Preparation and characterisation of whey protein fluid gels: The effects of shear and thermal history

Abstract With the potential use of fluid gels in structuring fluids, whey protein isolate (WPI) was used as a prerequisite to prepare fluid gels: owing to its high nutritional value and thermal stability on gelling. Fluid gels were prepared via heat-induced gelation of a 10 wt% protein solution under controlled temperature and shear. Physical properties of the resulting gel particles (e.g. size and particle–particle interactions) were found to be dependent on the combination of shear and thermal history. Discrete large aggregates (>120 μm) were obtained at low shear, with aggregate size decreasing (


Introduction
Particulate gel suspensions with particles in the micron-scale have received increasing interest in recent years, especially within the food industry where they have been applied to structure fluids (Fern andez Farr es, Moakes, & Norton, 2014;Sullo, Watson, & Norton, 2014). Further to this their textural role as fat mimetics has been recently proposed, as the presence of small, soft and smooth particles lowers the friction between oral surfaces providing enhanced fat perception (Fern andez Farr es, Douaire, & Norton, 2013). Additionally, it was shown that the rheology of full fat mayonnaise could be simulated by replacing the majority of oil droplet content with spherical agar gel particles (J. E. ).
An essential feature regarding their application is the novel rheological properties of fluid gels, since they behave in a solid-like fashion at rest, but flow above a critical value of applied stress (I. T. Norton, Jarvis, & Foster, 1999). The rheological properties can be finely controlled to meet requirements for desired applications by varying the composition (e.g. polysaccharide, protein), cross-link density (e.g. salt, concentration), particle size and/or degree of particleeparticle interactions. Both particle size and degree of interactions are directly related to the processing conditions used in production (Gabriele, Spyropoulos, & Norton, 2009;Garrec & Norton, 2012).
Fluid gels are formed when a separation process is applied to a biopolymer system undergoing its solegel transition (I. T. Norton et al., 1999). Amongst all the existing methods for the preparation of gel particles (e.g. emulsion route or biopolymer mixtures), shear induced fluid gels are typically more attractive due to relatively easy modulation of resulting properties, by varying two parameters: shear and thermal history. On the other hand, particles produced via a phase separation route (e.g. emulsion route) require additional steps; due to the transfer of produced particles into an aqueous phase, lowering efficiency.
The mechanism proposed for the formation of shear gels is based primarily on nucleation and growth (I. T. Norton et al., 1999). It has been proposed that thermodynamically driven spherical gel particles form mainly through growth of the initial gel nuclei via enrichment, when the applied shear is comparable to the time taken for polymer to diffuse through the system and order (Garrec, Guthrie, & Norton, 2013). Thus the rate of ordering is key to particle growth. Rapid ordering forces particle growth through particleeparticle structuring; this causes particles to rapidly reach sizes that become affected by the applied shear forces. Hence particle size becomes determined by the applied mechanical break-up and large (>100 mm) anisotropic particles are formed (Gabriele, Spyropoulos, & Norton, 2010).
Such changes in the microstructure led to varying flow properties, which fall in between the behaviour of colloidal particles and polymeric gels (Altmann, Cooper-White, Dunstan, & Stokes, 2004;Gabriele et al., 2010;Garrec et al., 2013). At low volume fractions (ɸ~0.2) polysaccharide fluid gels showed high viscosities and strong shear thinning behaviour as expected for highly aggregated suspensions, mainly dominated by colloidal forces. The prevention of a continuous gel network in this way leads to incomplete gelation at the particle surface. As a result, significant particleeparticle interactions can occur, reported to give rise to both yield stresses and subsequent shear thinning behaviour (Berli & Quemada, 2000;Garrec et al., 2013).
At high phase volumes where particles are in contact and so their movement becomes sterically confined, the fluid gels behave as a suspension of soft particles with rheology dependant on particle elastic modulus as well as particleeparticle interactions (Fern andez Farr es et al., 2013;Gabriele et al., 2009). The deformability of such soft gel particles, particularly at higher concentrations has been suggested as an explanation for effective volumes exceeding those characteristic of hard sphere suspensions (Adams, Frith, & Stokes, 2004).
Fat replacers based on biopolymer particles with controlled size and shape are already commercially available. Simplesse ® is a system of microparticulated, denatured protein spheres between 0.1 and 3 mm in diameter; owing to their size and morphology the protein are able to replicate organoleptic mouth feel. The protein particles are formed using a rapid heating process (up to 120 C) over a short period of time to denature the whey protein. Reducing the pH of the system below the isoelectric point increasing electrostatic repulsion, aided by aggregate blocking agents (i.e. lecithin) and high shear, led to an incomplete aggregation process resulting in discrete particles (Singer, Yamamoto, & Latella, 1988).
From an application point of view, coupled with the highly controllable rheological properties, thermal stability and formulation simplicity are equally important within industry: owing to many potential thermal processes and cost effectiveness within product production. This study therefore investigates the formation of thermally stable whey protein aggregates by applying shear gel technology to a gelling system. The research uses whey protein isolate (WPI), owing to its irreversible heat induced gelation, as a potential candidate for the preparation of fluid gels without additives. Therefore, the production of WPI fluid gels at native pH are investigated, with emphasis to the effect of shear and thermal history on particle intrinsic properties, and corresponding rheological properties thereafter.

Preparation of stock solutions
Whey protein stock solutions of 10 wt% were prepared by dispersing WPI powder in deionised water. Sodium Azide (0.02 wt %) was added to the solution to prevent bacterial growth. Solutions were then stirred for two hours at ambient temperature until completely dispersed and stored at 5 C until usage (pH remained unadjusted ca. pH 6.4).

Fluid gel preparation using a rheometer
A Malvern Kinexus Pro, stress controlled rheometer (Malvern Instruments Ltd, UK) equipped with a couette measuring system, cup diameter 27 mm and vane diameter 25 mm was used to prepare all WPI fluid gels. Aliquots of WPI solution were transferred using a syringe to the rheometer cup set to a temperature of 25 C. The geometry was then lowered and sample temperature left to equilibrate for 2 min before shear-heating profiles were undertaken. Heating profiles starting at 25 C and increasing to 80 C with a 10 min holding period were conducted at various heating rates (1, 3, 5 and 10 C min À1 ) under steady shear (200, 400, 600 and 800 s À1 ) before quiescent cooling at 3 C min À1 to 20 C. In all experiments a thermal cover was placed over the system to help prevent water loss. Samples were then removed from the geometry and stored at 5 C for 24 h before being prepared into known volume fractions and further tests carried out (as described in Section 2.7). Each sample was prepared in triplicate with apparent viscosity profiles showing the average of at least three repeats.

Static light scattering (SLS)
Particle size distributions were determined using static light scattering. A Malvern Mastersizer MS2000 (Malvern Instruments Ltd, UK) attached to a Hydro SM manual small volume sample dispersion unit was used to obtain both surface and volume weighted means, d 3,2 and d 4,3 respectfully. Particle sizes are calculated based on the Mie theory as such particles are assumed to be monodisperse homogeneous spheres. Samples were prepared by diluting gel particles in distilled water (RI ¼ 1.33) to avoid multiple scattering. Average distributions based on 3 runs were then taken in triplicate.

Optical light microscopy
An optical light microscope, Brunel SP300-fl (Brunel Microscopes Ltd, UK) fitted with an SLR camera (Canon EOS Rebel XS, DS126 191) was used to image fluid gel particles with objective lenses up to 40Â magnification (5Â, 10Â, 20Â and 40Â). Slides were prepared by firstly diluting fluid gel samples (1:4), then adding one drop onto a microscope slide (VWR, UK) and covered with a coverslip (Thickness No.1, VWR, UK).

Confocal laser scanning microscopy (CLSM)
Confocal laser scanning microscopy (CLSM) samples were prepared in the same manner as outlined in Section 2.5. Selffluorescing properties of whey protein were used in order to prevent any changes in the microstructure that could be induced through staining, or covalent tagging of the molecules. A Leica TCS-SPE confocal microscope (Leica Microsystems Ltd, UK) fitted with an argon laser was used for all CLSM analysis. Protein particles were excited at 405 nm and detected at 450e500 nm. Stacked images were obtained using 0.5 mm spacing through the gel particles under 40Â and 60Â magnification; Immersion oil (56822, SigmaeAldrich, UK) was used to bridge the gap between coverslip and objective lens. Images were again processed using a software package (ImageJ) to extract 3D images and enhance the contrast.

Rheological methods
A Protein content of 9% was calculated for gel particles based on limited denaturation (Croguennec, O'Kennedy, & Mehra, 2004). However, material response for fluid gels arises through the volume occupied by the particles; thus for all rheological testing samples were prepared in volume fractions (f FG ) from 55 to 90% on a gel particle basis. f FG were prepared by first dispersing the fluid gel particles using an overhead stirrer. Aliquots (10 ml) were removed using a syringe and added to centrifuge tubes. The free water was removed from the system using centrifugation, Sigma 3K30 (Sigma Laborzentrifugen GmbH, Germany), at 10,000 RCF for 10 min. Relative centrifugal force and duration was determined experimentally as the point at which continued water loss ceased for all fluid gels. The supernatant was removed and 100% volume fraction obtained. Gel particles were then diluted with deionised water to give desired f FG . Suspensions were further stored at 5 C for 24 h before testing. All fluid gel samples were analysed at 25 C and subjected to pre-shear (10 s À1 for 10 s) to ensure consistent loading conditions.

Dynamic oscillatory measurements
Frequency sweeps were conducted with a serrated 25 mm parallel plate and serrated bottom plate using a 1 mm gap. All frequency sweeps were obtained in controlled strain mode between 0.1 and 10 Hz. Strains were determined by amplitude sweeps and were set to within the linear viscoelastic region (LVR) for all samples tested (<1% strain).

Viscosity measurements
Thixotropic studies were undertaken using the same measuring system as described in Section 2.7.1. Fluid gel samples with f FG 90 were subjected to increasing shear rates from 0.01 to 800 s À1 before decreasing the shear rate (800e0.01 s À1 ). This method was then repeated using a 5 min recovery period in between sweeps.

Results and discussion
3.1. Formulation of WPI fluid gels 3.1.1. Effects of shear rate on the degree of aggregation Viscometry was used to study the formation of 10 wt% whey protein fluid gels at native pH (~6.4). Fig. 1 shows the changes in apparent viscosity for WPI solutions at different shear rates (200e800 s À1 ) as the temperature was increased from 25 to 80 C at 1 C min À1 and subsequently held for 10 min. On heating, the system's apparent viscosity slowly decreases (Arrhenius effect) until a critical temperature is reached, above which, a rapid increase in the viscosity is observed. The viscosity increases in a two-step fashion, which has previously been reported for other fluid gels such as k-carrageenan (Gabriele et al., 2009), where the mechanism for particle formation was of nucleation and growth. Growth of the initial nuclei has previously been reported to occur through one of two mechanisms: enrichment from the surrounding un-gelled polymer, or particleeparticle aggregation as two particles come together (coalesce) (Gabriele et al., 2009;Garrec & Norton, 2012). The formation of WPI gel particles can also be explained on the same grounds. Heat induced gelation of whey proteins is a multi-step mechanism primarily led with the unfolding of the native structure (Bauer, Hansen, & Øgendal, 1998;Cairoli, Lametti, & Bonomi, 1994;Hoffmann, Roefs, Verheul, Van Mil, & De Kruif, 1996;Sawyer, 1968). The WPI gel structure is mainly formed through the polymerisation of b-lactoglobumin, WPI's major protein constituent, which initially partially denatures whilst retaining its dimeric form. With further restructuring to the secondary structure, intermolecular b-sheets provide points of association leading to the formation of oligomers with higher molecular mass (Lef evre & Subirade, 2000) accompanied by an observed change in turbidity in accordance with Walkenstrom, Windhab, and Hermansson (1998). These then act as nuclei that grow via enrichment from unfolded neighbouring protein until a primary particle is formed. At this initial stage, the number and volume fraction of initial particles are mainly responsible for the increase in viscosity (Mudgal, Daubert, & Foegeding, 2011) (onset to the viscosity increase). As the onset of viscosity increase remains constant across all systems, it is suggested that the number and size of the initial particles is not affected by the shear applied (Fig. 1). This is in agreement with the current knowledge on protein denaturation/aggregation as the formation of the initial clusters occurs over a time scale of seconds (Morris, Watzky, & Finke, 2009). As the aggregation process continues the number of primary particles within the system rises until particleeparticle aggregation becomes dominant. This leads to the formation of large aggregates with increased phase volume. As a result the viscosity of the system increases as expected for highly flocculated suspensions (Barnes, 1994;Buscall et al., 1987;Chung et al., 2014;Goodwin & Reynolds, 1998). The application of a shear field during gelation is therefore the main factor responsible for limited aggregate growth preventing aggregation spanning across the entire volume. Correspondingly, the viscosity change throughout the solegel transition decreases with increasing shear rates from 200 to 800 s À1 , as shown in Fig. 1, as particleeparticle aggregation becomes increasingly hindered. At the later stage of the process strengthening of the matrix through covalent bonding i.e. disulphide bonds occurs until a final aggregate is attained (Le Bon, Nicolai, & Durand, 1999).
Confocal laser scanning micrographs of an individual WPI aggregate and also single particles are shown in Fig. 2a and b respectively. Fluid gels particles prepared at 200 s À1 were primarily large and highly aggregated, seeming to have formed through interconnected particles, resulting in a three-dimensional network as previously described. However, by increasing the shear rate to 800 s À1 whilst undergoing the solegel transition, it was observed that applied separation was sufficient to prevent any secondary aggregation, restricting the final aggregates to the size of the primary particles (Fig. 2b). This again reinforced the hypothesis that growth occurs via the bonding of primary particles to form larger aggregates that grow to a size dictated by the applied shear field.

Effects of heating rate
The effect of applied heating rate on the production of WPI fluid gels is depicted in Fig. 3. Here the apparent viscosity profiles of 10 wt% WPI solutions were measured on heating to 80 C at 10 C min À1 (Fig. 3a). The obtained data revealed that as the heating rate increased the overall viscosities of the systems increased when compared to those prepared at 1 C min À1 . It is suggested that a higher degree of aggregation occurs at higher heating rates, thus increasing the effective volume of the final aggregates and consequently the viscosity of the suspensions (Genovese, 2012).
Above the critical gelling concentration nucleation is no longer the rate-limiting step for aggregation, thus larger aggregates can rapidly form resulting in higher viscosities (Mudgal, Daubert, & Foegeding, 2009;Mudgal et al., 2011;Walkenstrom et al., 1998). As such, the relationship between viscosity and degree of aggregation was applied to study the rate of aggregation (Fig. 3b) in relation to both heating and shear rates, using apparent viscosity data obtained throughout the gelation process. The rate of aggregation (da/dt), was calculated as: where h i and h iÀ1 are the viscosity at time t i and t iÀ1 respectfully, and h 0 is the initial viscosity. Corresponding values of t i and t iÀ1 were determined from the linear region of the viscosity curve after the onset of viscosity had been reached. It was observed that the rate of aggregation was dependent on both the shear rate and the heating rate applied to the system. Fig. 3b clearly shows that increasing the shear rate during the solegel transition, slows the rate at which aggregates form across all heating rates studied: as the imposed shear hinders the denatured protein from coming into contact and bonding.
On the other hand, as the heating rate increases, the rate of aggregation increases (Fig. 3b), owing to the temperature dependence of aggregate growth being purely kinetic (Le Bon et al., 1999). Therefore, the number of primary particles that interact to form an aggregate is also expected to increase following an equation originally proposed by Berli, Deiber, and Añ on (1999): where N is the number of primary particles, t 0 is the initial time at which N ¼ 1, and k is the aggregation rate constant. Hence, as K increases, due to higher heating rates, the number of primary particles coming together also increases, resulting in the formation of larger aggregates. This explains the difference in viscosity profiles observed between solutions heated at 10 C min À1 and those heated at 1 C min À1 . However as suggested by Gabriele et al. (2009) increasing the applied heating rate would decrease the processing time, as a consequence the system experiences the given shear for less time, thus decreasing the likelihood that larger aggregates are not broken down.

Qualitative model for particle formation
Size distributions for WPI gel particles made at varying shear regimes and two heating rates 1 and 10 C min À1 are shown in Fig. 4a and b respectively. It was observed that fluid gels produced in this manner were mono-modal, with the exception of samples prepared at the lowest shear rate, where a small shoulder can be observed. The presence of this shoulder may be due to large aggregates becoming trapped in the proximity of the vane geometry during processing, thus avoiding the shear field (Fig. 4a). As shear Fig. 2. Diluted WPI fluid gel particles (1:4) viewed under confocal laser scanning microscopy: samples were excited at 405 nm using 40Â magnification. Particles were prepared at increased the distributions shifted towards lower particle sizes as growth during the gelation process becomes increasingly limited by the imposed shear. Distributions for particles prepared at 10 C min À1 have been shown in Fig. 4b. Again mono-modal peaks were observed; however, aggregates produced at higher heating rates were larger in size as predicted by Eq. (3).
Mean sizes for all WPI fluid gel particles obtained at varying shear and thermal histories have been shown in Fig. 5. The data obtained showed a relationship between the size of the WPI particles and shear rate employed during the process; with increasing shear forming smaller aggregates. Particle size also shifted towards higher values as the heating rate increased, for all applied shear rates, as earlier ascribed to faster aggregation kinetics and resulting increase in particleeparticle aggregation. To this end, it can be seen that particle size, and thus the degree of aggregation, is dictated by both the shear exerted, and heating rate applied to the system. This therefore shows that the rate of aggregation is a function of the two parameters and should be accounted for when considering "k" (Eq. (3)).
It is clear that the reduction in particle size as a function of applied shear rate is highly dependent on the heating rate adopted in the formation process; shown by particles made at 1 C min À1 becoming further reduced in size than those prepared at 10 C min À1 when compared at the same shear rate. This is clear evidence for a change in the aggregate formation mechanism. It is suggested that in the initial stages of aggregate formation,   relatively weak hydrophobic interactions hold the primary particles together. At this stage the shear-induced aggregate breakdown is more effective than in the later stages of formation where particles become more firmly structured through covalent bonding. Therefore, when energy is applied to the system at a faster rate (higher heating rates), aggregates are more rapidly formed that quickly becoming resistant to shear breakdown. Thus at 10 C min À1 aggregates are formed within the shear field, whereas at 1 C min À1 initial larger aggregates are formed that become subsequently broken down, resulting in smaller final sizes.
As the size of the aggregates increases through interlinking particles, system viscosity would be expected to increase in agreement with all data reported on highly aggregated suspensions (Barnes, 1994;Genovese, 2012;Goodwin & Reynolds, 1998). Fig. 6 shows the change in relative system viscosity across all shear and thermal histories calculated using Eq. (4).
Neglecting effects posed by the anisotropic morphology of the particle and subsequent orientation in flow, changes in the shear field result in the limitation of aggregation within the dispersed phase. Following the general dependence of viscosity on the dispersed phase volume fraction (Krieger & Dougherty, 1959), the viscosity of an aggregated suspension is assumed to be based primarily on the effective volume of the disperse phase, thus depending on the degree of aggregation (Barnes, 2000;Genovese, 2012). As explained earlier the effective volume fraction depends on the mechanism of aggregation leading to the number of aggregates per resulting particle. This can be clearly seen in Fig. 6 for suspensions prepared at 400, 600 and 800 s À1 , where the change in system viscosity mirrors the increase in particle size (Fig. 5). On the other hand suspensions prepared at 200 s À1 exhibit similar viscosity changes (over all heating rates), despite the large difference in particle size. It is suggested that for such large aggregates the effective volume fraction is close to the maximum packing fraction (highly concentrated regime), where flow is not only characterised by size, but also by particles moving, ''squeezing'', past each other. The bulk viscosity thus becomes a function of particle deformability and packing similarly to the behaviour of highly ''packed'' emulsions (Barnes, 1994). From this it is reasonable to assume that the deformability of the aggregate is a function of its size, with larger more aggregated particles being more deformable than smaller rigid ones (Barnes, 1994). This in turn would explain the constant viscosity observed for fluid gels made at 200 s À1 . From the above results we conclude that aggregation of self-similar primary particles to form large aggregates represents the mechanism for WPI fluid gels. Further building on the last statement, aggregate formation through this process can either occur within the shear field, or as gel formation and subsequent breakdown.

Fluid gel material properties
3.2.1. Small deformation rheology of fluid gels To further probe particle mechanical properties, WPI fluid gels were characterised under small deformation rheology. Fig. 7a shows frequency sweeps obtained for fluid gels prepared at different shear rates at a given volume fraction (f FG ) of 90%. The mechanical spectra for both gels is typical of an interconnected structure where elastic modulus (G 0 ) dominates the viscous component (G 00 ) as described by Ross-Murphy (1994), with both moduli only slightly dependant on frequency as previously reported for k-carrageenan fluid gels (Garrec & Norton, 2012). Fig. 7b shows amplitude sweeps for the same fluid gel samples. The linear viscoelastic region extends to a critical strain above which the material begins to flow. The critical strain required to initiate flow decreased as the shear rate used in the preparation of the fluid gels increased. Although WPI fluid gels made at higher heating rates are characterised by a higher degree of aggregation, the ability of the aggregates to deform may also play a role at such a high strain. Large aggregates with a more open network structure are expected to be softer, as such will deform under applied shear before any flow is induced. Deformation of soft particles could lead to a larger contact area and greater friction (Yoon, Singh, Oh, & Kong, 2005), in turn increasing their maximum strain within the linear viscoelastic region. Once flowing, particles produced at lower shear rates showed a lower dependence on strain implying greater interparticle structuring (Garrec et al., 2013). Fig. 8 shows that the elastic modulus (G 0 ) is highly dependent on particle phase volume (f FG ) where G 0 is reported for different volume fractions. Elasticity arises through the close packing of particles above a critical phase volume, 0.55. Fig. 8 clearly shows that the elasticity of the suspensions is a function of both the shear and heating rates used in the formation process. Fluid gels made at 10 C min À1 exhibit higher elastic moduli than those prepared at 1 C min À1 since higher heating rates favour particleeparticle interactions. On the other hand shear induced breakdown will limit the extent of particleeparticle interactions, which explains the lower storage modulus for particles made at 800 s À1 . Although suspension elasticity arises primarily through interactions between particles, the size of the aggregates should also be considered to negate against effects such as deformability and packing (Barnes, 1994). To this end, the extent of interactions between aggregates was further studied by comparing fluid gels of similar particle size ( Fig. 9). Although particles were of similar size (ca. 40 mm) it was observed that the two showed different elastic moduli. Again, elasticity was predominantly dictated by the level of shear applied during the formation process with fluid gels prepared at 400 s À1 having a higher elastic modulus than those made at 600 s À1 . This indicates that aggregates formed where size is limited more greatly through shear separation, interact with each other to a lower extent.

Fluid gel flow behaviour
Flow behaviours of WPI shear gels formed at 1 C min À1 are shown in Fig. 10a. Again, a comparison is made across suspensions with the same phase volume (90%). Fluid gels exhibit characteristic shear thinning behaviour typical of flocculated suspensions (Berli & Quemada, 2000;Quemada, 1998). At low shear rates systems show an apparent shear thickening until a small Newtonian plateau is reached; which increases as a function of the processing parameters. It was observed that smaller aggregates had lower viscosities than their counterparts prepared at lower shear rates. This clearly demonstrates that for the same degree of space filling, the more aggregated particles produce a higher viscosity. Viscosities at the first Newtonian plateau (0.03 s À1 for all samples) for WPI fluid gels prepared at different shear and heating rates are compared in Fig. 10b. Increasing the shear rate during production led to a decrease in the suspension viscosity at both heating rates, again due to the lower degree of particleeparticle interactions (Fig. 10b).
With increasing shear rate the viscosity decreases following shear-thinning behaviour. The nonlinearity in the systems is either due to alignment of suspended particles in the direction of flow or to the breakdown of the aggregates (Quemada, 1998). For this reason sample data is presented for both the upward and downward shear sweeps. The presence of hysteresis revealed a thixotropic behaviour arising from aggregate breakdown that led to an overall change in the phase volume; as such the viscosity decreases.
A second sweep separated by a 5 min recovery period was used to explore the reduction in viscosity with shear. After the recovery time it was observed that the second sweep followed a similar profile to the decrease in shear observed in the first sweep, shown in Fig. 11a. This suggested that no recovery of the system occurred within this time frame, inferring that the lowering in viscosity observed in the first sweep was caused by permanent damage to the aggregates: this was confirmed by SLS data that showed a shift in the size distributions from 184 to 156 mm. The degree of thixotropy (calculated as the area within the hysteresis loop) was compared across all fluid gels for different processing conditions:    9. Elastic moduli (G 0 ) plotted against volume fraction (ɸFG) for fluid gel particles of similar size (ca. 30 mm). Particles were prepared at 400 s À1 (1 C min À1 ) (C) and 600 s À1 (5 C min À1 ) (B). heating and shear rates (Fig. 11b). The evaluation of the hysteresis area revealed that the structural breakdown occurs mainly during the first ramp and is more apparent for fluid gels made at low shear rates 200 s À1 . This infers that large aggregates are more easily disrupted compared to smaller ones due to a higher probability of inherent weak spots.

Conclusions
This study has shown that fluid gels can be prepared using whey protein under controlled shear and thermal conditions. A qualitative model for the formation of WPI gel particles and their dependence on processing conditions has also been presented. This model is primarily based on the general understanding of growth and the aggregation process. Firstly, unfolded proteins interact to form primary particles. In the second stage these primary particles move into contact with each other and are held together by relatively weak forces. Ultimately, the particles become more firmly bound as result of covalent bonding, i.e. disulphide bonds, forming secondary aggregates/final particles. It is in this second stage that structural organisation can be controlled through manipulating the rate of aggregation. Fast heating rates will promote particles that interact quickly becoming rapidly resistant to shear and as such large aggregates are produced. On the other hand low heating rates or increasing shear rates lower the rate of aggregation slowing the strengthening stage, causing particles to be more affected by the flow and therefore smaller aggregates or even single resulting particles are produced. The bulk properties of the suspensions are ultimately governed by the interaction of the final aggregates once in close enough proximity (high volume fraction). The elasticity of the suspensions was shown to strongly depend on the degree of shear applied during formation with lower shear rates showing increased aggregateeaggregate interacting. The flow behaviour of the suspensions is characterized in terms of a reduced phase volume, where it is inferred that particles occupy different effective volumes as a function of the degree of aggregation. The suspensions showed marked shear thinning behaviour depending upon the breakdown of the aggregates. Finally this study has shown that whey protein can be used to prepare fluid gels that are both thermally stable and nutritionally desirable. Fluid gel viscosities at the first Newtonian plateau (0.03 s À1 ) having been prepared using varying shear rates at 1 C min À1 (-) and 10 C min À1 (▫).  11. (a) Flow curves for WPI fluid gels (prepared at 200 s À1 , 1 C min À1 ) sweep 1 (▫,-) and sweep 2 (B,C). A 5 min recovery period was held between sweeps. (Open symbols denote increasing shear rates, filled denote decreasing shear rate.). Data shows the average of three repeats. (b) Thixotropic analysis for WPI fluid gels formulated at 1 C min À1 (C,B) and 10 C min À1 (-,▫) as a function of processing shear rate (closed markers represent sweep 1 and open, sweep 2).