Milk protein concentrate (MPC) is a partially delactosed whole milk protein powder containing both casein and whey proteins in proportions similar to those found in milk. Production occurs through membrane filtration, where ultrafiltration retains caseins, whey proteins, colloidal minerals and fat residues, while removing part of the lactose, water and soluble salts. Next, diafiltration with water further increases the protein concentration in the MPC by allowing additional permeation of non-protein materials. Finally, evaporation (optional) and spray-drying are performed. The combination of these two filtration processes increases the protein content to up to 90% on a dry matter basis (Goulding et al., Reference Goulding, Fox and O'Mahony2020).
Cao et al. (Reference Cao, Zhang, Wu, Liu, Li and Li2015) evaluated the effects of nanofiltration and evaporation concentration technologies applied during the processing of MPC on its physicochemical properties. No significant differences were found in the chemical composition between the two processing methods; however, nanofiltration yielded better results in terms of solubility. These processing operations directly influence the techno-functional properties of MPC. Several approaches have been investigated to intentionally modify these properties, including CO₂ injection, the addition of chelating salts, enzymatic modification, high-pressure processing, among others (Cao et al., Reference Cao, Zhang, Wu, Liu, Li and Li2015; Khatkar et al., Reference Khatkar, Khatkar, Mehta, Kaur, Dhull and Prakash2023).
MPCs retain the native casein-to-whey protein ratio found in bovine milk, with caseins remaining in micellar form and whey proteins exhibiting varying degrees of denaturation depending on processing conditions. These concentrates are classified based on their protein content: formulations containing around 40% protein are considered low-protein MPCs, those with 60–70% are classified as medium-protein MPCs, and those with approximately 80% protein are considered high-protein MPCs. When the protein concentration exceeds 85%, the product is referred to as milk protein isolate. Higher protein concentrations and calcium ion activity are the causes of the low solubility and thermal stability of MPC powders (Lin et al., Reference Lin, Kelly, O'Mahony and Guinee2018; Meena et al., Reference Meena, Singh, Gupta, Borad and Parmar2018, Reference Meena, Singh and Gupta2021). Meena et al. (Reference Meena, Singh, Gupta, Borad and Parmar2018) demonstrated that the reduced solubility of MPC70 resulted mainly from its high calcium concentration and low pH, which led to protein aggregation during the ultrafiltration process and the subsequent denaturation and cross-linking of proteins during spray drying, culminating in the formation of aggregated dust particles with low solubility. Powdered dairy products rich in proteins, particularly caseins, are difficult to reconstitute in water. This difficulty persists even after long periods of rehydration, making it one of the most important attributes in the application of MPCs (Crowley et al., Reference Crowley, Desautel, Gazi, Kelly, Huppertz and O'Mahony2015).
Another crucial factor in the selection and application of MPCs is their thermal stability, which is affected by the ionic strength, salts and chemical properties of milk proteins. Therefore, the role of calcium in thermal stability and the relationship between the thermal coagulation time and pH must be considered. Therefore, adjusting the pH can influence the thermal stability of dispersions when reconstituted and subjected to high heat treatment (Sunkesula et al., Reference Sunkesula, Kommineni, Meletharayil, Marella and Metzger2021). Thermal processing during manufacturing can also trigger colour changes in the final product. According to Badin et al. (Reference Badin, Quevedo-Leon, Ibarz, Ribotta and Lespinard2021), colour parameters can be correlated with various physicochemical properties. Furfurals such as 5-hydroxymethylfurfural (HMF), 2-furaldehyde (F), 2-furyl-methylketone (FMC) and 5-methyl-2-furaldehyde (MF) are intermediate products of the Maillard reaction. They are found as free compounds (not bound to proteins) that can be used as indicators of heat-induced changes in dairy products, especially HMF (Lund et al., Reference Lund, Bechshøft, Ray and Lund2022; Pinto et al., Reference Pinto, Cunha, Mesa, Sad, Stephani, Perrone and Costa2024).
The application of MPCs in foods can provide them with superior techno-functional and rheological characteristics such as emulsification, foaming, viscosity and gelation. They also provide good nutritional properties owing to their relatively low lactose content, which may be of interest to lactose-intolerant consumers. As the protein content of the MPCs increases, lactose levels decrease. This high protein content and low lactose content make MPCs an excellent ingredient for the food industry, which requires foods rich in proteins and low in carbohydrates. The moderate taste of MPCs allows other flavours of the food to stand out; this palatability characteristic is very important for product acceptance (Khatkar et al., Reference Khatkar, Khatkar, Mehta, Kaur, Dhull and Prakash2023).
Therefore, MPCs are versatile and can be applied to a wide range of products, such as dairy desserts, nutritional drinks, ice cream, yogurt, meat products, confectionery and bakery products (McCarthy et al., Reference McCarthy, Kelly, Maher and Fenelon2014; Agarwal et al., Reference Agarwal, Beausire, Patel and Patel2015). The choice of application varies depending on the protein concentration of the MPCs (from 40 to 90%). For example, MPCs with lower protein levels can be used in cheese-related applications, whereas those with higher protein levels can be used in beverages and protein bars (Chamberland et al., Reference Chamberland, Brisson, Doyen and Pouliot2022; Dularia et al., Reference Dularia, Meena, Hossain, Khetra and Arora2023).
Although the use of MPCs has increased, there is still much to be understood regarding how the interaction of these proteins leads to a more economical relationship and better use. Physicochemical analyses combined with statistical data processing could contribute to the understanding of these issues. Based on this information, the industrial sector may be able to supply the market with products that use these proteins as technological ingredients (Ramdhan et al., Reference Ramdhan, Ching, Prakash and Bhandari2021). Technological innovations in the production of these proteins have enabled the creation of new products that can be adapted to food systems to maximize their nutritional, sensory and textural properties (Khalesi et al., Reference Khalesi, Cermeño and FitzGerald2023).
Owing to the large number of MPCs in the market, the aim of this study was to analyse their physicochemical properties statistically to provide an analysis protocol that can guide the dairy industry in terms of replacing MPCs without damaging the characteristics of the final product. Based on these results, a technological application diagram (TAD) was created to classify the samples into the following analysis groups: thermal stability, probability of browning, foaming capacity and rehydration capacity. For the industry, this analysis protocol and the application diagram provide guidance and support for selecting the MPC that best suits the intended application. They offer autonomy and clarity regarding the substitution of these products across different brands available on the market, benefiting the purchasing, supply, and research and development sectors.
Materials and methods
Samples
Nine MPC samples from different manufacturers (named 1 through 9) were analysed: samples 7 and 8 were MPC 80, and the others were MPC 70. Physical and chemical characterizations were carried out using scanning electron microscopy (SEM) analysis, pH and ionic calcium profiles, thermal stability by heat coagulation time (HCT), determination of Maillard reaction markers by high-performance liquid chromatography (HPLC), determination of the browning index (BI) by colorimetry tests, determination of the physical state of lactose by Raman spectroscopy, and foaming properties and rehydration profiles of the powders by particle size determination.
Powder reconstitution protocol
For analyses requiring prior reconstitution of the powders, the procedure was adapted from the methodologies proposed by Pinto et al. (Reference Pinto, Dias, Perrone, Oliveira and Stephani2025), Lima de Paula et al. (Reference de Paula, Teixeira, Francisquini, Stephani, Perrone, Carvalho and Oliveira2021) and Queiroz et al. (Reference Queiroz, Rezende, Perrone, Francisquini, Carvalho, Alves, Oliveira and Stephani2021). The powders were reconstituted at 10% (w/w) in distilled water using a mechanical stirrer for 30 min at 21°C and subsequently stored at 7 ± 2°C for 24 h to ensure complete rehydration.
Determination of pH, ionic calcium and HCT
The pH was measured using a pH meter (PG 1400, GEHAKA®) and ionic calcium was determined using the LAQUAtwin Ca2+ equipment (Horiba Kyoto, Japan). The HCT was carried out as described by Francisquini et al. (Reference Francisquini, Altivo, Diaz, Costa, Kharfan, Stephani and Perrone2024a) at a temperature of 155 ± 2°C. The details of the procedure are provided in the Supplementary Material.
Quantification of Maillard reaction markers and colour parameters
The presence of MR markers (HMF, F, FMC and MF) was evaluated by reversed-phase HPLC. The chromatographic conditions and sample preparation followed the method described by Pinto et al. (Reference Pinto, Cunha, Mesa, Sad, Stephani, Perrone and Costa2024) who investigated the presence of these markers in dairy samples, the supplementary material contains detailed information. Colour data were obtained using a CR-400 portable colorimeter (Konica Minolta) according to the Commission International de l'Eclairage (CIE) system. On this scale, the L* value represents the luminosity ranging from 0 (black) to 100 (white), and a* and b* correspond to the chromaticity coordinates (−a = green, +a = red, −b = blue and +b = yellow). These parameters were used to calculate the BI; the equations are described in the Supplementary Material (Pinto et al., Reference Pinto, Renhe, Viana, Perrone, Oliveira and Stephani2021; Queiroz et al., Reference Queiroz, Rezende, Perrone, Francisquini, Carvalho, Alves, Oliveira and Stephani2021). Analyses were carried out in duplicate. Statistical significance was assessed using Tukey's test (for BI) and Duncan's test (for HMF), both implemented in R software.
Raman spectroscopy
The Raman spectra were acquired on the Bruker spectrometer (RFS 100; FT-Raman Bruker®), equipped with a Ge detector employing liquid nitrogen as the cooling fluid and excited at 1064 nm by a Nd:YAG laser beam and the samples were spotted by a laser with a power of 110 mW. For each spectrum, an average of 512 scans were collected with a resolution of 4 cm−1 over a range of 200–3500 cm−1. The analytical data used were obtained from OPUS 6.0 Bruker Optik, Ettlingen®, Germany. To determine the physical state of lactose, a principal component analysis (PCA) was applied to the Raman spectra data in the region between 2800 and 3200 cm−1, using the Matlab software version 7.10.0 (R2010a). To minimize variations and normalize the data, the following pre-processing steps were applied: baseline correction, second derivative, Multiplicative Scatter Correction (MSC), normalization and mean centre, according to the methodology proposed by Queiroz et al. (Reference Queiroz, Rezende, Perrone, Francisquini, Carvalho, Alves, Oliveira and Stephani2021).
Foaming properties
Foaming properties were determined according to adaptations of the methodology proposed by Shevkani et al. (Reference Shevkani, Singh, Kaur and Rana2015). Then, 100 ml of the 10% reconstituted solution was added to a measuring cylinder (500 ml) and subjected to an O2 atmosphere for 10 s. Foam index (FI) was estimated as the difference between the height reached in the graduated cylinder after 10 s and the remaining foam height after 30 min, as shown in Figure S1 and Equation S6 (see Supplementary Material).
Particle size distribution
Particle size distribution analysis of the samples was performed using laser diffraction (Beckman Coulter LS 13 320, Miami, FL, USA) coupled with an aqueous module (Beckman Coulter, Miami, FL, USA). The particle size was measured using two methods according to the adaptations of the methodology proposed by Francisquini et al. (Reference Francisquini, Martins, Renhe, Oliveira, Stephani, Perrone and Carvalho2020).
Initially, powder samples without prior rehydration were slowly added to the analysis module filled with water at 25°C, and data were collected every 90 s at five different time points: 1.5, 3.0, 4.5, 6.0 and 7.5 min. In the second method, samples previously rehydrated at 10% (w/w) were analysed. For both approaches, measurements were performed until signal levels of 50 ± 5% were reached on the PIDS (polarization intensity differential scattering) photodetectors, and data were collected in the range of 0.04–2.000 μm. The equation used to calculate the polydispersity index (PDI) in solution is provided in the Supplementary Material, according to the methodology proposed by Danaei et al. (Reference Danaei, Dehghankhold, Ataei, Hasanzadeh Davarani, Javanmard, Dokhani, Khorasani and Mozafari2018), Pinto et al. (Reference Pinto, Dias, Perrone, Oliveira and Stephani2025) and de Paula et al. (Reference de Paula, Teixeira, Francisquini, Stephani, Perrone, Carvalho and Oliveira2021).
Scanning electron microscopy
The microstructures of the samples were analysed using SEM (Hitachi TM 3000, Hitachi Ltd, Tokyo, Japan) at 500× magnification. ImageJ 1.53e software (National Institutes of Health, USA) was used to evaluate the length of the particles obtained from the images, and the PDI was obtained according to the method described by Danaei et al., (Reference Danaei, Dehghankhold, Ataei, Hasanzadeh Davarani, Javanmard, Dokhani, Khorasani and Mozafari2018), Pinto et al. (Reference Pinto, Dias, Perrone, Oliveira and Stephani2025) and de Paula et al. (Reference de Paula, Teixeira, Francisquini, Stephani, Perrone, Carvalho and Oliveira2021). The equation applied is described in the Supplementary Material.
Statistical analysis
Statistical analyses were conducted at a significance level of 0.05 using the Tukey test with the assistance of Software R® version 4.1.2 (The R Foundation for Statistical Computing). Correlation tests were performed to evaluate the relationship between the results obtained for pH and ionic calcium. PCA was applied as a method of multivariate statistics by Matlab® (software version 7.10.0 (R2010a)); the data were pre-processed using the baseline correction and mean centre.
Results and discussion
Influence of pH, ionic calcium on HCT
The thermal stabilities of nine MPC samples rehydrated at 10% (w/w) were assessed. Samples 1 and 2 exhibited lower stability, as they coagulated more rapidly (Table 1). The pH and ionic calcium values of these samples differed significantly from those of the other samples. Sample 1 had a higher pH and lower ionic calcium value, whereas sample 2 had a lower pH and higher ionic calcium value. This suggests that there is a pH and ionic calcium range that favours the stability of these samples under the conditions in which the test was applied. Therefore, pH control after rehydration is an important factor affecting the thermal stability of the solutions.
Ionic calcium profile, pH, HCT, FI, HMF content and browning index (BI). With 1, 2, 3, 4, 5, 6 and 9 being commercial samples of MPC70 and 7, 8 MPC80
Table 1 Long description
Nine milk protein concentrate samples are compared for pH, ionic calcium, heat coagulation time, foam index, hydroxymethylfurfural, and browning index, reported as means with variability. pH ranges from 6.75 in sample 2 to 7.25 in sample 1, while ionic calcium spans 69 ppm in sample 1 to 170 ppm in sample 2, with sample 9 also high at 163 ppm. Heat coagulation time values are broadly similar across samples, from 12.6 minutes in sample 2 to 26.0 minutes in sample 3, and all are marked as not significantly different within that column. Foam index varies widely, lowest in sample 8 at 2.0 cm and highest in sample 6 at 22.0 cm, with sample 5 also high at 20.0 cm. Hydroxymethylfurfural is highest in sample 1 at 0.126 mg per 100 g; samples 6, 7, and 9 are reported below the quantification limit, and the remaining quantified values fall roughly between 0.043 and 0.108 mg per 100 g. Browning index is highest in sample 9 at 14.66 and lowest in sample 7 at 8.15, with samples 5 and 6 also low near 9.4. Letter groupings indicate statistical comparisons within each column, and different tests were used for hydroxymethylfurfural and browning index, so significance should be interpreted per column rather than across measures.
Notes: Averages followed by the same capital letter in the same column do not differ by Tukey's test at the 5% probability level. Duncan's test was applied to the HMF results and Tukey's test was applied to the BI results. HMF quantification limit (LQ): 0.012 mg/100 g (Pinto et al., Reference Pinto, Cunha, Mesa, Sad, Stephani, Perrone and Costa2024).
In which: HCT (heat coagulation time), FI (foam index), HMF (hydroxymethylfurfural) and BI (browning index).
Dumpler and Kulozik (Reference Dumpler and Kulozik2015) evaluated the thermal stability of concentrated skim milk samples. HCT values decreased for pH values lower than 6.3 and higher than 6.7 (both values extremes in the sample population), demonstrating the dependent relationship between milk pH and thermal stability. HCT is directly linked to protein concentration and pH. Concentrated milk with 7–9% protein has a much shorter clotting time than non-concentrated milk. The pH influences the protein charge and the concentration of free calcium ions, negatively affecting protein stability due to heat treatment (de Kort et al., Reference de Kort, Minor, Snoeren, van Hooijdonk and van der Linden2012). Sunkesula et al. (Reference Sunkesula, Kommineni, Meletharayil, Marella and Metzger2021) analysed the final coagulation time of MPCs with a calcium reduction of up to 40% and pH values of 6.5, 6.7, 6.9 and 7.1 and showed that the thermal stability at pH 7.1 is the lowest, tripling with a 30% reduction in Ca2+. At pH values 6.5–6.9, a 30% reduction in calcium content also resulted in greater thermal stability.
Determining the FI
Foaming is a characteristic that can affect technological processes and is a factor that can cause complications when applied to products in some dairy production lines. Shevkani et al. (Reference Shevkani, Singh, Kaur and Rana2015) reported that foaming properties, including foam formation and stability, are key functional attributes of protein isolates, determining their suitability for application in food systems requiring aeration and expansion, such as whipped toppings and ice cream mixes. Zhao et al. (Reference Zhao, Fan, Shao, Cheng, Wang, Jiang, Zhang and Yuan2022) further indicated that milk protein solubility significantly influences foaming capacity, as solubility is a prerequisite for both effective foaming and emulsifying properties. Therefore, the FI should be assessed based on the intended application of the MPC, as its functional performance may vary depending on the specific food system.
Table 1 shows the differences in cm between the values before and after foam stabilization. Lower values represent more stable foams, and their formation is maintained for a longer period. Sample 8 showed the least variation, indicating that the application of this product should consider its greater likelihood of foaming. On the other hand, sample 6 is less likely to cause problems associated with foaming, as it has the highest FI cm value, indicating less stable foaming compared to the other samples. Samples 8 and 6 were significantly different from the other samples.
Markers of the browning reaction and colour parameters
To monitor the progression of the Maillard reaction in dairy products, four furfural compounds, hydroxymethylfurfural (HMF), 2-furaldehyde (F), 2-furylmethylketone (FMC) and 5-methyl-2-furfural (MF), are commonly analysed. Among these, HMF is the most frequently observed, as it is formed during the intermediate stage of the reaction. In contrast, F, FMC and MF are typically generated in the advanced stages or through interconversion processes, particularly under extreme temperature conditions or during prolonged storage (Wei et al., Reference Wei, Liu, Zhang, Kou, Du, Kong, Xie and Shen2024). In this study, only HMF was detected and quantifiable quantities in the analysed samples, the values are shown in Table 1. This methodology was originally developed by Pinto et al. (Reference Pinto, Cunha, Mesa, Sad, Stephani, Perrone and Costa2024) for application in a dairy product, specifically doce de leite, and was subsequently employed in several studies involving different dairy matrices. Notable applications include the works of Celestino et al. (Reference Celestino, Sad, Silva, Cunha, Stephani and Perrone2024) and Mesa et al. (Reference Mesa, Pinto, Schwarzenbolz, Förster, Henle, Wolfschoon-Pombo, Costa, Perrone and Stephani2025) on doce de leite; Francisquini et al. (Reference Francisquini, Cunha, Diaz, Barbosa, Altivo, Alves, Brito, Stephani and Perrone2024b) on lactose-free skimmed milk powder; and Cunha et al. (Reference Cunha, Pinto, Sad, Stephani, Pombo, Costa and Perrone2025), who applied the method to yogurt analysis.
The CIE system defines colour based on three elements: luminosity or clarity, tonality, and saturation or chromaticity (Pinto et al., Reference Pinto, Renhe, Viana, Perrone, Oliveira and Stephani2021). The colour coordinates L*, a* and b* can be used to calculate other parameters that help interpret the results of colorimetric tests, such as the BI. BI is one of the most widely used indicators of sugar-containing food products. It indicates changes in colour in relation to browning and is defined as the purity of the brown colour (Pathare et al., Reference Pathare, Opara and Al-Said2013). These values are listed in Table 1.
The Maillard reaction is initiated by the condensation of a reducing sugar with an amino group (an amino acid or protein), leading to the formation of a Schiff base. This is followed by rearrangements that lead to the formation of other structures such as melanoidins (Lund et al., Reference Lund, Bechshøft, Ray and Lund2022). In MPC samples, lactose levels decrease proportionally as the protein content increases (McCarthy et al., Reference McCarthy, Kelly, Maher and Fenelon2014). As a result, the MPC80 samples (7 and 8) had a lower lactose content, that is, a lower amount of reducing sugar available to participate in the Maillard reaction; therefore, they had the lowest BI values. According to the statistics applied to the BI values, the samples were grouped as follows: Samples 7 and 8 had the lowest BI values, followed by Samples 5 and 6. Sample 9 had the highest BI value, followed by Samples 1 and 3. Considering the HMF content results, Sample 1 contained a significant amount of this marker.
According to Gaiani et al. (Reference Gaiani, Francius, El-Kirat-Chatel, Ray, Alexander and Burgain2026), the BI is commonly used to evaluate powder colour, with higher BI values indicating greater accumulation of Maillard reaction products. For powders of comparable composition but produced at pilot and industrial scales, structural modifications may influence colour coordinates, particularly lightness L. Variations are also observed in the b* coordinate, which reflects yellowish hues, likely due to a lower content of brown Maillard pigments formed during processing. Furthermore, the degradation of light-sensitive compounds such as riboflavin (vitamin B2), β-carotene and vitamin A contributes to the yellow-greenish appearance of the powders. These surface compositional changes can significantly affect colour measurements, especially the b* value and the overall BI.
Determination of the physical state of lactose
Raman spectroscopy provides information on the molecular structure. Combined with PCA, it can be applied to investigate adulterants in foods such as milk, milk powder, meat and butter (Zhou et al., Reference Zhou, Li, Wang, Wang and Yu2024).
According to the work carried out by de Paula et al. (Reference de Paula, Teixeira, Francisquini, Stephani, Perrone, Carvalho and Oliveira2021), the physical states of lactose, crystalline, amorphous or glass transition, can be determined by the profile of the bands in the region between 2800 and 3200 cm−1. According to the author, this region reflects CH stretching vibrations, which are more pronounced for carbohydrates. In the amorphous state, lactose shows two poorly defined peaks. During crystallization (glass transition), new, sharper bands appear, and fully crystallized lactose exhibits well-defined spectral bands. Queiroz et al. (Reference Queiroz, Rezende, Perrone, Francisquini, Carvalho, Alves, Oliveira and Stephani2021) also identify this region as suitable for distinguishing between amorphous and crystalline lactose.
The Raman spectra of the nine samples under study showed a lactose band profile indicative of their amorphous state, represented by the formation of two ill-defined bands in the region between 2800 and 3200 cm−1, the spectra of which can be seen in Figure S2 of the supplementary material.
The dataset presented in the PCA was used to evaluate the products at different stages of the lactose crystallization process, as shown in Fig. 1. The technique was applied considering modifications to the band profile in the characteristic region of lactose (2800–3200 cm−1), highlighting differences that could not be observed in the visual analysis of the bands.
Separation of the physical states of lactose by principal component analysis. With 1, 2, 3, 4, 5, 6 and 9 being commercial samples of MPC70 and 7, 8 MPC80.
Two distinct groups of samples were observed: group A comprising samples 7 and 8 (80% protein concentrate) and group B with samples 1, 2, 3, 4, 5 and 6 (70% protein concentrate). MPCs with 80% protein concentration had a lower lactose content and were grouped separately from the other samples, demonstrating the applicability of Raman spectroscopy associated with PCA in classifying samples in terms of protein content. Sample 9 was the only sample that did not fit into either of the two groups; this behaviour may indicate that this sample was moving to another stage in the crystallization process, such as a vitreous transition.
Relationship between powder morphology profile and particle size evolution in rehydration capacity
Morphological profile of MPC powders
Parameters such as protein and mineral content, spray drying temperature, powder storage conditions and milk heat treatment can affect the rehydration process of MPC powders (McCarthy et al., Reference McCarthy, Kelly, Maher and Fenelon2014).
Figure 2 shows the morphological analysis images for nine MPC samples and the PDI value, according to de Paula et al. (Reference de Paula, Teixeira, Francisquini, Stephani, Perrone, Carvalho and Oliveira2021); PDI values <0.4 indicate that the sample has a monodisperse characteristic with a high degree of uniformity, and PDI values >0.4 indicate a low degree of uniformity. The results show that sample 4 had more dispersed particles and a higher PDI value. Sample 1 had large individual particles and damaged fine particles, which may indicate a problem related to high outlet air temperatures in the drying process. The properties of the powders can be determined by the type of agglomeration. In sample 9, larger aggregates were formed than in the other samples; some aggregates could also be observed in sample 6. However, as the samples were of commercial origin, it was not possible to establish a relationship between the morphological profile and formulation, processing, or storage.
Images obtained by scanning electron microscopy (magnification 200×). 1, 2, 3, 4, 5, 6 and 9 are commercial samples of MPC70 and 7 and 8 MPC80.
Particle size distribution for powder samples
The particle size distribution was analysed at five different times; it was obtained by adding powder to the liquid module of the analyser. The aim of this analysis was to provide a better understanding of the rehydration process because the readings were taken simultaneously with the dissolution of the powder. A graph of the evolution of particle size over time is shown in Fig. 3. Sample 6 exhibited the most differentiated profile; it can be seen that, over time, the particle distribution curve approaches the region below 50 µm. Samples 2, 3 and 9 exhibited the particle distribution profile closest to the 100 µm region, among which sample 9 showed the largest aggregates in the morphological analysis.
To better understand the differences in the rehydration capacity of the samples, statistical analysis was performed, and the PDIs of the particles in solution were determined. These results are shown in Table 2 and were applied to the last reading time of 7.5 min.
Particle size evolution over time (1.5; 3.0; 4.5; 6.0 and 7.5 min).
Figure 3 Long description
AM 1 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 4.5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves rise from near 0 at low diameters, increase steeply between about 20 and 60, form a broad peak between about 70 and 120, then drop toward 0 by about 300 to 400. The five curves are separated around the peak region, with the highest curve near the top of the axis around the peak. AM 2 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 6.5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves stay near 0 at low diameters, then rise sharply after about 40 to 60, reaching a narrow peak around about 90 to 120, then fall back toward 0 by about 200 to 300. The curves largely overlap, with small separations near the peak. AM 3 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves rise gradually from low diameters, then rise more steeply after about 30 to 60, reaching a peak around about 100 to 150, then decrease toward 0 by about 300 to 400. The curves are close together, with slight differences around the peak and the descending side. AM 4 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves rise from near 0, increase steeply after about 40 to 70, peak around about 120 to 180, then fall toward 0 by about 400 to 600. The curves overlap closely, with small separations near the peak. AM 5 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves show a low, broad rise at smaller diameters, then a steeper rise after about 40 to 70, reaching a peak around about 120 to 200, then decreasing toward 0 by about 400 to 600. The curves are very similar, with minor differences around the peak height. AM 6 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 5.5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves rise from near 0, then form a peak centered around about 60 to 120, with visibly separated peak positions and heights among the five time curves. After the peak, all curves drop toward 0 by about 200 to 300. AM 7 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 4.5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves rise gradually at low diameters, then rise steeply after about 30 to 70, reaching a broad peak around about 100 to 200, then decreasing toward 0 by about 400 to 600. The curves overlap closely with small differences near the peak. AM 8 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves rise from near 0, increase steeply after about 30 to 60, peak around about 80 to 140, then fall toward 0 by about 200 to 300. The curves are nearly overlapping, with slight separation near the peak. AM 9 Title: Differential Volume Legend: 1.5 min, 3.0 min, 4.5 min, 6.0 min, 7.5 min. X-axis label: Part icle Diameter (um). X-axis ticks shown from 1 to 1000. Y-axis label: Volume (%). Y-axis ticks shown from 0 to 5. A vertical dashed reference line is drawn at 50 on the x-axis. All five curves rise from near 0, increase steeply after about 40 to 80, reach a peak around about 120 to 200, then decrease toward 0 by about 400 to 600. The curves are close together, with small differences around the peak and the descending side.
Particle size distribution for MPC and PDI samples in solution at the last analysis time (7.5 min). 1, 2, 3, 4, 5, 6 and 9 being commercial samples of MPC70 and 7, 8 MPC80
Table 2 Long description
Particle size distribution metrics are reported for nine MPC and PDI samples at the final measurement time, including Dv10, Dv50, Dv90 (micrometers), percent volume below 50 micrometers, and polydispersity index. Median particle size (Dv50) ranged from 47.83 micrometers in sample 6 to 100.80 micrometers in sample 2, with sample 8 also low at 53.70 micrometers. The percent volume below 50 micrometers was highest for sample 6 (52.55%) and next highest for sample 8 (46.25%), while sample 2 was lowest (21.65%). Upper-tail size (Dv90) was largest for sample 1 (237.15 micrometers) and smallest for samples 6 and 8 (about 115 micrometers). Polydispersity index ranged from 1.73 (sample 2) to 2.45 (sample 7), with sample 1 also high at 2.37. Values are means with variability shown, and letter groupings indicate which differences were statistically significant under the study’s multiple-comparison test.
Notes: Averages followed by the same uppercase letter in the same column and averages followed by the same lowercase letter in the row do not differ according to Tukey's test at a 5% probability level.
In which: (PDI) polydispersity index and (MPC) milk protein concentrates 70 and 80% protein.
The statistics applied classified the samples into three groups based on the percentage of particles smaller than 50 µm. Samples 6 and 8 had a higher percentage of particles below 50 µm and were statistically different from the other samples, indicating a higher rehydration profile than the others. They were followed by samples 5 and 7, with values statistically similar to each other and different from the others, and samples 1, 2, 3, 4 and 9 exhibited the lowest values for percentage of particles below 50 µm. Sample 1 had the highest Dv90 value, which was significantly different from the others, indicating that after 7.5 min this sample still had larger particles than the others, a potential indication of its lower rehydration capacity. Samples 6 and 8 had the lowest Dv90 and Dv50 values, suggesting that these samples showed better rehydration at the last analysis time. The PDIs calculated for the particles in solution were significantly different only for sample 7, which had the highest value. In other words, this sample showed greater variability in the size distribution of the particles dispersed in the solution. In the morphological analysis, this sample showed the lowest PDI value (Fig. 3) and minimal aggregate formation, which may have contributed to its lower rehydration capacity.
According to Crowley et al. (Reference Crowley, Desautel, Gazi, Kelly, Huppertz and O'Mahony2015), the slow rehydration of MPC powders containing 80% protein occurs due to a poorly dispersible casein fraction. This condition can be aggravated during storage, particularly at high temperatures and/or relative humidity.
Particle size distribution for rehydrated samples
The particle size distribution of the samples rehydrated at 10% (w/w) for 24 h, stored at 7 ± 2°C, was measured; the graphical behaviour is shown in Fig. 4A. Samples 1, 6 and 9 had particles in the region smaller than 1 µm, suggesting a better rehydration capacity for these powders under the conditions evaluated. Casein micelles were present in the <1 µm regions and non-agglomerated particles formed during the atomization process were observed in the >10 µm regions (Crowley et al., Reference Crowley, Desautel, Gazi, Kelly, Huppertz and O'Mahony2015).
Particle size distribution (A) and principal component analysis (B) for samples rehydrated at 10% (m/m) for 24 h at 7 ± 2°C. 1, 2, 3, 4, 5, 6 and 9 being commercial samples of MPC70 and 7 and 8 MPC80.
Figure 4 Long description
Line graph titled Differential Volume. The horizontal axis is labeled Part icle Diameter (micrometer) with tick labels 0.04, 0.06, 0.1, 0.2, 0.4, 0.6, 1, 2, 4, 6, 8, 10, 20, 40, 60, 80, 100, 200, 400, 600, 800, 1000. The vertical axis is labeled Volume (percent) with tick labels from 0 to 7. A legend lists 1.10 percent, 2.10 percent, 3.10 percent, 4.10 percent, 5.10 percent, 6.10 percent, 7.10 percent, 8.10 percent, 9.10 percent. A vertical dashed line is drawn at particle diameter 1 micrometer. Curves are plotted for the legend entries. Text labels near the left–side peaks include 6.10 percent, 1.10 percent and 9.10 percent. The 6.10 percent curve has a peak near particle diameter about 0.1 micrometer at about 6.8 on the volume axis. The 1.10 percent curve has a peak near particle diameter about 0.2 micrometer at about 1.0 on the volume axis. The 9.10 percent curve has a peak near particle diameter about 0.1 micrometer at about 0.5 on the volume axis. Multiple curves rise between particle diameter about 6 micrometers and 200 micrometers, with several peaks between about 40 micrometers and 120 micrometers reaching about 4 to 6 on the volume axis. Several curves drop to near 0 between about 100 micrometers and 200 micrometers. Scatter plot with the horizontal axis labeled Scores on PC 1 (51.53 percent) and the vertical axis labeled Scores on PC 2 (18.07 percent). The horizontal axis shows tick labels –2000, –1500, –1000, –500, 0, 500, 1000, 1500, 2000. The vertical axis shows tick labels –1000, –500, 0, 500, 1000. A dashed horizontal line crosses at 0 on the vertical axis and a dashed vertical line crosses at 0 on the horizontal axis. Nine numbered points are shown: 1 near (about 600, about –100), 2 near (about –900, about 300), 3 near (about –600, about 100), 4 near (about –500, about 350), 5 near (about –400, about –250), 6 near (about 1800, about –150), 7 near (about –200, about 500), 8 near (about –900, about –200), 9 near (about –300, about –700). A closed outline surrounds points 2, 3, 4, 5, 7 and 8. A large dashed oval outline encloses the plotted area.
The principal component graph is shown in Fig. 4B. Samples 2, 3, 4, 5, 7 and 8 were grouped together, suggesting that they had similar profiles in terms of rehydration capacity in this experimental model. Sample 6 was located in a more distant region, indicating a greater difference in relation to the rehydration profile.
Among the samples analysed, sample 6 showed outstanding results compared to the other samples, both in the morphological analysis and particle size distribution evaluation. In the particle size distribution of the powder in solution, this sample showed the highest percentage of particles <50 µm at the final time of 7.5 min. In the 10% (m/m) rehydrated solution, this sample had the highest number of particles in the region of less than 1 µm, morphologically presented aggregates and had one of the highest PDI values, demonstrating the heterogeneous characteristics of the powder. Based on these results, we conclude that sample 6 has the highest rehydration capacity. Samples 2, 3 and 4 may have lower rehydration capacities as per the results; they also have a lower percentage of particles below 50 µm. In the PCA for samples rehydrated at 10%, these samples are in the same group. Crowley et al. (Reference Crowley, Desautel, Gazi, Kelly, Huppertz and O'Mahony2015) evaluated the particle size distribution after two different rehydration times, 90 min and 24 h, confirming the presence of large particles and difficulty in solubilization even after considerable periods of rehydration. This behaviour was observed for MPCs with protein contents greater than 80% and for MPC60 and 70.
Development of the TAD for dairy products
To create the TAD, all analyses were divided into four groups. The first group included analyses of pH, ionic calcium and HCT, referring to the thermal stability of the samples. The second group included analyses of the Maillard reaction markers, BI and the physical state of lactose; the determination of the physical state of lactose was integrated into this group because with the transition from the amorphous to the crystalline state, the lactose molecule becomes more mobile and more available to participate in browning reactions. In the third group, we considered the stability of the foam and included only an analysis to determine the FI. In the last group, the relevant criterion was the rehydration capacity, including morphological profile analyses using SEM and particle size determination experiments.
Statistical treatment was applied to each analysis; based on this result, a score was assigned to each sample for each analysis. If the sample was statistically different from the others, demonstrating superior characteristics to the others, it was assigned the value ‘+1’, if it had statistically different results, but with negative performance within the analysed parameter, it was assigned the value ‘−1’, for results that did not differ statistically, the value ‘0’ was assigned. The values assigned to each analysis were added together in each group, as shown in Table 3, making it possible to classify the samples within the TAD, respecting the following determination:
Scores attributed to each analysis group
Table 3 Long description
Nine samples are scored across four analysis groups: thermal stability (pH, ionic calcium, heat coagulation time), browning likelihood (HMF concentration, browning index, lactose state), foaming capacity (foam index), and rehydration capacity (polydispersity index and particle size distribution in powder and solution). Thermal stability scores are lowest for Sample 1 and Sample 2, while Samples 3 through 9 are neutral overall in that group. Browning scores are highest for Sample 7, followed by Sample 6 and Sample 8, whereas Sample 1 and Sample 9 have negative browning scores. Foaming capacity is negative for Samples 1, 2, 3, and 8, and neutral for Samples 4 through 9. Rehydration capacity is strongest for Sample 6, positive for Sample 8, neutral for Samples 1, 5, 7, and 9, and negative for Samples 2, 3, and 4. Most individual metric entries are neutral, so differences are driven by a small number of positive or negative values, and the scores reflect relative direction rather than absolute performance.
Note: In which: Ca2+ (ionic calcium), HCT (heat coagulation time), HMF (HMF concentration), BI (browning index), Lac (the physical state of lactose), PDI (polydispersity index), PSD (particle size distribution).
• Score < 0: Samples in which the sum has negative values in a given group are considered ‘not recommended’.
• Score = 0: Samples in which the sum is equal to zero in a given group are considered ‘recommended’.
• Score > 0: Samples in which the sum has positive values in a given group are considered ‘highly recommended’.
For the first group, better thermal stability and the Tukey test were used for the three analyses. For example, samples 1 and 2 showed statistically different results for pH and sample 1 for ionic calcium (Table 1), negatively influencing thermal stability, as discussed above, so a value of ‘−1’ was assigned. The HCT values showed no statistically significant differences, so this analysis was assigned a value of ‘0’ for all the samples. In the ‘Score’ column we have the final value that evaluates the samples within this group, with samples 1 and 2 having values lower than zero or ‘not recommended’. In other words, for technological applications where thermal stability is critical, these MPCs should not be the first option.
In the second group, with a lower probability of browning according to the results shown in Table 1, Duncan's test was applied to analyse the determination of HMF, sample 1 showed the highest value of this analyte and was assigned the value ‘−1’ and samples 6, 7 and 9 did not detect HMF, which is a positive characteristic in relation to the others, so +1 was assigned. For BI, sample 9 was statistically different with the highest BI value, and this had the greatest possibility of browning, which is a reason why it was assigned ‘−1’, as this is an unfavourable condition for the group analysed. Also, in group 2, the physical state of the lactose was considered, as demonstrated in Fig. 2, and ‘+1’ was assigned to samples 7 and 8, MPC80 with the lowest lactose content. In other words, a lower carbohydrate content was available for the Maillard reaction. A value of ‘−1’ was assigned to sample 9, which, according to the PCA graphical analysis, is isolated from the others, suggesting a possible transition from an amorphous to a crystalline state. In effect, sample 9 had a greater chance of gaining mobility and participating in the browning reactions. As a result of the sum of these assigned values, we can conclude that samples 1 and 9 obtained a sum of less than 0; therefore, they are not good options for applications in which a change in colour would be detrimental to the final product, for example, in yoghurt processing.
For group 3, only the foam determination test was considered; therefore, sample 8 was the one with the most stable foam and was assigned a value of ‘−1’ along with samples 2, 3 and 8, which according to the Tukey test (Table 1) are statistically similar to it. Consequently, these samples are not recommended for technological processes where foaming can be an obstacle to production.
Group 4 evaluated the most important characteristics of the MPCs and their rehydration capacity. Three tests were considered, the first was the morphological profile through the PDI (Fig. 3), for which all the samples were given ‘0’, as the PDI results were very varied and it was not possible to conclude on the impact of this variable on the criterion evaluated within the group. The second analysis considered the particle-size distribution of the powder, which was evaluated over five times. For this, a value of ‘+1’ was assigned to statistically similar samples with a higher percentage of particles smaller than 50 µm, following the grouping obtained by the Tukey test, shown in Table 2. Finally, the particle size distributions of the samples reconstituted at 10% were evaluated. According to the PCA graph and rehydration profile (Fig. 5), values ‘0’ were assigned to samples 2, 3, 4, 5, 7 and 8, which were grouped together by the PCA, and ‘+1’ for samples 1, 6 and 9, which were not grouped together, but showed a rehydration profile with more particles in the region smaller than 1 µm, indicating better solubility. Therefore, by evaluating the sum of this group, we can consider that samples 6 and 8 are highly recommended (sum greater than 0) for technological processes where the rehydration capacity is extremely important.
Technological application diagram for milk protein concentrate substitution in in dairy products.
From Table 3, the DAT was created, as shown in Fig. 5, in which the samples were distributed in four quadrants representing the four groups: thermal stability, less stable foaming capacity, less contribution to browning reactions and better rehydration capacity. The samples included in the smallest circle (red) scored less than zero or ‘not recommended’, the samples grouped in the middle circle scored zero or ‘recommended’ and the samples in the outermost circle (blue) scored more than zero or ‘strongly recommended’.
The application diagram presented brings together the results obtained through the physicochemical techniques applied in conjunction with statistical tools, relating them to some of the main requirements for the technological application of MPCs. Its purpose is to simplify the presentation of the results obtained from the various analyses and statistics applied. It is a tool that makes it easy to visualize and understand the possible substitutions between MPCs and ascertain the MPCs that are most suitable depending on the processing situation. The developed analysis protocol can be applied to each set of MPCs available for use within an industrial dairy plant and can prevent losses, help the research and development sector and contribute to the acquisition of supplies.
Conclusions
Physicochemical analyses were complementary to sample characterization. After discussing the data, Tukey, Duncan and PCA statistical tests were applied to select parameters to group the samples in a diagram considering the thermal stability, the possibility of browning, foaming capacity and rehydration. When the results were combined, they showed three different profiles of MPCs. Sample 1 had the highest pH and particle size distribution compared with the lower values of all other parameters; sample 6 had the opposite behaviour, and all other samples had similar results. In this regard, the results obtained in this study offer guidance to the industry and provide support so that the MPC that best suits the application can be used, which could lead to improvements in the production process and avoid problems on the production line. It is important to consider that the commercial samples analysed were at different stages of shelf life, came from different suppliers and were produced on different dates. It is also important to implement the same protocol for a larger number of batches to further validate the proposed TAD.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022029926102209.
Acknowledgements
The authors express their gratitude to CNPq, FAPEMIG and CAPES for funding the research scholarships. This work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (research project nos. 140,695/2023-0, 140,709/2023-1 and MAI – DAI 403,602/2020-3).
