Research materials

MB and WB were obtained from a dairy plant (Spółdzielnia Mleczarska Rospuda in Filipów, Warmia and Mazury voivodeship, Poland) as the final products of periodic churning performed under the same industrial conditions. WB was produced from whey obtained from Gouda cheese production. MB was produced from sweet milk cream. MB and WB were collected in triplicate during the week from different batches in the amount of 1 kg (n=3).

Water was removed from MB and WB to determine its influence on the thermal characteristics of fat. Butters were clarified at 60°C; the fat phase was separated from the aqueous phase, and the fat phase was filtered through filter paper containing anhydrous sodium bisulphate roasted at a temperature of 150°C. The above procedure was applied to obtain milk butter fat (MBF) and whey butter fat (WBF). The samples of MBF and WBF were obtained from one clarification process for each sample.

Samples of MB and WB were analysed for chemical composition, FA profile, colour parameters, texture properties, and sensory attributes. Samples of MB, WB, MBF, and WBF were used in the DSC analysis.

Chemical composition analysis

The content of total solids, solids-non-fat, and water of WB and MB was determined with the FoodScan Dairy Analyser (Foss, Hillerøed, Denmark). Fat content was determined with the gravimetric method by Rose-Gottlieb according to ISO 1211 [ISO, 2010].

Fatty acid profile analysis

Extraction of milk fat was conducted by the Rose-Gottlieb method [ISO, 2010]. Fatty acid esters were prepared according to ISO 15884 [ISO, 2002]. The 7890A gas chromatograph with a flame ionisation detection (Agilent Technologies, Santa Clara, CA, USA) was used to determine the FA profile on the CP-Sil 88 column (100 m × 0.25 mm, 0.20 μm; Agilent Technologies) with a thermal gradient from 60°C (1 min) to 180°C at 5°C/min; injector and detector temperature, 225°C and 250°C, respectively; flow rate of carrier gas (helium), 0.8 mL/min; split ratio, 1:50; injection volume, 1 μL; liner, 0.4 mm. The reference material was BCR-164 anhydrous milk fat (LGC Standards, Kiełpin, Poland). The proportions of fatty acids were calculated and divided into the following groups: saturated fatty acids (SFAs), monounsaturated fatty acids (MUFAs), polyunsaturated fatty acids (PUFAs), short-chain fatty acids (SCFAs; C4–C8), medium-chain fatty acids (MCFAs; C10–C15), long-chain fatty acids (LCFAs; C16–C19), and branched-chain fatty acids (BCFAs; C13:0 i, C13:0 ai, C14:0 i, C15:0 i, C15:0 ai, C16:0 i). The following abbreviations were used in the description of fatty acid isomers: ai – anteiso, i – iso, c – cis, t – trans.

Differential scanning calorimetry analysis

The Q10 DSC analyser with a closed refrigerated cooling system (TA Instruments, New Castle, DE, USA) was calibrated with indium. The DSC analysis (sample weight: 10±1 mg; type of measuring crucibles: biconvex hermetically sealed aluminium crucible; gas: nitrogen; flow rate: 50 mL/min, reference sample: empty crucible) was divided into two consecutive and uninterrupted cycles to determine the influence of the thermal history of butter and butter fat on the melting behaviour of milk fat. In the first cycle, measurements were conducted in the following sequence: I – heating from –40°C to 95°C, II – cooling from 95°C to –40°C. The sequences from the first cycle were repeated three times in the second cycle. In each stage, temperature was changed at a rate of 10°C/min, and at the end of each stage, the samples were maintained at the final temperature for 1 min. The results of the DSC analysis were processed in the Universal Analysis 2000 program (TA Instruments). The following parameters of the identified phase transition peaks were determined according to Brożek et al. [2022b]: maximum peak temperature (T_max), phase transition onset temperature (T_onset), transition temperature range as peak width at half height (ΔT_1/2), enthalpy (ΔH), peak height at maximum (P_height), starting (P_s) and ending (P_e) points of the peak.

Colour analysis

The colour analysis was performed in the CIELab space. Three colour parameters were measured: L^*, lightness from black (0) to white (100); a^*, greenness (–) to redness (+); b^*, blueness (–) to yellowness (+) with the CR-400 Chroma Meter (Konica Minolta Sensing Americas, Inc., Ramsey, NJ, USA) calibrated against a white plate (L^*=95.42; a^*=4.89 and b^*=–2.43). The results were used to calculate chromaticity (C^*), the whiteness index (WI), and the yellowness index (YI) according to equation (1), (2), and (3), respectively.

The total difference in colour chroma (ΔC^*) between MB and WB was calculated with the use of formula (4), and the total difference in colour (ΔE) between MB and WB was calculated with the use of formula (5).

All parameters were calculated according to Pathare et al. [2013].

Texture analysis

Instrumental texture analyses were conducted with the use of the TA.XT.plus Texture Analyser (Stable Micro Systems, Godalming, United Kingdom) with Texture Exponent 32 software. Butter samples were stored in a refrigerator at 6±1°C, and texture analyses were conducted at the same temperature in the micro thermostatic chamber (Temperature Applied Sciences Ltd, West Sussex, United Kingdom) cooled with liquid nitrogen from the XL-120 cylinder (Taylor-Wharton Gas Equipment Sdn. Bhd., Shah Alam, Selangor, Malaysia). The sample was positioned centrally under a P/6 cylindrical flat probe. Butter firmness (N), consistency (N×s), cohesiveness (N), and the viscosity index (N×s) were determined in a penetration test (speed: 2 mm/s; distance: 20 mm). Firmness was measured as the maximum force needed to press the probe into the sample. Cohesiveness was described as the maximum force needed to overcome the resistance of the sample when the probe returns to its initial position. Consistency and the viscosity index were expressed by the area under the force vs time curve when the probe penetrated the sample and returned to its initial position, respectively. Butter cohesiveness and the viscosity index were described by negative values, and they were related to the direction of probe movement [Sánchez & Pérez, 2012; Tarapata et al., 2021].

Sensory evaluation

The sensory profiles of MB and WB were determined on a descriptive scale according to ISO 13299 [ISO, 2016]. The evaluation was conducted by a panel of eight trained specialists with confirmed taste sensitivity according to ISO 8586 [ISO, 2014]. The descriptors were selected by the panel. Each panellist received a product sheet and a score sheet. The following quality descriptors were considered in the evaluation: appearance (1 descriptor – colour uniformity), aroma (4 descriptors – milky, pasteurisation, nutty and cheesy aroma), consistency (3 descriptors – firmness, cohesiveness, brittleness), and taste (4 descriptors – milky, pasteurisation, nutty and cheesy). Sensory attributes were evaluated on a five-point descriptive scale (1 to 5 points). The intensity of each sensory attribute was determined (1 – absence of descriptor, 5 – very intense).

Statistical analysis

The results of fatty acid profile, colour parameters, texture properties, and sensory attributes for MB and WB were analysed with the Student’s t-test (p≤0.05). One-way ANOVA (normal distribution was verified using the Shapiro-Wilk test) was performed for the results of thermal properties, and the significance of differences between means was determined in Duncan’s test at α=0.05. The relationship between the percentage content of individual FAs in total FAs in each sample and parameters of phase transition peaks was determined by principal component analysis (PCA). Data were processed statistically in StatSoft Inc. Statistica v. 13.1 software (Tulsa, OK, USA).

Fatty acid profile

An analysis of the FA profile revealed that C14:0, C16:0, C18:0, and C18:1 Σc FAs dominated and accounted for 75.44 and 78.25 g/100 g of total FAs in WB and MB, respectively (Table 1). The remaining FAs were detected in smaller quantities. Samples of MB and WB differed significantly (p≤0.05) in the proportions of most FAs. Whey butter was significantly (p≤0.05) more abundant in C4:0, C15:0 i, C15:0 ai, C17:1, C18:0, C18:1 Σt, C18:1 Σc, C18:2, C18:3, C18:2 c9, t11, MUFAs, PUFAs, SCFAs, and LCFAs, but less abundant in C10:0, C10:1, C12:0, C13:0 i, C13:0 ai, C13:0, C14:0, C14:1, C15:0, C16:0 i, C16:0, C16:1, SFAs, and MC-FAs than MB. The FA profiles of both products were typical of MB and WB [Gómez-Mascaraque et al., 2020; Nadeem et al., 2014; Staniewski et al., 2021].

The observed differences in the FA profiles of WB and MB probably resulted from the whey origin or whey cream production technology. The production of whey cream involves a larger number of processing operations than the production of milk cream, which promotes changes in milk fat globules (MFG) and milk fat globule membrane (MFGM) components in whey cream [Brighenti et al., 2021; Jukkola & Rojas, 2017]. The FA profiles of whey depend on the cheese from which whey originates. WB was obtained from the whey of Gouda cheese, which is commonly produced in Europe. The available literature provides only reports from the studies of whey butter produced from the whey of regional cheeses or experimental products; therefore, some differences could be noticed between the FA profile of WB in the present study and the works of other authors. Çetinkaya [2021] found a higher percentage of unsaturated fatty acids (UFAs) in whey fat (from Kashar cheese) than in milk fat, but Kasapcopur et al. [2021] found no differences between whey fat (a mixture of whey from rennet-coagulated White-brined and Kashar and acid-coagulated Lor cheeses) and milk fat in the content of UFAs. Brighenti et al. [2021] reported greater differences in the size of MFG and a higher content of phospholipids in whey cream from soft cheese (Munster) than in milk cream. They observed that MFGM could be damaged during cheesemaking and whey processing operations, which can promote the formation of larger MFG aggregates. The FA profile of the MFG core differs from the FA profile of MFGM. The FA composition of MFGM phospholipids is characterised by a higher content of UFAs than the FA composition of core TAGs [Jhanwar & Ward, 2014]. According to Fauquant et al. [2005], MFGM are more abundant in UFAs, in particular PUFAs whose content is approximately six times higher than in the MFG core. Based on these studies, it can be assumed that the higher content of UFAs-rich phospholipids in WB compared to MB was responsible for the different FA profile in MB and WB, but this requires further studies.

Differential scanning calorimetry (DSC)

The shape of DSC curves differed subject to butter type, water content, and the sample’s thermal history (Figure 1, Figure 2, Figure 3). These differences affected the parameters of phase transition peaks describing both exothermic (Table 2) and endothermic (Table 3) transitions. Butter samples and the corresponding fat samples were compared to determine which phase transition peaks correspond to changes in milk fat and water. MBF and WBF were also used in the analysis to minimise differences in the fat content of samples which could have influenced the shape of DSC curves and the parameters of phase transition peaks.

Figure 1

Differential scanning calorimetry (DSC) curves of milk butter and whey butter (a) and the corresponding fats (b) with an indication of the phase transitions of water, obtained in both cycles of the DSC analysis.

Figure 2

Exothermic peaks of fat crystallisation (A, B) and water freezing (X) in milk butter and whey butter (a) and the corresponding fats (b).

Figure 3

Melting curves of milk butter (MB), whey butter (WB) (a) and the corresponding fats (MBF and WBF, respectively) (b) in cycle I and cycle II of the differential scanning calorimetry (DSC) analysis. Endothermic peaks represent: C – low-melting point fraction (LMPF) and middle-melting point fraction (MMPF) melting; D and G – LMPF melting; E – MMPF melting; F – high-melting point fraction (HMPF) melting. Peaks X denote ice melting (a).

The DSC curves plotted for MBF and WBF had a similar shape, but differed from the curves generated for MB and WB due to the presence of water in the analysed samples (Figure 1). The phase transitions of water in the butter samples (water freezing, ice melting, evaporation) induced much greater changes in heat flow than the phase transitions of fat. Phase transitions of water were visualised by large peaks in DSC curves (Figure 1, Figure 2, Figure 2). These peaks overlapped the peaks representing the phase transitions of milk fat, in particular endothermic peaks associated with fat melting (Figure 2). The parameters of crystallisation and ice melting peaks are consistent with those given in the literature [Tomaszewska-Gras, 2012].

Two exothermic peaks were observed in DSC curves in the range of –40°C to 16.5°C during cooling from 95°C to –40°C. A first peak (A) occurred in all samples between around –40°C to around 14°C, and a second peak (B) occurred between around 10°C to around 15°C (Figure 2). Peak B was not completely separated from peak A at a cooling rate of 10°C/min, therefore peaks A and B were determined jointly as a single exothermic peak (undifferentiated peaks A and B) (Table 2). Exothermic peaks (Figure 2) had similar characteristics to those reported by Smet et al. [2010]. In a mentioned study, onset temperature was determined at 15.35°C in the first crystallisation peak and at 12.07°C in the second crystallisation peak. Exothermic peaks (A–B) were associated with crystallisation that occurs within a temperature range of –40°C to 20°C [Brożek et al., 2022a, b; Hokkanen et al., 2021; Smet et al., 2010]. Truong et al. [2014] noted two peaks between 0°C to 15°C in hard milk fat fraction. A first peak within the temperature range of 0°C to 10°C and a second additional smaller peak between 10°C to 15°C.

An analysis of the exothermic peaks of milk fat crystallisation demonstrated that water removal had no significant (p>0.05) effect on T_max values in MB samples or T_max and T_onset values in WB samples (Table 2). The remaining parameters of the exothermic peaks of fat crystallisation differed significantly (p≤0.05) between MB and MBF as well as between WB and WBF. Tonset was the only parameter that did not differ significantly (p>0.05) between MBF and WBF. Values of parameters ΔH and Pheight were higher in the crystallisation peaks of MBF and WBF than in the crystallisation peaks of MB and WB, respectively, due to a higher content of milk fat.

Endothermic peaks were observed in DSC curves in the range of –40°C to 40°C, especially in the range of –10°C to 40°C, during heating within a temperature range of –40°C to 95°C (Figure 3). The DSC curves presenting fat melting parameters differed across DSC cycles. Two endothermic peaks (C and F) were noted in cycle I, whereas four endothermic peaks (D–G) that did not change in successive heating and cooling treatments were observed in cycle II. The first peak (C) with an unclear onset temperature and its endset at 16–21°C was observed only in cycle I. Peak C was associated with unseparated LMPF and MMPF. In cycle II of the DSC analysis, peak C was transformed to peaks D and E. Peak D was identified as LMPF; it did not have a clear onset temperature, and its endset temperature was determined at 13–15°C. Peak G that was not clearly separated from peak D was observed only on WBF curves within a temperature range of 10–13°C. Peaks G and D represented LMPF melting; therefore, they were considered jointly as a single peak D during the determination of endothermic phase transition peaks (Table 3). Peak E (from 13–14°C to 18–19°C) was identified as MMPF and peak F (from 18–24°C to 34–42°C) was identified as HMPF. Endothermic peaks (C–F) were identified as overlapping LMPF and MMPF peaks (C), and as separate LMPF (D and G), MMPF (E), and HMPF peaks (F) based on published data [Brożek et al., 2022a; Hokkanen et al., 2021; Kasapcopur et al., 2021; Smet et al., 2010; Staniewski et al., 2021]. Our previous study demonstrated that the characteristics of the crystallisation and melting peaks of fat from freezedried milk, cream and buttermilk were affected by the content of fat and FAs in samples [Brożek et al., 2022a]. Two fat crystallisation peaks in the DSC curves of buttermilk and milk, and three crystallisation peaks in the DSC curve of cream were reported due to a higher concentration of SFAs in cream than in buttermilk and milk. Małkowska et al. [2021] also observed three crystallisation peaks of the solid fraction of milk fat with a high content of SFAs.

The presence of water in MB and WB had a greater influence on the parameters describing the melting peaks of LMPF and MMPF than HMPF (Table 3) due to the presence of ice melting peaks overlapping LMPF and MMPF melting peaks in DSC curves (Figure 3). Ice melting peaks overlapped milk fat melting peaks C–E in the DSC curves of MB and WB; therefore, the corresponding parameters could not be determined. For this reason, only the parameters of endothermic peaks denoting HMPF melting (peak F) could be compared between cycle I and cycle II. The identification of peaks D and E in DSC curves depended on the size of the ice melting peak. Due to the higher water content in WB than in MB, ice melting peaks were larger on the DSC curve of WB, and they showed greater overlap with peaks C–E than on the DSC curve of MB (Figure 1). Tomaszewska-Gras [2012] found that ΔH of the ice melting peak and the water freezing peak were dependent on the water content of butter.

The parameters of endothermic peaks (C–F) (Figure 3) are presented in Table 3. A comparison of all products and products within each group (MB and MBF; WB and WBF) revealed that the parameters of the endothermic peaks (C–F) of butters and the corresponding fats differed significantly (p≤0.05) in most cases. In cycle I of the DSC analysis, the Pheight of peak C differed between MBF and WBF (p≤0.05). In peak C, lower values of Pheight indicate that the phase transition of MBF components was a less energy-intensive process than the phase transition of WBF components. In cycle I of the DSC analysis, the parameters describing peak F differed significantly (p≤0.05) across products (excluding Pe). In cycle I, differences (p≤0.05) in Tonset and Ps values were noted between MB and MBF and between WB and WBF, whereas no differences (p>0.05) in the values of Pe were observed between MB and WB, or in the values of Ps and Pe between MBF and WBF. The above indicates that fat type did not determine the beginning and end of HMPF melting in cycle I, unlike water which influenced the beginning of HMPF melting in cycle I.

In cycle II, peak C was transformed to LMPF (D) and MMPF peaks (E) (Figure 3). A comparison of MBF and WBF did not reveal significant (p>0.05) differences in the T_onset and P_s values of peak D (Table 3), which indicates that the type of butter fat had no effect on the onset of LMPF melting. In turn, MBF and WBF differed significantly (p≤0.05) in all parameters of peak E, which indicates that the type of butter fat determined the melting characteristics of MMPF. In cycle II of the DSC analysis, the parameters describing peak F differed (p≤0.05) between MBF and WBF (except for ΔT_1/2 and P_e). The differences (p>0.05) were found between MB and MBF (T_max, T_onset and P_s) and between WB and WBF (T_max and T_onset), which indicates that water content influenced the mentioned parameters of the HMPF peak in cycle II.

The thermal history of butters and the corresponding fats affected DSC curves showing their melting behaviour (Figure 3) and the parameters of endothermic peaks (Table 3). The greatest differences were noted between LMPF and MMPF. After the first stage of heating, cooling and reheating, the overlapping melting peaks of LMPF and MMPF (peak C) split into separate LMPF (peak D) and MMPF (peak E) peaks. Peak C was completely transformed to peak D (peaks D and G in WBF) and peak E; therefore, the corresponding parameters could not be accurately compared between cycle I and cycle II, and only a general change trend was confirmed. The applied thermal processing regime also influenced HMPF melting parameters (peak F). The endothermic peaks of MBF and WBF differed in Tonset and Ps values, and the endothermic peak of MBF differed (p≤0.05) in the value of Pe between cycle I and cycle II (Table 3). In the DSC analysis, the peak parameters of all products were less scattered in cycle II than in cycle I, which indicates that the milk fat melting process was more stable in cycle II. The transition of peak C to peaks D and E (Figure 3) and the smaller scatter of results in cycle II than in cycle I (Table 3) could be explained by the transition of fat crystals from the polymorphic form α to forms β’ and β. According to the literature, the hexagonal α form is irreversibly transformed to the more stable orthorhombic β’ and triclinic β forms during thermal processing [Fredrick et al., 2011; Hondoh & Ueno, 2016; Lopez & Ollivon, 2009; Staniewski et al., 2021; Viriato et al., 2019]. Our previous study demonstrated that thermal history influenced the parameters of phase transition peaks and fat melting peaks of cream, milk, and buttermilk [Brożek et al., 2022b]. In turn, Staniewski et al. [2020] found that thermal history affected the melting process of anhydrous milk fat.

The following variables were subjected to PCA to identify variables that were most highly correlated with MBF and WBF, as well as variables that were bound by the strongest correlations: the FA profile of samples (g/100 g total FAs) (Table 1) and the peak parameters (Table 2 and Table 3). The first two principal components, PC1 and PC2, explained 80.50% of total variance in input data, but only PC1 explained 69.78% of the variance, whereas PC2 explained only 10.72% of the variance (Figure 4). Two clusters of variables (left and right) were formed at the ends of the PC1 axis. The variables from each cluster were strongly correlated with each other.

Figure 4

Principal component analysis biplot of milk butter fat (MBF) and whey butter fat (WBF), and the contribution of the parameters of peaks C and F in cycle I and peaks A, D, E and F in cycle II of the differential scanning calorimetry (DSC), and the content fatty acids (g/100 g total fatty acids) to principal component 1 (PC1) and principal component 2 (PC2). The variables were arranged in ascending (cluster left, L) and descending (cluster right, R) order based on loadings. The L cluster contains C15:0 ai, C18:1 Σt, D-ΔT_1/2, F_II-P_height, C18:3, C18:2 c9, t11, C15:0 i, MUFA, PUFA, C18:2, C18:1 Σc, LCFA, E-P_height, A-ΔT_1/2, C18:0, F_I-P_height, C-P_height, C4:0, SCFA, C8:0, D-P_height, C6:0, F_II-ΔT_1/2. The R cluster contains C16:0, C14:1, C16:1, C14:0, F_II-T_max, C12:0, C13:0 ai, C13:0, SFA, MCFA, A-P_height, F_II-Ps, F_II-ΔH, C10:0, F_II-T_onset, C10:1, A-ΔH, D-T_max, C15:0, E-P_e, E-ΔH, F_I-ΔH, A-Ps, E-T_max, F_I-T_max, E-T_onset, A-T_max, C13:0 i, E-ΔT_1/2, C16:0 i, D-ΔH, E-P_s, D-P_e, C-P_e, C-ΔH, F_I-P_e and F_I-T_onset. Symbols I, II in the subscript of peak F denote cycles I and II of the DSC analytical sequence, respectively.

The loading vectors indicate that the variables were correlated. The greater the acute angle, the stronger the correlation between the variables in a given direction, whereas an obtuse angle points to an inverse correlation, and a straight angle indicates an absence of a correlation between variables. To facilitate the description of variables, peaks were marked with letters A and C–F, and the symbols I and II in the subscript denote cycles I and II, respectively. The cluster on the left side of the PC plot contains points representing variables with negative loadings on the PC1 axis in the range of –0.99 to –0.75 (Figure 4). These variables were higher in WBF than in MBF. The left cluster includes the following variables: C15:0 ai, C18:1 Σt, D-ΔT_1/2, F_II-P_height, C18:3, C18:2 c9, t11, C15:0 i, MUFA, PUFA, C18:2, C18:1 Σc, LCFA, E-P_height, A-ΔT_1/2, C18:0, F_I-P_height, C-P_height, C4:0, SCFA, C8:0, D-P_height, C6:0, and F_II-ΔT_1/2, which were arranged in an ascending order based on the loadings. The cluster on the right side of the PC1 axis contains points denoting variables that were bound by inverse correlations relative to the left cluster and were characterised by highly positive loadings in the range of 0.73 to 0.99. The values of these variables were higher in MBF than in WBF. The right cluster includes following variables: C16:0, C14:1, C16:1, C14:0, F_II-T_max, C12:0, C13:0 ai, C13:0, SFA, MCFA, A-P_height, F_II-P_s, F_II-ΔH, C10:0, F_II-T_onset, C10:1, A-ΔH, D-T_max, C15:0, E-P_e, E-ΔH, FI-ΔH, A-P_s, E-T_max, F_I-T_max, E-T_onset, A-T_max, C13:0 i, E-ΔT_1/2, C16:0 i, D-ΔH, E-P_s, D-P_e, C-P_e, C-ΔH, F_I-P_e, and F_I-T_onset, which were arranged in a descending order based on the loadings.

A higher content of SFAs and MCFAs, especially C12:0, C14:0 and C16:0, in MBF than in WBF increased the values of T_max, ΔH and Ps and decreased the value of ΔT_1/2 in crystallisation peak A (which indicates that MBF crystallisation was characterised by higher enthalpy, occurred at a higher temperature, and proceeded less rapidly than WBF crystallisation), whereas an increase in the content of MUFAs, PUFAs, and LCFAs, including C18:1, and, to a much lesser degree, an increase in the content of SCFAs, including C4:0, C6:0 and C8:0, induced a reverse trend (which indicates that WBF crystallisation was characterised by lower enthalpy, occurred at a lower temperature, and proceeded more rapidly than MBF crystallisation). In our previous study, we found that T_onset increased, whereas ΔT_1/2 decreased with a rise in the content of SFAs and MCFAs, in particular C14:0 and C16:0, in freeze-dried milk [Brożek et al., 2022b]. In turn, Truong et al. [2014] reported that anhydrous milk fat had a crystallisation peak in the temperature range from –5°C to 10°C. The observed peak was smaller and characterised by higher values of T_max and T_onset in the hard fraction compared to soft fraction of milk fat.

The content of SFAs and MCFAs, especially C12:0, C14:0 and C16:0, influenced the value of P_e of the melting peak corresponding to the merged LMPF and MMPF peaks (peak C); the values of T_max, ΔH and Pe of the LMPF peak (peak D); the values of T_max, T_onset, ΔH, ΔT_1/2, P_s and P_e of the MMPF peak (peak E); the values of T_max, T_onset, ΔH and P_e in cycle I; and the values of T_max, T_onset, ΔH and P_s in cycle II of the HMPF melting peak (peak F). The above implies that the maximum difference in heat flow in all endothermic peaks (based on the absolute values in cycles I and II), the enthalpy of LMPF, MMPF and HMPF (in cycles I and II), and the melting rate of MMPF were lower in WBF than MBF, whereas the beginning of MMPF (peak E) and HMPF (cycles I and II, peak F) melting and the end of melting in the overlapping peaks of LMPF and MMPF (peak C), LMPF (peak D) and MMPF (peak E) occurred at a lower temperature in WBF than MBF. The content of MUFAs, PUFAs, and LCFAs, including C18:0, and to a much smaller extent, the content of SCFAs, including C4:0, C6:0 and C8:0, induced a drop in the above parameters, but increased P_height values of all endothermic peaks, and ΔT_1/2 values of LMPF and HMPF (cycle II) melting peaks. The above implies that LMPF and HMPF (cycle II) melted more rapidly in WBF than MBF. The presence of peak G can probably be attributed to a higher content of MUFAs, PUFAs, and LCFAs, including C18:0, in WBF than in MBF. According to Staniewski et al. [2021], the FA profile of TAGs can affect the melting behaviour of milk fat, in particular the value of P_s, and it can influence the textural properties of butter. Anankanblil et al. [2018] demonstrated that the P_s value of MMPF and HMPF crystallisation and melting peaks decreased with a rise in the content of C18:0, C18:1, C18:2, and C18:3, and increased with a rise in the content of C14:0 and C16:0. Larsen et al. [2014] and Smet et al. [2010] made similar observations also in the P_s value of the LMPF peak. The saturated FA C18:0 decreased the P_s values of milk fat crystallisation and melting peaks despite the fact that it has a higher melting temperature than C18:1, C18:2, and C18:3. The above could be attributed to mammary desaturase activity which transforms C18:0 into C18:1, thus lowering melting temperature [Anankanblil et al., 2018; Larsen et al., 2014]. Buldo et al. [2013] reported a correlation between the melting temperature (P_s) of MMPF and the content of C14:1, C16:0, C18:1 Σc, C18:1 Σt, and C18:2 c9, t11. In a PCA conducted in our previous study, the content of many FAs was correlated with the parameters of crystallisation and melting peaks of fat from freeze-dried milk, including the values of T_onset of the crystallisation peak, ΔT_1/2 of the LMPF melting peak, and T_max and T_onset of the MMPF peak [Brożek et al., 2022b]. These variables were inversely corelated with the ΔT_1/2 of the crystallisation peak and the content of LCFAs, MUFAs, and PUFAs, including C18:0, C18:1, C18:2 and C18:3.

Colour

Samples of MB and WB differed significantly (p≤0.05) in the colour parameters a^*, b^*, C^*, WI and YI, but not in parameter L^* (p>0.05) (Table 4). The values of b^* and YI were higher in MB, which indicates that it was more yellow in colour than WB. Colour chroma was also higher in MB, whereas the values of parameter a^* and the whiteness index were higher in WB.

The values of L^*, a^* and b^* in MB were similar to those given in the literature [Gómez-Mascaraque et al., 2020]. In turn, WB was characterised by similar lightness, a higher contribution of greenness, and a significantly higher contribution of yellowness relative to other studies [Jinjarak et al., 2006; Kasapcopur et al., 2021]. The total colour difference between MB and WB reached 2.33 (Table 4). The detection of colour difference could be easily possible even by an inexperienced observer if the ΔE=2.0–3.5 [Dobrzańska & Cais-Sokolińska, 2014].

Texture

The textural properties of MB and WB are presented in Table 5. Milk butter and whey butter differed significantly (p≤0.05) in firmness. Internal stickiness (described by cohesiveness values) was significantly (p≤0.05) higher in MB than WB. The examined products did not differ significantly (p>0.05) in consistency or viscosity. The obtained lower values of firmness and cohesiveness for WB compared to MB are in agreement with the fatty acid profiles of the samples. In turn WB was more abundant in MUFAs and PUFAs (Table 1) which contribute to butter softness. Amal [2009] and Jinjarak et al. [2006] reported lower hardness and cohesiveness in whey butter than in sweet cream butter. In the work of Staniewski et al. [2021], the rheological properties of butter were correlated with the composition of TAGs in the fat phase, and butter firmness decreased with a rise in the content of C18:1, C18:2, and C18:3 PUFAs in TAGs.

Sensory evaluation

The sensory evaluation revealed that MB and WB differed significantly (p≤0.05) in the values of consistency descriptors: firmness, brittleness, and cohesiveness (Figure 5) as well as in the intensity of milky and nutty aroma, and cheesy taste (p≤0.05). The remaining descriptors of taste, aroma, and colour uniformity did not differ significantly (p>0.05) between the compared butters.

Figure 5

Sensory profile of milk butter (MB) and whey butter (WB).

In a study by Mallia et al. [2008], butter enriched with UFAs was characterised by higher spreadability and a somewhat less intense milky taste and pasteurisation aroma than conventional butter after 1 week of storage at 6°C. After 8 weeks of storage at 6°C, these differences, excluding spreadability, were levelled out between the samples. Jinjarak et al. [2006] found that sweet cream butter and whey butter differed significantly only in yellowness, shininess, porosity, melting rate, cardboard odour, and nutty flavour.