Get the 411 on Edible Oils: New Spectroscopic Tool Can ID Edible Oils

Each track has a pattern that produces a reflected beam with a unique sinusoidal modulation for each individual wavelength. The reflected beams are brought to an image on a single detector, which generates an interferogram type of signal. The intensity contribution for each wavelength component is obtained by applying a Fourier transform to the interferogram.

Identifying Oil Samples

An EP-NIR spectrometer covering the 1,375-2,750 nanometer spectral range was coupled to an external halogen NIR source and a 2-millimeter path-length process transmission multimode fiber probe to positively identify samples of canola, corn, and olive oils without false positive responses.

Commercially available samples of canola, corn, and olive oils were analyzed in four replicates, re-collecting a new background between each assay. All spectra were collected using a chemometrics software package (Aspectrics). Data collection parameters were 30-second time integration. Data treatment was performed in three steps:

• Calculating absorbance spectra (from single beam intensity spectra) using open beam configuration as a background;
• Calculating the second order derivative for each absorbance spectrum; and
• Developing principal component analysis (PCA)-based methods for the identification of each type of oil.

Initial chemometrics testing was conducted on the absorbance spectra themselves (Figure 1, p. 36). This approach yielded no significant result; the NIR spectroscopic differences among the three types of oils were too small to be identified and modeled using a PCA approach. When these small variations were enhanced by calculating the second order derivative spectrum of each of the sample absorbance spectrum, however, spectral ranges of 1,600-1,950 nanometers and 2,000-2,600 nanometers (Figure 2, p. 36) were clearly identified as relevant for the identification of material.

Results demonstrated that when projecting the EP-NIR spectrometer response in a space designed to identify a type of oil, all oil spectra are plotted within 99% probability of being the specific type of oil. In addition, the remaining oil samples are plotted within 99% probability of not being the specific type of oil (Figure 3, p. 36). It should be noted that it was the access to spectral information in the 2,000-2,600 nanometer (5,000-3,850 cm-1) range that enabled this application.

The results from this experiment confirm that EP-NIR spectroscopy, together with an NIR source and a 2-millimeter path-length process transmission multimode fiber probe, can identify each type of oil. The results also confirm that the model is solid enough to prevent even partial false positive responses. A significant source of spectral information is obtained in the range between 2,000-2,600 nano-meters (5,000-3,850 cm-1).

Lanher is director of applied sciences and technology at Aspectrics. For more information, call (925) 931-9270 or visit www.aspectrics.com.