Freezing kinetics and microstructure of ice cream from high-pressure-jet processing of ice cream mix

Pasteurized skim milk, skim milk powder, 40% fat cream, and sugar were purchased from the Penn State Berkey Creamery (University Park, PA) and used to formulate a 13% milk solids-not-fat, 6% fat, and 13% sugar low-fat ice cream mix containing 32% total solids. A control containing an emulsifier (polysorbate 80 at 0.075% wt/wt) was also prepared. Skim milk and cream were heated to 43°C before adding skim milk powder and sugar and mixing until a homogeneous mix was obtained. All samples were pasteurized (71°C for 30 min) under continuous stirring at a moderate speed that did not introduce air, then homogenized in a 2-stage homogenizer (Gaulin, Lake Mills, WI) with the first stage set at 13.8 MPa and the second stage set at 3.4 MPa. Following homogenization, samples were immediately cooled to <4°C and aged for at least 24 h before HPJ processing. Non-HPJ-treated samples with (C-P80) and without (C-0) emulsifiers were considered as controls.

The mixes (4°C) were processed using a Hyperjet 94i-S pump system (Flow Internationals Corp., Kent, WA) set at the appropriate pressure: 100, 200, 300, 400, or 500 MPa. The pressure was achieved by an intensifier pump and maintained by an attenuator before the liquid sample exited the system through a diamond nozzle (10-µm inner diameter). Immediately after this nozzle, the samples were collected in a custom-made tube-in-tube heat exchanger with a counter-current flow of cold water (∼2°C), which cooled the samples to an outlet temperature of <30°C. The samples were then transferred to a refrigerator and held ∼24 h before dynamic freezing.

The total solids and fat content of each replicate mix were tested. The total solids and fat content were determined with a Smart-Trac-II (CEM Corp., Matthews, NC).

The ice cream mixes (3 L) were frozen in a Taylor batch freezer (Taylor C117 Batch Freezer, Taylor Freezers, Rockton, IL). The mixes were agitated in the freezer until reaching a temperature of −5°C. When this temperature was reached, the condenser was shut off to continue agitation without further temperature reduction. During this isothermal whipping period, samples were extracted every 3 min for 12 min (starting at time 0) and tested for overrun. Mix samples and dynamically frozen samples at 0, 6, and 12 min were collected and stored (4°C) for particle size determination and confocal scanning laser microscopy (CSLM). After 12 min, the samples were collected in pint containers and stored for 1 wk at −25°C before rheological testing, hardness, and drip-through rate determination and transmission electron microscopy (TEM).

The TEM procedure was modified from . A sample (1 mm³) was extracted at −25°C from the inner section of hardened ice cream samples using a razor blade. The samples were immediately immersed in liquid nitrogen. The samples were then transferred using precooled tweezers from liquid nitrogen into vials containing a fixative cocktail of 2.0% (wt/vol) uranyl acetate, 3.0% (vol/vol) glutaraldehyde, and 1.0% (wt/vol) osmium tetroxide in absolute methanol, which was stored in liquid nitrogen before use. These vials then went through a gradual temperature increase to allow for complete freeze substitution using the following Leica EM AFS (Leica Microsystems, Buffalo Grove, IL) program: hold at −80°C for 4 d, increase to −40°C (2°C/h), hold at −40°C for 18 h, and increase to −20°C (2°C/h). It was noted that the ice cream cubes remained floating on the top of the fixative cocktail during the initial temperature stages, so a small piece of paper towel saturated with fixative cocktail was placed over the vials to ensure that all of the sample blocks were immersed in the fixation solution. Furthermore, the samples were gently agitated twice daily to ensure even fixation. At −20°C, the fixation solution was discarded and the samples were washed with precooled methanol 3 times. The gradual infiltration with low-temperature embedding resin Lowicryl HM20 (Marvac Ltd., Halifax, NS, Canada) was completed at −20°C as follows: 1 wash with HM20: 100% methanol 1:2 for 30 min; 1 wash with HM20: 100% methanol 1:1 for 30 min; 1 wash with HM20: 100% methanol 3:1 for 30 min; and 3 changes of 100% HM20 with each wash occurring for 30 min. The samples were then polymerized with a UV light (360 nm) at −20°C for 72 h. Further polymerization was completed at 0°C for 24 h. Resin blocks were trimmed, and thin sections were cut into 80-nm thickness using a diamond knife (Diatome, Hartfield, PA) fitted on a Leica UC6 Ultramicrotome (Leica). The sections were then mounted onto 300-mesh formvar-carbon-coated copper grids (EMS, Hartfield, PA). Poststaining was completed by submersing samples into a drop of 2% aqueous uranyl acetate solution for 8 min followed by an immediate wash in ultrapure water. Then, samples were placed onto a drop of lead citrate solution (0.04%) for 5 min to complete the poststaining. Samples were viewed using a FEI Tecnai 12 BioTwin electron microscope (120 kV; FEI Company, Hillsboro, OR) and representative images were taken with a 4k Eagle CCD camera after viewing multiple fields per sample.

The average total solids and fat content for all ice cream mixes was 32.6 ± 0.6% and 6.1 ± 0.2%, respectively. No significant differences in ice cream mix density were observed except for the 400 MPa–treated (0.71 ± 0.08 g/mL) and 500 MPa–treated (0.75 ± 0.02 g/mL) ice cream mix samples, which were significantly (P < 0.05) less dense than the C-0 sample (1.07 ± 0.03 g/mL; Table 1). We have previously reported similar low densities in low-fat ice cream mix processed by HPJ at 400 MPa due to the incorporation of stable air bubbles sucked into the fluid stream as it exits the nozzle (). Samples treated at lower HPJ pressures (100–300 MPa) had an average density similar to the controls (i.e., C-0, C-P80).

The flow behavior indexes and consistency coefficients of the ice cream mixes treated at 400 and 500 MPa were significantly different from the controls and from the samples treated with lower HPJ pressures (P < 0.05, Table 1 and Figure 1), with higher consistency coefficients and smaller flow behavior indices (i.e., more viscous and more shear-thinning). These results are consistent with previously reported data in HPJ-treated low-fat ice cream mix ().

The apparent volume-moment mean diameter (D[4,3]) values were significantly larger for the 400 MPa–treated (13.06 ± 2.91 µm) and 500 MPa–treated (12.91 ± 0.77 µm) ice cream mixes compared with the controls (C-0 = 0.50 ± 0.06 µm, C-P80 = 0.51 ± 0.17 µm) (Table 1). However, reported apparent particle sizes (especially at HPJ pressures > 300 MPa) should be treated with caution because the protein–lipid complexes have poorly defined refractive indices and shapes. Therefore, it was necessary to support the particle size determinations from light scattering with direct particle observation using CSLM (Figure 2). As expected, both control mixes before freezing (C-0 and C-P80) showed a well-dispersed mixture of protein and fat (Figure 2a and b), whereas ice cream mix processed at 400 MPa and 500 MPa showed aggregation of particles in which fat and protein were present together as a complex along with small, stable air bubbles (Figure 2c and d). These results are similar to previously reported data on the 400-MPa HPJ treatment of low-fat ice cream mix (). The presence of the large protein–lipid complexes in the samples treated at 400 MPa and 500 MPa was presumably responsible for the change in rheological properties of these samples (Figure 1), whereas the presence of air was responsible for their lower densities (Table 1).

The effect of HPJ treatment on fat droplets has not previously been visualized using TEM. Transmission electron microscopy images of the control and HPJ-processed hardened ice cream samples are shown in Figure 7, Figure 8, respectively. In the C-0 sample, casein micelles appeared as almost perfectly circular dark structures, whereas small fat globules were lighter in color but maintained well-defined edges as described by . Distinguishing air cells from coalesced fat droplets was more difficult, but air cells were identified as large structures with rougher edges. In the C-0 sample, some casein micelles appeared to be adsorbed to the surface of fat globules, as they were likely acting as protein emulsifiers. The TEM of the C-0 sample was considered as a baseline for evaluating and comparing the other sample micrographs.

The C-P80 sample had more evidence of fat destabilization and aggregation than the C-0 sample. This destabilized fat network tended to congregate around air bubbles in the ice cream. observed similar features in ice cream with polysorbate 80. The destabilized fat network likely contributed to the increased hardness and decreased melting rate of the C-P80 sample (Figure 9).

The ice creams made from mix processed at 300 MPa were expected to be similar to the C-0 ice cream sample because many of the properties (e.g., particle size, apparent viscosity) studied in the current work did not significantly differ from the C-0 sample. Indeed, the microstructure seen in the TEM images of the 300 MPa–treated sample had no major differences except for the presence of some larger casein micelles (>200 nm). Similarly sized micelles were visualized by in skim milk processed at an HPJ pressure of 300 MPa.

In contrast, the microstructure of the 400 MPa–treated and 500 MPa–treated ice creams had notably different features than the C-0 sample (Figure 8, items 4 and 5). Fewer intact casein micelles were present in these samples, and dispersed protein was scattered throughout the serum of these ice creams. Smaller, less regular fat structures also seemed to be present in these samples. Again, some ambiguity was present regarding the identification of fat globules versus air cells, but CSLM of the melted ice creams (Figure 6) provided fluorescence-based identification of large coalesced fat globules. In agreement with CSLM, TEM revealed a very thick coating of protein around fat globules in the 400 MPa–treated ice cream. The protein-coated fat globules seen in the 400 MPa–treated sample and CSLM (Figure 6) were not observed in the TEM micrograph of the 500 MPa–treated sample. The disruption of casein micelles and fat droplets likely contributed to the altered rheological properties (Figure 4) and increased hardness (Figure 9).

Processing skim milk using HPJ ≥400 MPa has previously been shown to disrupt native casein micelle structure (). proposed that the disruption of micelles allowed for the release of surface-active casein protein monomers, which contributed to enhanced foamability and emulsification ability in HPJ-treated skim milk. The TEM images of 400 MPa–treated and 500 MPa–treated ice creams also showed the disruption of casein micelles and, in the 400 MPa–treated ice cream, this dispersed protein appeared to congregate around fat globules, which could aid in their stabilization (i.e., emulsification ability).

Given the ice cream made from mix processed at 400 or 500 MPa HPJ had different rheological properties, particle sizes, and microstructures after freezing than both the control sample (C-0) and the low-pressure-regimen samples (≤300 MPa), we anticipated that the finished ice creams would also have different physical properties.

Significant differences in hardness were noted between the C-P80 ice creams (3,887 ± 1.68 g) and the 200 MPa–treated (2,254 ± 603 g) and 300 MPa–treated (1,589 ± 212 g) ice creams (P < 0.05, Figure 9). The 400 MPa–treated ice creams were the only samples that had significantly (P < 0.05) higher melting rates than the C-0 ice creams (2.91 ± 0.13 and 2.17 ± 0.13 g/min, respectively); however, the 300 MPa–treated (2.83 ± 0.31 g/min), 400 MPa–treated, and 500 MPa–treated (2.61 ± 0.31 g/min) ice creams melted significantly faster than the C-P80 ice creams (1.62 ± 0.25 g/min).

By plotting ice cream hardness as a function of melting rate for the finished ice cream (Figure 9), 4 distinct sample groupings became apparent: C-0, 100 MPa–treated, and 200 MPa–treated ice creams (intermediate hardness and meltdown); C-P80 (harder and more resistant to meltdown); 300 MPa–treated ice cream (softer and less resistant to meltdown); and 400 MPa–treated and 500 MPa–treated ice creams (harder and less resistant to meltdown).

Our results for the C-P80 sample were consistent with data reported by that showed that an ice cream with added polysorbate 80 (80:20 blend of mono- and diglycerides and polysorbate 80) was both harder and more resistant to meltdown than an ice cream with no emulsifier due to the formation of a destabilized fat network.

As mentioned above, the physical properties of ice creams that were HPJ-treated at pressures ≤300 MPa did not differ from the C-0 sample (Figure 4, Figure 5), so the freezing process and development of structural attributes in the 100 MPa–treated and 200 MPa–treated ice creams were not expected to vary from those of the C-0 sample. The 300 MPa–treated ice creams had significantly lower hardness values and greater melting rates than the C-P80 ice creams. This reduction in hardness and increase in melting rates for the ice creams made from mix processed at 300 MPa were unexpected because none of the physical properties varied significantly from the C-0 samples (Figure 4, Figure 5). Therefore, other structures that were not quantified in the present work (e.g., ice crystals, air cells) likely contributed to these differences.

We hypothesized that the development of large fat–protein complexes in the HPJ would alter ice cream hardness and meltdown properties, similarly to C-P80 ice cream samples in which a matrix of destabilized fat formed. Although the 400 MPa–treated and 500 MPa–treated ice creams had similar hardness values to the C-P80 samples, their melting rates were significantly greater. With this, the large particles observed in the 400 MPa–treated and 500 MPa–treated ice creams clearly did not behave as a typical polysorbate 80–based destabilized fat network. The dispersed protein seen in the TEM images of 400 MPa–treated and 500 MPa–treated ice creams likely contributed to the increased hardness values, but the large coalesced fat globules seen in the TEM images likely failed to support the ice cream structure during melting, increasing the melting rate compared with the C-P80 sample.