Abstract

Melamine was quantified in liquid milk without performing the usual liquid-liquid or solidphase extraction steps. Following a simple clean up procedure (extraction by acetonitrile/water mixture, centrifugation and filtration by 0.45 μm membrane), melamine was accurately quantified in fullfat liquid milk down to 0.057 mg kg^-1 using second order multivariate calibration. Second order data was created from absorbance-pH measurements (batch-experiment method) taking the advantage of acid/base properties of melamine. The study was carried out over the spectral domain 200-250 nm and pH 2.0-8.0. Although the number of pH points was lower than the spectral points (4 vs. 51), the data was satisfactorily analyzed by second order multivariate calibration methods. Quantification of melamine in spiked milk samples (0.5-2.0 mg kg^-1) is carried out using unfolded partial least squares-residual bilinearization U-PLS/RBL, parallel factor analysis PARAFAC and multivariate curve resolution-alternating least squares MCR-ALS. U-PLS/RBL outperformed other methods with mean recoveries and relative errors of prediction of 100.1 (2.3), 99.3 (6.2), and 93.8 (16.4) for U-PLS/RBL, MCR-ALS, and PARAFAC, respectively. The reason for success of U-PLS/RBL is attributed to its effectiveness for resolving melamine-signal (the basic form) from the signal of milk-extract. PARAFAC and MCR-ALS showed a modest performance and this is attributed to the rank-deficiency in the collected matrices (for test samples), linear dependency (in pH-dimension), and intense spectral-overlapping in the spectral mode. The performance of the proposed method is critically compared with other published methods.

Keywords: Melamine, matrix effect, pH-spectra, PARAFAC, MCR-ALS, U-PLS/RBL.

Graphical Abstract