Title Science and Technology Indonesia e-ISSN:2580-4391 p-ISSN:2580-4405 Vol. 6, No. 1, January 2021 Research Paper Facile Detection of Oil Adulteration using UV-Visible Spectroscopy Coupled with Chemometric Analysis Nina Gusti1, Dinda Oktarina1, Rina Elvia1, Euis Nursa’adah1, Rendy W. Wardhana1, Agus Sundaryono1, M. Lutfi Firdaus1* 1Graduate School of Science Education, University of Bengkulu, Indonesia *Corresponding author: lutfi@unib.ac.id Abstract Engine and machine oils, be�er known as lubricant, is a fast-moving part within the motorcycle and automobile industries. Due to its high demand, these oils are o�en counterfeited by irresponsible people to get more profit. The thing most o�en done to adulterate oil is by mixing it with other low-quality or used oil. Here, we propose a simple analytical method to identify oil adulteration by using UV-Visible spectroscopy coupled with chemometric analysis. A number of 425 genuine and adulterated oils were used as samples. A�er appropriate dilution using n-hexane, the samples were analyzed by UV-Visible spectrophotometer followed by Principle Component Analysis (PCA) and Principle Component Regression (PCR) as part of the chemometric analysis. The results show that prediction samples were accurately classified into their corresponding groups with PCA scores of 49% and 27% for principal component 1 and 2, respectively. PLS model achieved a good prediction to detect lubricant oil adulteration, with R-square of predicted and reference samples were 0.9257 and 0.9204, respectively. The proposed method shows a promising alternative to the conventional chemical method using a more sophisticated instruments such as GC-MS and HPLC for oil or other organic compound identification. Keywords Lubricant oil, adulteration, UV-Visible spectroscopy, chemometrics, multivariate analysis Received: 14 October 2020, Accepted: 18 December 2020 https://doi.org/10.26554/sti.2021.6.1.14-18 1. INTRODUCTION Lubricant, generally, contains 90% base oil and less than 10% additives. Base oil may derive from petroleum fractions, mineral oils, vegetable oils or synthetic liquids such as hydrogenated polyole�ns, esters, silicones, �uorocarbons and many others (Iłowska et al. (2018); Liu et al. (2019a)). Synthetic oil is a man- made lubricant that consists of chemically modi�ed materials such as petroleum components (Liu et al., 2019b). Oil is used as lubricant due to it reduces friction; the lubricant is more slippery than the components it is applied to. A reduction in friction gen- erate easier movement, less wear on components, and reduced energy needs of the system (Díaz Tovar et al., 2018). Currently, renewable lubricants with tailored molecular architecture are also introduced to the market (Liu et al., 2019a). High popula- tion growth followed by the increase of purchasing power has resulted in more vehicles being produced and used worldwide. Engine (lubricant or machine) oil is one of the fast-moving items in a vehicle whose production is increasing due to the high de- mand. This results in increasing oil price, accompanied by the oil adulteration that is becoming more widespread. Therefore, it is necessary to develop an analytical method to detect oil adul- teration. Up to today, the qualitative and quantitative analysis of oil have been conducted using one or combination of HPLC, GC-MS and FT-IR (Escandar et al. (2006); Mobaraki and Hem- mateenejad (2011); Wiberg (2006)). Chemometrics technique were often embedded to the analysis of organic compounds for a more comprehensive and present additional insight on the analysis (Hanrahan (2008); Miller and Miller (2018)). In this pa- per, we report the development of UV-Vis spectrometry coupled with chemometrics technique to discriminate lubricant-oil adul- teration that provide a robust and low-cost alternative to the conventional chemical method. 2. EXPERIMENTAL SECTION 2.1 Materials Dilution of oil samples were conducted by using n-hexane as a solvent. This n-hexane and other chemicals were purchased from Merck (Darmstadt, Germany). Various lubricant oils were pur- chased from local market in Bengkulu city, Indonesia. Cheaper and used oils that is often used to adulterate lubricant were ob- tained from traditional motorbike service and workshop (Figure 1 ). A Genesys 20 UV-visible Spectrophotometer from Thermo Fisher Scienti�c (Massachusetts, USA) was used to obtain the https://doi.org/10.26554/sti.2021.6.1.14-18 Nina Gusti et. al. Science and Technology Indonesia, 6 (2021) 14-18 Figure 1. The photograph of genuine oil X (a), genuine oil Y (b), and adulterated oil (c) spectra. Quartz and glass cuvettes, beaker glass, erlenmeyer were used for samples preparation and analysis. The chemomet- rics analysis was done by using The Unscrambler X software from CAMO Analytics (Oslo, Norway). 2.2 Methods As much as 5 genuine oils from di�erent companies were pur- chased from the authorized motorcycle workshop. In addition, we purchased various low-quality oils and used oils from tradi- tional motorcycle workshop as samples of mixture and adulter- ated oils. Before analyzed, oils were �ltered and diluted with n-hexane in order to get appropriate absorbance when analyzed by UV-Visible spectrophotometer. Dilution of oil with n-hexane was conducted from 10% (9:1) to 90% (1:9) dilution factors. The prepared samples in cuvettes were scanned from 420 to 920 nm using UV-Visible spectrophotometer. A total number of 425 samples were analyzed in this research, including 5 genuine lu- bricant oil samples. The csv �les from spectrophotometer were converted and processed with MS Excel software (Microsoft, USA) and further analyzed in Unscramble X software using two models, i.e. Principal Components Analysis (PCA) and Partial Least Squares (PLS), as chemometrics analysis (Nunes (2014); Tan et al. (2015); Wold et al. (2001)). 3. RESULTS AND DISCUSSION 3.1 UV-Visible spectroscopy Figure 2 shows the typical spectra of UV-visible for genuine and adulterated oils at wavelength from 220 to 920 nm. The mix- ture of oil, including those of adulterated ones, shows distinct absorbance between 350 to 390 nm that we used as the basis to di�erentiate various lubricant and its mixtures using this simple least square technique. In general, a good separation between each sample were observed from 300 to 420 nm wavelength. Therefore, a good linear correlation between oil impurity and absorbance were noticed, as shown in Figure 2. However, at low wavelength from 220 to 300 nm, the noise of spectra was high and thus it is impossible to di�erentiate between genuine and adulterated ones using a simple and traditional linear regres- sion. At high wavelength from 520 to 920 nm, the spectra were uniform almost for all samples. In order to obtain a more compre- hensive results of adulterated oil identi�cation, a more complex computation of the whole UV-Visible spectra using chemomet- ric technique is necessary. Therefore, we use chemometric as multivariate analysis to elucidate the UV-Visible spectra. Figure 2. Typical UV-visible spectra of non- and adulterated-oil at wavelength from 220 to 920 nm. The dotted-line box shows the best spectra separation that can be used to draw a simple linear regression Figure 3. Simple linear regression (SLR) of oil impurity versus absorbance at 380 nm wavelength A simple linear regression of oil impurity versus absorbance at wavelength of 380 nm is shown in Figure 3. Since the color of diluted oil is pale yellow, it shows a good linearity in ultra violet range with R-square = 0.9891 with linear equation of y = 1.1694x + 0.0437. Genuine oil shows a lower absorbance due to a clearer liquid compared to the adulterated ones. Absorbance ranges from 0.2 to 1.1 that shows a wide dynamic range and in turn it will provides a high sensitivity. 3.2 Chemometrics analysis Two models of chemometrics, i.e. Principal Components Analy- sis (PCA) and Partial Least Squares (PLS), were chosen as mul- tivariate techniques to analyze the principal component in oil samples. The PCA technique is well known to predict adul- teration of samples with appropriate precision and sensitivity (Herrero-Latorre et al., 2019). Figure 4 shows the PCA plot de- rived from UV-visible spectra from 220 to 920 nm wavelength of oil using a projection onto the �rst two principal components. A good separation of genuine and adulterated oils was achieved with percentage of the �rst principal component (PC-1) and the second principal component (PC-2) were 49% and 27%, respec- tively. These results show that both PC-1 and PC-2 components represent 76% of the total variability, and successfully captur- ing the data structures. According to Sugianti et al. (2016), if the number of variants of PC 1 and PC 2 are greater than 70%, © 2021 The Authors. Page 15 of 18 Nina Gusti et. al. Science and Technology Indonesia, 6 (2021) 14-18 Figure 4. Principal Components Analycsis (PCA) derived from UV-visible spectra of oil using a projection onto the �rst two principal components Figure 5. Partial Least Squares (PLS) derived from oil UV-Visible spectra then the score plot shows a good two-dimensional visualization. Therefore, prediction samples were accurately classi�ed into their corresponding groups of PC-1 and PC-2. Various genuine oils were identi�ed as a cluster between -5 to -10 (PC-1) and -2 to +1 (PC-2), while those of adulterated oils were located between +5 to +10 (PC-1) and +2 to +7 (PC-2). The second model of chemometrics technique used in this research was Partial Least Squares (PLS), as shown in Figure 5. The UV-Visible spectra from 220 to 920 nm wavelength of 425 samples were applied using a full cross-validation classi�- cation model. The R-square of predicted and reference samples were 0.9257 and 0.9204, respectively. The calibration model with R-square greater than 0.91 is treated to be an excellent predic- tion (Suhandy and Yulia, 2017). These high value of R-square, i.e., close to 1.0, con�rm that the PLS model is appropriate to discriminate the genuine lubricant oil and adulterated ones. Fur- thermore, the value of mean squared error (RMSECV) was low enough for both predicted and reference samples, i.e. 0.0052 and 0.0053, respectively. These low values show that PLS model is not over �tting which is a good result for the purpose of this research (Sirisomboon and Posom, 2019). Table 1 shows the application of chemometric techniques to analyze various samples that mixed with impurities. Most of the samples are organic compounds and thus HPLC, GC-MS, FT-IR and UV-Vis spectrometry were employed as an initial instrumen- tation part. Detected samples include pharmaceuticals, food, cof- fee, wine, and oil. The advantage of UV-Vis spectrometry among other methods is its simple operation procedure. Furthermore, colored samples such as oil can be easily detected using visible spectrometry. Further works will include the digital image col- orimetry coupled with chemometric techniques and smartphone applications (Firdaus et al. (2019); Herrero-Latorre et al. (2019)). The most common chemometrics techniques used for organic analysis were PCR and PLS, including those of the present study. The results of present study show that both techniques, i.e., PCR and PLS, can be used as qualitative and quantitative analytical method to identify the adulteration of lubricant oils. 4. CONCLUSIONS We have successfully discriminated genuine and adulterated oil samples using a simple and rapid method of UV-Vis spec- troscopy coupled with chemometrics technique. Both model of PCA and PLS achieved good prediction to detect lubricant oil adulteration, with R-square of predicted and reference samples were 0.9257 and 0.9204, respectively. The results suggest that the proposed method is a promising alternative with a low-cost and less chemicals for assessment of lubricant oil adulteration. 5. ACKNOWLEDGEMENT This work was funded by Indonesian Kemenristek-BRIN, under grant number 8/E1/KPT/2020 and 036/SP2H/LT/DRPM/2020 for master thesis (M.Sc) research project. REFERENCES Barbosa, M. F., D. S. DO NASCIMENTO, M. GRÜNHUT, H. V. Dantas, B. S. F. BAND, M. C. U. DE ARAÚJO, and M. Insausti (2017). Fast determination of biodiesel content in commercial diesel/biodiesel blends by using digital images and multivari- ate calibration. Analytical Sciences, 33(11); 1285–1289 Bhaskar, R., R. Bhaskar, M. K. Sagar, V. Saini, and K. Bhat (2012). Simultaneous Determination of Verapamil Hydrochloride and Gliclazide in Synthetic Binary Mixture and Combined Tablet Preparation by Chemometric-Assisted Spectroscopy. Journal of Analytical Sciences, Methods and Instrumentation, 02(03); 161–166 Chen, Y., S.-B. Zhu, M.-Y. Xie, S.-P. Nie, W. Liu, C. Li, X.-F. Gong, and Y.-X. Wang (2008). Quality control and original discrimina- tion of Ganoderma lucidum based on high-performance liquid chromatographic �ngerprints and combined chemometrics methods. Analytica Chimica Acta, 623(2); 146–156 Costa, G. B., D. D. S. Fernandes, V. E. Almeida, T. S. P. Araújo, J. P. Melo, P. H. G. D. Diniz, and G. Véras (2015). Digital image-based classi�cation of biodiesel. Talanta, 139; 50–55 Díaz Tovar, J. S., S. Valbuena-Duarte, and F. Racedo-Niebles (2018). Study of non-linear optical properties in automobile lubricating oil via Z-Scan technique. Revista Facultad de Inge- niería Universidad de Antioquia, (86); 27–31 Didham, M., V. K. Truong, J. Chapman, and D. Cozzolino (2020). Sensing the Addition of Vegetable Oils to Olive Oil: The Ability of UV–VIS and MIR Spectroscopy Coupled with Chemometric Analysis. Food Analytical Methods, 13(3); 601–607 Dinç, E. and Ö. Üstündağ (2003). Spectophotometric quantita- tive resolution of hydrochlorothiazide and spironolactone in © 2021 The Authors. Page 16 of 18 Nina Gusti et. al. Science and Technology Indonesia, 6 (2021) 14-18 Table 1. Application of chemometric techniques for various samples analysis Analytical Chemometric Sample Reference method model HPLC PCA Pharmaceutical (Wiberg, 2006) PLS Pharmaceutical (Chen et al., 2008) GC-MS PCA Pharmaceutical (Escandar et al., 2006) (Pan et al., 2011) (Gröger and Zimmermann, 2011) PLS Metabolomics FT-IR PLS Carbonyl (Mobaraki and Hemmateenejad, 2011) PCA Food (Foca et al., 2011) UV-Vis PCR Pharmaceutical (Hadad et al., 2008) spectrometry PCR Wine (Martelo-Vidal and Vázquez, 2014) PLS Pharmaceutical (Bhaskar et al., 2012) PCA, PLS Co�ee (Souto et al., 2015) MLR Pharmaceutical (Dinç and Üstündağ, 2003) PCA, PLS Olive oil (Didham et al., 2020) PCA, PLS Lubricants oil Present study Digital image PLS Biodiesel (Costa et al., 2015) Colorimetry (Barbosa et al., 2017) LDA Vegetable oil (Milanez et al., 2017) tablets by chemometric analysis methods. Il Farmaco, 58(11); 1151–1161 Escandar, G. M., P. C. Damiani, H. C. Goicoechea, and A. C. Olivieri (2006). A review of multivariate calibration methods applied to biomedical analysis. Microchemical Journal, 82(1); 29–42 Firdaus, M. L., A. Aprian, N. Meileza, M. Hitsmi, R. Elvia, L. Rah- midar, and R. Khaydarov (2019). Smartphone coupled with a paper-based colorimetric device for sensitive and portable mercury ion sensing. Chemosensors, 7(2); 25 Foca, G., F. Masino, A. Antonelli, and A. Ulrici (2011). Prediction of compositional and sensory characteristics using RGB digital images and multivariate calibration techniques. Analytica Chimica Acta, 706(2); 238–245 Gröger, T. and R. Zimmermann (2011). Application of parallel computing to speed up chemometrics for GC× GC–TOFMS based metabolic �ngerprinting. Talanta, 83(4); 1289–1294 Hadad, G. M., A. El-Gindy, and W. M. Mahmoud (2008). HPLC and chemometrics-assisted UV-spectroscopy methods for the simultaneous determination of ambroxol and doxycycline in capsule. Spectrochimica Acta Part A: Molecular and Biomolecu- lar Spectroscopy, 70(3); 655–663 Hanrahan, G. (2008). Environmental chemometrics: principles and modern applications. CRC Press Herrero-Latorre, C., J. Barciela-García, S. García-Martín, and R. M. Peña-Crecente (2019). Detection and quanti�cation of adulterations in aged wine using RGB digital images combined with multivariate chemometric techniques. Food chemistry: X, 3; 100046 Iłowska, J., J. Chrobak, R. Grabowski, M. Szmatoła, J. Woch, I. Szwach, J. Drabik, M. Trzos, R. Kozdrach, and M. Wrona (2018). Designing lubricating properties of vegetable base oils. Molecules, 23(8); 2025 Liu, S., T. R. Josephson, A. Athaley, Q. P. Chen, A. Norton, M. Ier- apetritou, J. I. Siepmann, B. Saha, and D. G. Vlachos (2019a). Renewable lubricants with tailored molecular architecture. Science advances, 5(2); eaav5487 Liu, S., B. Saha, and D. G. Vlachos (2019b). Catalytic production of renewable lubricant base oils from bio-based 2-alkylfurans and enals. Green Chemistry, 21(13); 3606–3614 Martelo-Vidal, M. and M. Vázquez (2014). Determination of polyphenolic compounds of red wines by UV–VIS–NIR spec- troscopy and chemometrics tools. Food chemistry, 158; 28–34 Milanez, K. D. T. M., T. C. A. Nóbrega, D. S. Nascimento, M. In- sausti, B. S. F. Band, and M. J. C. Pontes (2017). Multivari- ate modeling for detecting adulteration of extra virgin olive oil with soybean oil using �uorescence and UV–Vis spectro- scopies: A preliminary approach. LWT-Food Science and Tech- nology, 85; 9–15 Miller, J. and J. C. Miller (2018). Statistics and chemometrics for analytical chemistry. Pearson education Mobaraki, N. and B. Hemmateenejad (2011). Structural char- acterization of carbonyl compounds by IR spectroscopy and chemometrics data analysis. Chemometrics and Intelligent Laboratory Systems, 109(2); 171–177 Nunes, C. A. (2014). Vibrational spectroscopy and chemomet- rics to assess authenticity, adulteration and intrinsic quality parameters of edible oils and fats. Food Research International, 60; 255–261 Pan, R., F. Guo, H. Lu, W.-w. Feng, et al. (2011). Development of © 2021 The Authors. Page 17 of 18 Nina Gusti et. al. Science and Technology Indonesia, 6 (2021) 14-18 the chromatographic �ngerprint of Scutellaria barbata D. Don by GC–MS combined with Chemometrics methods. Journal of pharmaceutical and biomedical analysis, 55(3); 391–396 Sirisomboon, P. and J. Posom (2019). On-line measurement of activation energy of ground bamboo using near infrared spectroscopy. Renewable Energy, 133; 480–488 Souto, U. T. d. C. P., M. F. Barbosa, H. V. Dantas, A. S. de Pontes, W. da Silva Lyra, P. H. G. D. Diniz, M. C. U. de Araújo, and E. C. da Silva (2015). Identi�cation of adulteration in ground roasted co�ees using UV–Vis spectroscopy and SPA-LDA. LWT-Food Science and Technology, 63(2); 1037–1041 Sugianti, C., N. Apratiwi, D. Suhandy, M. Telaumbanua, S. Waluyo, and M. Yulia (2016). STUDIES ON THE USE OF UV-VIS SPECTROSCOPY FOR IDENTIFICATION OF BLEND- ING OF CIVET COFFEE WITH ARABICA COFFEE. Jurnal Teknik Pertanian Lampung, 5(3); 167–176 Suhandy, D. and M. Yulia (2017). The use of partial least square regression and spectral data in UV-visible region for quanti�- cation of adulteration in Indonesian palm civet co�ee. Inter- national journal of food science, 2017 Tan, J., R. Li, and Z.-T. Jiang (2015). Chemometric classi�cation of Chinese lager beers according to manufacturer based on data fusion of �uorescence, UV and visible spectroscopies. Food chemistry, 184; 30–36 Wiberg, K. (2006). Quantitative impurity pro�ling by principal component analysis of high-performance liquid chromatography–diode array detection data. Journal of Chro- matography A, 1108(1); 50–67 Wold, S., M. Sjöström, and L. Eriksson (2001). PLS-regression: a basic tool of chemometrics. Chemometrics and intelligent laboratory systems, 58(2); 109–130 © 2021 The Authors. Page 18 of 18 INTRODUCTION EXPERIMENTAL SECTION Materials Methods RESULTS AND DISCUSSION UV-Visible spectroscopy Chemometrics analysis CONCLUSIONS ACKNOWLEDGEMENT