Microscopic water dispersion and hydrogen-bonding structures in margarine spreads with Raman hyperspectral imaging and machine learning

Abstract

Margarine, a water-in-oil (W/O) emulsion, offers advantages such as lower costs in comparison to similar products, but large amounts of saturated fats pose health risks. Reduction of saturated fat content is difficult and often leads to "oil-off," i.e., the seepage of liquid oil from the mixture, resulting in undesirable appearance and texture. Investigations into the phenomenon have often focused on morphology at the water-oil interfaces, and this work establishes Raman imaging as a powerful application for observing microscopic morphologies of W/O emulsions. We analyze morphologies of 5 distinct margarine spreads that differ in manufacturing date, formulation, and manufacturing process. More robust H-bonding in the oil phase of the emulsions co-occurred with smaller amounts of oil-off, suggesting that H-bonding interactions between emulsifier molecules, water, and crystallized fats in the lipid phase of the W/O emulsions results in an emulsion that is less susceptible to the production of oil-off.

Keywords: Fat crystals; Machine learning; Margarine; Morphology; Raman hyperspectral microscopy; Water-in-oil emulsion.