Comparison of the Dispersed Phase Coalescence Mechanisms in Different Tablespreads



ABSTRACT:  Temperature-induced destabilization of the dispersed phase in butter and margarine was compared by following changes in droplet size (d3,3), solid fat content (SFC), and fat crystal spatial organization in the 28–34 °C range. At 28 °C, both butter and margarine were stable, with similar d3,3 values (approximately 6 μm) and droplet size distributions. As the storage temperature was raised above 30 °C, notable droplet coalescence was observed (for example, at 32 °C d3,3 values of approximately 10 μm for butter and approximately 12 μm for margarine were obtained). Dispersed phase coalescence in butter was dominated by coagulation, with the fat crystal network-limiting droplet–droplet contact until a minimum SFC was reached (approximately 2.5%). In margarine, the rate-limiting step for coalescence was the melting of Pickering crystals present around the dispersed aqueous droplets. Unlike butter, there was no sharp change in stability at a particular temperature or critical SFC. With these differences, coalescence in butter could be modeled as a 2nd-order process and as a 1st-order process in margarine. Overall, these results demonstrated that the kinetic stability of the dispersed aqueous phase in butter and margarine depends on SFC and the spatial distribution of fat crystals within the spreads.


Common tablespreads such as butter and margarine are water-in-oil emulsions in which the dispersed aqueous phase (approximately 20%[w/w]) is kept kinetically stable by a continuous solid fat crystal and oil network (approximately 80%[w/w]). At temperatures where the solid fat content (SFC) is low (< approximately 3%), dispersed phase destabilization may occur due to droplet flocculation, coagulation, coalescence and/or sedimentation. Sedimentation (or creaming) is the macroscopic separation of the 2 phases induced by differences in density. Flocculation involves the attraction of neighboring droplets via weak colloidal interactions, resulting in loosely structured flocs. When flocculated, droplets maintain their structural integrity (Dukhin and Sjöblom 1998; Dukhin and others 2003), and may be redispersed with slight agitation (Dukhin and others 2001; Mishchuk and others 2002). If attractive forces are sufficiently strong, flocculated droplets will not be easily redispersed. This distinct step, known as coagulation, is the initial stage leading to coalescence, which is the complete fusion and subsequent droplet size increase associated with inter-droplet film thinning and rupture.

The kinetic stability of the dispersed phase in tablespreads relies on the concerted effects of the fat crystal network and interfacially active crystals, also known as Pickering crystals (Pickering 1907). These may originate as a result of surfactant solidification at the oil–water interface (for example, monoacylglycerols [MAGs]) and/or the migration of previously formed crystals toward the droplet surface (Friberg 1997). Fat crystals lacking surface activity will be more likely to form a plastic network throughout the continuous phase of the emulsion, thus encasing the dispersed phase and reducing droplet diffusion and phase separation (Lucassen-Reynders 1962). Buchheim and Dejmek (1997) found that aqueous droplets in butter were surrounded by plate-like fat crystals whereas those in low-fat spreads were covered by a thin interfacial layer of crystalline fat and/or emulsifier. Heertje and others (Juriaanse and Heertje 1988, 1993, 1998) found distinct differences between the fat crystal network structure in butter and margarine. Via cryo-SEM, in margarine, they found clearly discernible crystalline shells surrounding aqueous droplets as well as the presence of a continuous fat matrix. Conversely, in butter, lipid crystalline shells around the aqueous phase were not very prominent. Rather, its microstructure consisted of a mottled crystal network incorporating interglobular crystals and some fat globules that had survived churning and not undergone oil-in-water to water-in-oil phase inversion. Key factors influencing the stabilization efficacy of fat crystals in spreads will include the composition of the aqueous and fat phases, the fat crystal size, polymorph, and aggregation behavior as well as the SFC–temperature relationship (Hodge and Rousseau 2003, 2005; Rousseau and Hodge 2005; Macierzanka and others 2006).

Previously, our group studied the temperature-induced destabilization of 3 tablespreads in the 10 to 35 °C range and determined that a critical SFC exists for successful stabilization of the dispersed phase, though this was dependent on whether Pickering or network stabilization dominated (Rousseau and others 2003). Here, the destabilization kinetics of the dispersed phases in butter and margarine are compared, showing that SFC and fat crystal microstructure strongly influence the emulsion breakdown of these spreads.

Materials and Methods


Butter and tub margarine were purchased from a local grocery store and stored at 5 °C until required. Both spreads consisted of 80% (w/w) fat and 16% (w/w) moisture. The ingredients in butter were cream and salt. In margarine, the key ingredients were liquid canola oil, water, hydrogenated canola oil, buttermilk, soy lecithin, mono- and di-acylglycerols, and salt. No treatment was performed on the spreads prior to experimental use.

SFC and droplet size distribution

Samples were obtained by pressing thin plastic cylinders (ID = 0.5 cm, L = 2 cm) into the spreads at random locations by hand. The filled cores were then removed with tweezers and placed directly into NMR tubes (ID = 0.8 cm, L = 20 cm). Sampling was performed on refrigerated spreads (8–10 °C) and the plastic cylinders had no effect on SFC or droplet size measurements (results not shown). SFCs and droplet size distributions were determined using a Bruker Minispec Mq pulsed nuclear magnetic resonance unit (Bruker Canada, Milton, ON, Canada), equipped with a pulsed field gradient unit that allows unimodal characterization of emulsion droplet size distributions (pfg-NMR), via measurement of the restricted diffusion of water molecules within emulsion droplets (Packer and Rees 1972; van den Enden and others 1990; Li and others 1991; Balinov and others 1994; Lee and Dungan 1998). Briefly, pfg-NMR self-diffusion measurements of hydrogen nuclei take place when 2 equal field gradient pulses are applied within a standard spin-echo pulse sequence, where the intensity of the observed echo is reduced by the effect of molecular diffusion. With emulsion droplets, diffusion is restricted by the presence of droplet interfaces. Once this boundary is reached, the echo intensity is no longer reduced. Signal attenuation is recorded as a function of gradient strength and a log-normal droplet size distribution function is best-fitted to the measured data by adjusting the mean droplet diameter and standard deviation of the distribution.

It has been shown that the water droplet size distribution in tablespreads is best described with a log-normal function (Packer and Rees 1972; van den Enden and others 1990; Balinov and others 1994). A volume-weighted log-normal distribution F can be expressed as a function of droplet diameter (di) according to the following equation (Van Lent and others 2008):


The frequency distribution F(di) is characterized by 2 parameters: the volume-weighted geometric mean diameter (d3,3) and the geometric standard deviation (σg), where d3,3 is defined by (Alderliesten 1991):


Here, ni is the number of droplets of diameter di. The geometric standard deviation (σg) can be calculated as (Alderliesten 1990):


where N is the total number of droplets and d0,0 is the number-weighted geometric mean diameter and is related to d3,3 as follows (Alderliesten 1990):


Thus, droplet sizes can be reported as d3,3 or d0,0 values and the standard deviation of the distribution (σg).

The pfg-NMR field gradient strength was calibrated with CuSO4-doped water (diffusion coefficient = 1.31 × 10−9 m2/s at 5 °C). The true self-diffusion coefficient was taken as the bulk molecular diffusion of the aqueous phase of the individual spreads, separated via centrifugation. As this technique relies on the molecular movement of water molecules within droplets, it detects size increases in the droplets themselves and not the clustering of droplets, thereby differentiating between coalescence and flocculation/coagulation.

Droplet sizes were determined at 28–34 °C in 1–2 °C increments for up to 96 h at each temperature. This technique is well suited for droplet size determinations in spreads, particularly when compared to other means (for example, light scattering), as no dilution, sample preparation or pretreatment is necessary, and the optical appearance of the samples is inconsequential.


Polarized light microscopy (PLM) was used to characterize spread morphology as a function of temperature. Samples were placed on viewing slides at room temperature (approximately 23 °C) (Fisher, St. Louis, Mo., U.S.A.), gently spread with a cold spatula, and covered with a cover slip (Fisher). A 63× water-immersion objective was used for visual examination, using an upright Zeiss LSM-510 confocal laser-scanning microscope using in transmitted polarized light mode (Zeiss Inc., Toronto, Ontario, Canada). Images were processed with Photoshop CS2 (Adobe Canada, Etobicoke, Ontario, Canada). Sample temperature was controlled and adjusted using a Peltier-controlled temperature stage (model TS60 with STC200 controller, Instec, Boulder, Colo., U.S.A). The sequences presented demonstrate the typical emulsion and fat crystal network evolution seen as a function of temperature.

Statistical analyses

Triplicate analyses were performed on all SFC and droplet size measurements. Analyses of variance and posthoc tests were performed and statistical differences were considered significant at P= 0.05.

Results and Discussion

Dispersed phase destabilization

The initial droplet size distributions (DSDs) of the spreads (Figure 1) presented similar d0,0 and d3,3 values and breadths, though butter was slightly more polydisperse than margarine. The mean d0,0 values for butter and margarine were 5.10 ± 0.09 and 5.56 ± 0.08 (P > 0.05) whereas their mean d3,3 values were 6.69 ± 0.47 μm and 6.39 ± 0.15 μm (P > 0.05), respectively. Figure 2 shows the evolution in d3,3 in butter and margarine as a function of storage temperature and time. In butter (Figure 2A), there was little change in droplet size at 28 °C over time. At 30 and 32 °C, 2–3 μm increases in d3,3 occurred during the first few hours (t= 6 h), followed by a plateau. This was likely associated with the slight decrease in SFC, which allowed some droplet movement and minimal coalescence. At 32.5 °C, d3,3 values significantly increased over 48 h (P < 0.05). After 48 h at 32.5 °C and 6 h at 34 °C, droplets were no longer detected, likely due to sedimentation of the remaining dispersed phase. Though normally not a problem in spreads, with little or no solid fat at higher temperature, the dispersed phase may sediment due to the influence of gravity.

Figure 1—.

Droplet size distributions in butter (solid line) and margarine (dashed line) stored 1 h at 28 °C. The data represents the best fit of a log-normal distribution of the diffusion data from pfg-NMR.

Figure 2—.

Mean droplet size (d3,3) of dispersed aqueous phase in butter (A) and margarine (B) following storage at 28–34 °C for 96h. 28 °C (-•-); 30 °C (-○-); 31 °C (-▾-); 32 °C (-Δ-); 32.5 °C (-▪-); 34 °C (-□-).

The evolution in the d3,3 values of the dispersed phase in margarine substantially differed from that of butter (Figure 2B). At 28 °C, there was no change in d3,3 values over 96 h (P > 0.05). At temperatures ≥ 30 °C, d3,3 values remained unaffected for the first few hours, followed by an exponential growth in droplet size. At 30 °C, d3,3 values significantly increased after 96 h (P < 0.05) whereas at 31 °C, they increased significantly within 48 h, likely indicating partial breakdown of the fat crystal network at these temperatures (P < 0.05). Storage at ≥ 32 °C led to more rapid destabilization of the dispersed phase.

SFC-microstructure–droplet size relationship

The changes in d3,3 with SFC and time for butter and margarine are reported in Figure 3A and 3B, respectively. Butter SFCs ranged from approximately 7% at 28 °C to approximately 2% at 34 °C whereas margarine had a lower SFC than butter at all temperatures, varying from approximately 4.5% at 28 °C to approximately 1% at 34 °C. All equilibrium SFCs decreased in a statistically significant manner in both spreads as a function of temperature (P < 0.05).

Figure 3—.

Solid fat content and mean droplet size (d3,3) of butter (A) and margarine (B) as a function of time and temperature. 28 °C (-•-); 30 °C (-○-); 31 °C (-▾-); 32 °C (-Δ-); 32.5 °C (-▪-); 34 °C (-□-).

In butter (Figure 3A), the critical SFC for emulsion destabilization was approximately 2.5%. At higher SFCs, only slight changes in droplet sizes occurred whereas at SFCs below this critical value, the dispersed phase coalesced rapidly. Thus, at 32 °C, butter SFC reached 3.32%± 0.13% and there was only a slight change in SFC over 96 h (P > 0.05). However, at 32.5 °C, as SFC decreased from 2.68%± 0.22% to 1.85%± 0.13% over 48 h, d3,3 values increased to a level that was beyond the instrument detection limit indicating extensive coalescence of the dispersed phase. At 34 °C, with a further drop in SFC (to 1.53%± 0.03%), dispersed phase destabilization occurred within 6 h. At these SFCs, the fat crystal network was incapable of preventing coalescence and thus significant increases in d3,3 values were observed.

In margarine, some dispersed phase destabilization occurred at most temperatures (Figure 3B). Unlike butter, there was no sharp change in stability at a particular temperature or critical SFC. Rather, at each temperature ≥ 30 °C, significant increases in d3,3 with time were observed (P < 0.05), though at different rates. For example, at 30 °C, d3,3 values increased from 6.8 ± 0.05 μm at 0.5 h to 10.8 ± 1.35 μm at 96 h (P < 0.05), with SFCs decreasing from 3.0%± 0.18% to 2.7%± 0.23% (P > 0.05). At 34 °C, d3,3 values increased from 8.2 ± 0.03 μm at 0.5 h to 25.8 ± 2.99 μm at 24 h (P < 0.05) (with SFC dropping from 1.4%± 0.12% to 1.0%± 0.09 %) (P < 0.05).

PLM yielded substantial information on the role of microstructure in dispersed phase destabilization (Figure 4). Corresponding differential interference contrast (DIC) images are also shown to identify the location of the water droplets. At 28 °C, butter's aggregated fat crystals consisted of a rather uniformly distributed solid and liquid fat network (Figure 4A-i), which corresponded to an SFC of 7.21%± 0.67%. The typical fat crystals in butter were 3–5 μm in length. Dark areas corresponded to either liquid butterfat or aqueous droplets. Bright circular regions in the PLM image corresponded to fat globules that had not coalesced and phase-inverted during butter production (no water droplets can be seen in the corresponding DIC image, Figure 4A-ii). At the same temperature, margarine had a lower SFC (4.38 ± 0.37%) (Figure 4B-i, 4B-ii), and consisted of larger crystals (length approximately 5–10 μm), many of which were present at the water droplet periphery. An increase in temperature to 31 °C led to a loss in birefringence in both spreads, indicative of lower SFCs and a sparser fat crystal network compared to 28 °C (Figure 4A-iii, 4A-iv, 4B-iii, and 4B-iv, respectively). The corresponding SFCs for these samples were 4.1%± 0.18% and 2.3%± 0.15%. At 34 °C, with the spreads fully destabilized, little solid fat remained in either spread (Figure 4A-v, 4A-vi, 4B-v, and 4B-vi).

Figure 4—.

Photomicrographs of butter (A) and margarine (B) as a function of temperature. Butter: (i and ii) 28 °C, (iii and iv) 31 °C, (v and vi) 34 °C. Margarine: (i and ii) 28 °C, (iii and iv) 31 °C, (v and vi) 34 °C. Pictures (i), (iii), and (v) represent polarized light micrographs (PLM) and pictures (ii), (iv), and (vi) represent the corresponding differential interference contrast (DIC) images. The bar represents 50 μm.

Destabilization kinetics

Emulsion destabilization may be described as a process combining coagulation or flocculation followed by coalescence (Dukhin and Sjöblom 1998; Dukhin and others 2001, 2003). The main criterion distinguishing these processes is the characteristic time of the destabilization event, which is a function of many properties, notably the emulsion droplet size distribution and composition (for example, droplet surface charge and electrolyte concentration in the continuous phase). For flocculation, the characteristic time is defined as the average time taken between 2 independent collisions, and this is known as the Smoluchowski time (τs).  After a collision, a droplet doublet (or floc) may be redispersed into individual droplets or alternatively may grow in size with the approach and collision of 3rd droplet (thus forming a triplet). However, this will depend on the time required for the doublet to fragment into individual droplets (τd). If τs << τd, the doublet will grow into a larger floc, with the coagulated droplets retaining their identity or coalescing to minimize their contact area (Dukhin and others 2001). In this scenario, the rate of coalescence would be proportional to the inverse of the lifetime of the membrane separating the droplets (coalescence characteristic time, τc) (Dukhin and Sjöblom 1998).

Simovic and Prestidge (2004) confirmed the applicability of this coupled coagulation-coalescence model for micron-sized, uncharged droplets, which is a similar situation to butter and margarine where the dielectric constant of the continuous fat/oil phase is very low. Here, we adapted this model to distinguish between dispersed phase destabilization in butter and margarine and to demonstrate how fat crystals present in either the continuous phase network or as Pickering species were a controlling factor for emulsion breakdown.

Figure 5 demonstrates that differences existed in the spatial distribution of the fat crystal network in the spreads, with the presence of a more uniform crystal distribution in butter compared to margarine where more crystals were associated with the droplet interface (Figure 5A and 5B, respectively). These microstructural differences were fundamental to the different temperature-induced destabilization behaviors observed. The schematic in Figure 6 summarizes the observed dispersed phase breakdown mechanisms for butter (Figure 6A) and margarine (Figure 6B). At 28 °C, the dispersed phase in each spread was stable due to the presence of the fat crystal network, irrespective of crystal spatial distribution. In butter, as the temperature was raised, the butterfat crystals began melting and the aqueous droplets gradually began to coagulate and coalesce, with the remaining sparser fat crystal network only partially limiting movement (Figure 6A). At 32.5 °C, however, the critical SFC was reached with too little crystalline mass present to prevent extensive droplet migration, droplet–droplet collisions (coagulation), and coalescence. In this case, the rate of coalescence was much faster than the rate of coagulation (τc << τs). Hence, destabilization in butter was modeled using 2nd-order Smoluchowski kinetics (Dukhin and Sjöblom 1998; McClements 2005), where fast droplet coalescence was coupled with slow coagulation, and the rate-limiting step was the coagulation of the droplets:


where dt is the volume-weighted droplet diameter (d3,3) at time t, d0 is the initial d3,3, and k1 is the rate of droplet coagulation. Calculated rate constants and confidence intervals (Table 1) determined with Eq. 5 yielded r2 values of 0.65 to 0.99, and were proportional to the inverse of the average time between droplet collisions (τs) (Mishchuk and others 2002; Sæther and others 2004; Mishchuk 2005). As expected, the 2nd-order rate constants increased with temperature as the fat crystal network became sparser (Table 1).

Figure 5—.

PLM micrographs showing the typical fat crystal network spatial distribution in butter (A) and margarine (B). Images are 20 μm across.

Figure 6—.

Diagram of temperature-induced destabilization in butter (A) and margarine (B).

Table 1—.  Coalescence rate constants and corresponding 95% confidence intervals for butter (2nd order) and margarine (1st order).
Temperature (°C)kbutter(μm−1h−1)95% Confidence intervalkmargarine(h−1)95% Confidence interval
283.43 × 10−42.23 × 10−44.63×10−43.49 × 10−42.22 × 10−44.75 × 10−4
303.39 × 10−32.21 × 10−34.56×10−36.12 × 10−34.40 × 10−37.84 × 10−3
31---1.41 × 10−21.26 × 10−21.56 × 10−2
325.30 × 10−33.82 × 10−36.78×10−33.99 × 10−23.51 × 10−24.47 × 10−2
32.51.07 × 10−39.09 × 10−41.22×10−3---
349.69 × 10−37.73 × 10−31.17×10−25.97 × 10−25.42 × 10−26.53 × 10−2

Aqueous phase destabilization in margarine followed a different mechanism. With SFCs lower than in butter, dispersed phase destabilization was initially expected to be more rapid. However, with the interfacial crystals mostly consisting of saturated MAGs (Heertje 1993, 1998), they melted at a higher temperature than the fat crystal network, and prevented direct contact between droplets, even at low SFCs (Figure 6B). As a result, the rate of coalescence was much slower than the rate of coagulation (τc >> τs) and it was only upon melting of the interfacial crystal layer that extensive breakdown took place. This process was modeled as a 1st-order process, where the rate-limiting step for aqueous phase destabilization was the average time taken for the interfacial crystal layer to melt (Dukhin and Sjöblom 1998):


where k2 is the rate of coalescence of the aqueous droplets, which is proportional to the inverse of coalescence time (τc). As mentioned by Sæther and others (2004), Eq. 6 is not valid for the initial stages of emulsion destabilization, given the lack of change in droplet size. This is also evident from Figure 2B, where droplet size was constant prior to an exponential increase. First-order rate constants and confidence intervals (Table 1) obtained using Eq. 6 yielded r2 values ranging from 0.79 to 0.98. Similar to butter, coalescence rate constants in margarine increased with temperature.


The above results highlight the differences in the microstructure and destabilization mechanisms in butter and margarine. The dispersed phase of butter underwent gradual destabilization followed by sudden coalescence at a critical SFC. In margarine, some destabilization occurred at most temperatures (≥ 30 °C) and higher temperatures led to extensive breakdown. PLM images showed that the butterfat crystal network was more homogeneous and did not contain many crystals associated with droplet interfaces. In margarine, there was also a well-developed network; however, many droplets were coated with a crystalline layer leading to Pickering stabilization. Mechanistically, margarine was stabilized by a combination of Pickering species and the presence of a fat crystal network, resulting in a 1st-order process for droplet growth, where the rate-limiting step was the melting of the interfacial crystalline layer. In butter, there was little evidence of Pickering stabilization, and dispersed phase stability depended on the density of the fat crystal network and its capacity to prevent droplet–droplet collisions. Here, the rate of droplet destabilization was predicted by a 2nd-order process.


Financial support from the Natural Science and Engineering Research Council (NSERC) of Canada, and Ryerson Univ. (from the Office of Research Services and from the Faculty of Community Services SRC grant program) is acknowledged. Helpful discussions with Mrs. Belinda Elysée-Collen and Mrs. Atrayee Basu Ghosh is acknowledged.