Thermal Fingerprints: How DSC and TGA Are Exposing Food Fraud

Food fraud has become one of the most quietly expensive problems in the global supply chain, with adulterated olive oil, diluted honey, mislabeled meat, and stretched dairy products costing the industry billions each year while eroding consumer trust. A comprehensive review published in Food Chemistry: X by Amir Pourmoradian, Alireza Mohammadi, and Mohsen Barzegar of Tarbiat Modares University now makes the case that two laboratory workhorses from materials science—differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA)—are emerging as fast, cheap, and remarkably effective weapons in the fight against food adulteration. The review, which surveys literature from 2020 to 2026, argues that these thermal techniques deserve a central place in next-generation food authentication platforms.

The principle behind thermal analysis is elegantly simple. Every food has a chemical signature, and that signature dictates how it behaves under heat. DSC measures heat flow into or out of a sample relative to a reference as temperature changes, capturing melting points, crystallization temperatures, glass transitions, and protein denaturation events. TGA, by contrast, continuously records changes in sample mass under a controlled atmosphere, revealing moisture content, decomposition temperatures, and degradation pathways. Because adulteration, substitution, or raw material variation alters these thermal events—shifting melting temperatures, changing enthalpy values, or reshaping mass-loss curves—each food generates a characteristic thermal fingerprint that fraud cannot easily disguise.

What makes these techniques so attractive compared with conventional methods is practicality. Chromatographic techniques such as gas chromatography-mass spectrometry and liquid chromatography-mass spectrometry remain the gold standard for definitive molecular identification, but they demand extensive sample preparation, expensive instrumentation, solvents, and long analysis times. DSC and TGA require minimal or no sample preparation, tiny sample quantities, no solvents, and only minutes of analysis time. This aligns squarely with the principles of green analytical chemistry, and it makes the techniques well suited to high-throughput routine screening where hundreds of samples must be triaged quickly.

Edible oils illustrate the power of the approach most vividly. Premium oils such as extra virgin olive oil are among the most frequently adulterated commodities in the world, and their market value depends on botanical origin, geography, and production method. Because each oil’s triacylglycerol composition is unique, each produces a distinctive crystallization profile that adulteration measurably distorts. Recent studies reviewed by the authors show DSC combined with chemometrics detecting sunflower, corn, and soybean adulteration in extra virgin olive oil with quantitative prediction errors as low as a few percent. In flaxseed oil work, linear discriminant analysis achieved 99.5 percent classification accuracy for rapeseed adulteration, while artificial neural networks delivered quantitative predictions with a correlation coefficient of 0.996. Data fusion of DSC with GC-MS in walnut oil studies pushed support vector machine classification to 100 percent accuracy.

Honey, another fraud magnet, has proven equally amenable to thermal fingerprinting. The glass transition temperature of honey is highly sensitive to sugar composition and water content, so adding cheap syrups such as high-fructose corn syrup, rice syrup, or inverted sugar systematically lowers the glass transition and alters heat capacity changes. One research team used DSC with partial least squares regression to successfully predict adulterant concentrations in sunflower honey contaminated with five different syrups at levels of 5 to 20 percent. More strikingly, recent work has coupled DSC with artificial intelligence: convolutional neural networks boosted by synthetic minority over-sampling improved classification accuracy of adulterated stingless bee and Tualang honeys from a baseline of 24 to 67 percent up to 60 to 91 percent, while a separate study using graph-based semi-supervised learning achieved accurate classification of corn syrup adulteration with only a limited number of labeled samples.

Beyond oils and honey, the review documents thermal analysis succeeding across a remarkable range of matrices. In dairy, DSC coupled with machine learning models including gradient boosting machines and multilayer perceptrons achieved 100 percent classification accuracy in detecting formaldehyde, whey, urea, and starch adulteration in raw bovine milk. Thermogravimetry combined with chemometrics detected starch in commercial yogurt above the legal 1 percent threshold with over 83 percent accuracy. DSC has resolved species-specific thermal fingerprints in beef, pork, rabbit, and chicken for solvent-free halal authentication, and TGA with derivative thermogravimetry quantified corn, husk, and straw adulteration in ground roasted Arabica coffee with a detection limit of just 0.36 percent by mass.

The techniques also serve quality control well beyond fraud detection. DSC-derived oxidation induction times and onset temperatures correlate strongly with conventional peroxide and anisidine values, providing rapid indicators of oil deterioration during storage. Glass transition measurements have been shown to predict chemical stability: in condensed biopolymer matrices, storing lipids below the glass transition slowed oxidation by restricting molecular diffusion, suggesting DSC can forecast shelf life without months of storage trials. DSC further characterizes cocoa butter polymorphism—critical to the gloss and snap of chocolate—starch gelatinization, and protein denaturation, with recent nano-DSC work resolving the thermal transitions of six individual whey proteins by manipulating pH and ion chelation conditions.

The authors are candid about the limitations. Thermal profiles are influenced by cultivar, geography, climate, processing history, oxidation, and experimental conditions such as heating rate, sample mass, and thermal history, all of which can complicate comparisons between laboratories. Overlapping thermal events in complex matrices can obscure individual transitions, and sensitivity to low-level adulteration is limited when adulterants behave thermally like the authentic product. The review also raises a pointed methodological question the field has largely ignored: what is the minimum detectable thermal shift below which DSC- and TGA-based claims become unreliable? Without systematic detection thresholds for each food matrix, cross-study comparisons remain largely qualitative. High classification accuracies reported after machine learning integration, the authors caution, should not be mistaken for universal detection capability, particularly when models rest on small datasets without independent external validation.

The future, however, looks increasingly intelligent. The review outlines a roadmap in which deep learning extracts discriminative features directly from raw thermal curves, explainable AI techniques such as SHAP reveal which temperature ranges drive classification decisions, and transfer learning compensates for small thermal databases. Federated learning could allow laboratories and regulators to train shared models without sensitive data ever leaving its site of origin, while digital-twin frameworks could couple thermal sensors to real-time, in-line quality monitoring under Industry 4.0 infrastructure. Portable miniaturized DSC devices may eventually bring authentication to the point of collection, and hyphenated systems combining thermal analysis with FTIR, Raman, or mass spectrometry promise multimodal fingerprints capturing variations no single technique can detect.

The review’s bottom line is measured but optimistic: DSC and TGA will not replace chromatography or spectroscopy as confirmatory reference methods, but positioned as rapid, solvent-free screening tools within a complementary analytical strategy, they offer a practical route to catching fraud before it reaches consumers. Realizing that vision will require standardized protocols, large and diverse reference databases built on FAIR data principles, interlaboratory validation, and eventual regulatory acceptance. If those pieces fall into place, the humble thermogram—once confined to polymer labs—may become one of the food industry’s most powerful anti-fraud weapons.

Subject of Research: Applications of differential scanning calorimetry and thermogravimetric analysis in food quality control and authenticity verification

Article Title: Recent applications of DSC and TGA in food quality control and authenticity verification: Current progress and future directions

Article References: Pourmoradian, A., Mohammadi, A., & Barzegar, M. (2026). Recent applications of DSC and TGA in food quality control and authenticity verification: Current progress and future directions. Food Chemistry: X, Article 104587. https://doi.org/10.1016/j.fochx.2026.104587

Image Credits: AI Generated

DOI: Not provided

Keywords: food fraud, food authentication, differential scanning calorimetry, thermogravimetric analysis, thermal analysis, adulteration detection, olive oil, honey, chemometrics, machine learning, oxidative stability, glass transition