Study of the synergistic effect of sodium citrate and phosphates on the colloidal stability of skim milk under controlled severe heat treatment conditions

Background Colloidal stability is crucial for milk processing at very high temperatures. Intense heat unbalances mineral and protein equilibrium, often leading to coagulation, sedimentation and quality loss. Chelating salts, such as sodium citrate and sodium phosphates, are frequently used to control ionic calcium concentration and enhance thermal stability. However, data on the combined effects of multiple salts on milk stability under severe heat treatments are still scarce. Aim This research aimed to identify optimal salt combinations to improve the stability of skim milk under high‐temperature conditions. To achieve this, the interaction between different salts was analysed to determine their impact on colloidal and thermal properties. Methods A four‐component simplex‐centroid mixture design was used to formulate skim milk from commercial milk powder with varying concentrations of chelating salts. Samples were subjected to a simulated treatment at 155°C and analysed using mixture regression models and analysis of variance (ANOVA), followed by multi‐response optimisation and response surface analysis. Major Findings Both before and after heat treatment, sodium triphosphate was the most effective salt for reducing ionic calcium levels. Conversely, sodium monophosphate proved most effective for prolonging thermal stability. The results demonstrated that the salts exhibited synergistic effects when combined, showing a clear synergy that balanced pH, free calcium and heat coagulation time. Scientific or Industrial Implications The observed synergy between salts is fundamental for maintaining the mineral and protein balance. The optimal formulation was found to be a mixture of 46.5% sodium citrate, 37.6% sodium monophosphate and 16.0% sodium triphosphate. Validation tests showed that experimental results were closely aligned with predicted values, with relative errors of less than 5%, confirming the robustness of the optimisation model.