icated manner. Fo r 12 % il-lactoglobulin gels (Stadi ng and He rm a nsson , 199 1) , the fine stranded gels at pH 7.5 show a low er brea kin g stress (4 kPa) than that of th e pa rti cle gel close to the isoe lect ri c point at pH 6.0 (approximatel y 15 kPa). The opposite be hav ior is obse rved when the prote in concentration is inc reased to 14 % (Table I , Lan gley era/., 1986) . In th is case , the particle gel shows a lower breaking st ress (23 kP a) th a n the fine stranded gel (26 kPa). Thi s is ascribed to the la rge d iffere nce in concentrat ions requi red for pro pe r gel fo rmation (Stading and Her man sso n , 199 1). Form ation of th e open w id e-pore part icle gel occurs already at a concen tration of I % protein , whereas for mation of the fine stranded gel requires a protein concentration of about 10 %. Below this conce ntration , a visco us system is obta in ed . Co nseq uen tl y, when a fin e stranded gel is mad e close to its critical gelat ion co nce ntration, it is soft and has a low breaking st ress. Ho wever, when the p rotein co ncent ra tion is wel l above the criti cal co nce ntration for gela ti on, the hom oge neous fin e st ra nded gels wi ll be stronger than th e inhomogeneous particle gels, because regio ns o f low protein concentration in the inhomoge neous gels will ac t as weak points , resu lting in a low breaking stress.
Th e same ph eno me non is observed with th e 20% ovalbu min gels far above the c riti cal gel concentration for bo th gel types. Unde r th ose ci rcu mstances, the b reak ing st ress for the fine st ra nded ho moge neous gel ::.1 pH I 0 (600 kPa) is considerab ly highe r than that of the particle gel at pH 5 (20 kPa). From th ese data, it is apparen t that th e increase in protein co nce ntration does not strongly affect th e breaking stress of th e parti cle gel, but has a very great innucnce on the ultimate property of the fine stran ded gel. This st ru c tural in fo rm ation is hig hly relevant in understanding such func ti o nal prope rt ies like water-binding and melting behavior of protein ge ls.
Marga rine and butt e r Products like margarine and butte r con ta in, apa rt from oi l and fat , about 20% wate r whi c h is prese nt as fin ely dispersed drop le ts whic h are seve ra l micromete rs in diameter. Fat sp reads containin g 80 % fat derive th ei r consistency main ly from the continuous fat ph ase rath e r th a n from the dispersed wate r ph ase. In a marga rine 1 the con ti nuous fat phase appears to be a n interconnec ted network str uctu re (see th e section Fat Cr)'stal / fat cryst al int e raction ) composed of sin gle c rysta ls and shee t-like crystal agg re gates (F ig. SO, J uri aa nse and Hee rtj e, 1988).
Butte r shows a completely different microstructure: it has a discont inuous structu re of fa t gl o bules (F ig. 51) . Thi s exam ple represent s an extreme case, in wh ic h many mil k fat glo bu les of the o ri gi nal c ream persisted during th e c hurning process. In o th er cases, depending on the rip e ning procedure of the crea m and on working co ndit io ns (Prec ht and Pete rs, 198 l a ,b) fewe r globules and large r a moun ts of int e rglob ular fa t phase    Figure 52 . Stress strain cu rves for butter and margarine, obtained from parallel plate comp re ssion. h 0 and h: height of sample before and after compression, respec ti ve ly. The product softening, which is de termi ned by the ratio umar./CJ 00 , is much more pronounced for margarine than for butter.
Hcertj e, 1988) (sec a ls o the sect ion Oi l dropl c l s under stru ctur a l c le m e nt s). Summa rizing thi s in fo rm ation on the microstructure of butter and margarine , it appears th at margari ne is composed o f a co ntinuous network of fat cry stals or fat crystal aggregates. whereas butter has a much more discontinuous struc ture, contain ing fat globules with no interac t ion or a limited int erac tion with the re st of the matrix. Thi s is renected in functional properties , suc h as ha rdn ess, sp read ab ilit y. mouth-feel, emu ls ion stab ility and sal t release of both products. By pa ra llel plate compression (F ig. 52). some of these properti es ca n be de te rmined. Th e maximum stress (umax) is a me asure of th e product hardness. The produc t softening or pl astici ty, which is determined by the ra t io of th e str ess at in fi nit y (o 00 ) and the maximum st ress. appears to be much higher for margar ine th an for butte r. It shows that , on deformation , many mo re bonds are broke n in the connected marga ri ne st ructure than in th e d iscon t inu ous butter s tru ct ure. whi ch is in accor dance wi th the obser ved microstru ct ure.

Co nclu s ions
Water. air , I ipids. proteins and polysaccharides a re the m::~in components of all food produc ts. Th ey a re often prese nt in a speci fic aggregat ion or di spers io n sta te: water as dispe rsed water dropl ets; -oil as dispersed o il droplets; -fat as crystals, crysta l aggregates. g lobu les and networks; -polysacc har ides and proteins as parti c les. strands. stra nds of panicles and networks.
Functional properties arc obta in ed by specific intera ctions be tween th e various str uc tural e lements. Those properties are stron g ly influenced by interac tio ns such as between: -the di spe rsed phase and the continuous phase. e.g .. in oi l-in -protein ge ls; -fat crys ta ls by fo rming con tinuous netwo rk s (e.g .. mar ga rin e) or weak ly-in te ract in g glob ul ar aggregates (e.g ., butte r) ; -protein stra nds by forming fi ne stranded ge ls or agg regated particle gels.
On the basis of this type of information. num ero us deve lopments have taken place in the past decade in th e design of new food produc ts and materials. In parti c ular , biop olyrnc rs have been used in low-c alori e applications as fat substitlltes. These developments will continue an drequi re a SOlJ!ld know ledge of the relation between struc ture and function and how structure can be manipulated in orde r to achi eve prope r functionality . Also , new ways of st ructuring may be env isaged in our co nt inui ng e fforts to manu fac ture hig h-quality , health y and tasty food produ cts. t\cknow lcdgmc nt The author g ratefully acknow ledges the co ntr ibution of ma ny co ll eagues in Unile ver Researc h Laboratorium. in particular the participation an d scientific discussions with P.J.M.W.L. Birker , J .C .G. Blank, A. C. Juri aanse, F.S.M. van Kl eef, P. Smits , J. Vi sser and B.