NIR Spectroscopy application to determine the carbohydrate and lipid content in green bean coffee

Coffee flavor is influenced by the chemical concentration including the carbohydrate and lipid content. Determination of the chemical content is still using the chemical method that spend time, expensive and destructive, then a suitable method which is fast and non destructive is required especially to fulfill the needs of coffee industries. This study aimed to apply NIRS method to determine the carbohydrate and lipid content from Indonesia green bean coffee from several origin to increase the data distribution, NIR wave of 1000-2500 nm, followed by determination of the carbohydrate and lipid content using chemical method. After that, NIR data pretreatment using multiple scatter correction (MSC), first and second derivative of savitzky golay, normalization and the combinations then calibrate both the spectra and chemical data by Partial Least Square (PLS) method. Result showed that NIR spectroscopy can be used to determine the carbohydrate and lipid content in green bean coffee could be seen by the high coefficient correlation, RPD and consistency value, while the CV is lower than 2.


Introduction
Coffee quality is influenced by several things that the coffee through, such as the seedling to cultivation process, harvesting and also post harvest stage.In post harvest stage, the coffee beans will be dried untill maximum 12.5% water content [1], and during the process the other component is also exist inside the beans.The quality of coffee beans is not only infuenced by the water, but also with all the chemical component inside including the lipid and carbohydrate content.
Carbohydrates are include into the major components of coffee and usually its percentage is approximately > 18% to more than 50% of the dry weight [2][3][4][5].Sucrose is an important component for coffee flavor which will influenced the cup quality of coffee [1].In roasting stage, the Maillard reaction and caramelization will be occured and this is an important stage that will change the color of coffee and emit the aroma of roasted coffee, while in this case the carbohydrate is the precursors of those reactions.
On the other hand the lipid component, which also include as the major component in coffee beans, generally contributes approximately about 10% to 18% in coffee beans [4][5].In arabica coffee, total lipid concentration could be two times that of robusta beans.Lipid consist of fatty acid, and it is 1302 (2024) 012114 IOP Publishing doi:10.1088/1755-1315/1302/1/012114 2 important for coffee to keep it fresh and avoid the staleness caused by hydrolisis and oxidation of triacylglycerols [2].
Therefore, determination of carbohydrate and lipid inside of coffee is required due to their influence to the coffee quality.A fast determination and environmentally friendly method which are not use chemicals or anything that could increase the polution on earth is required for nowadays industry, Near Infrared Spectroscopy (NIRS) could be used as one of the method.Just like the other agricultural products, coffee also consist of carbon, oxygen, hydrogen and nitrogen which make it could be detected by NIR Spectroscopy.When we give NIR light to a material, then the light will be absorbed, transmitted and also reflected both specular and diffuse.Then in NIR spectroscopy, the diffuse reflectance will be measured in the wavelength of 780-2500 nm.The measurement could trasnlated into an absorbtion spectra since the molecules inside the matter will vibrate as it absorb the NIR light corresponding to its frequency.Then, the energy absorbtion is translated into absorbtion spectrum form using NIR Spectrometer [6].
Furthermore, this method nowadays generally used as a fast coffee beans determination in Indonesia, namely NIR can be used to predict caffeine composition in green bean coffee [7,8], determination of trigonelline and chlorogenic acids inside the coffee beans [9], for determining the water content [10], classify the coffee beans from several varieties [11].

Material and Methods
Three types of arabica coffee from Sumatera island, Indonesia which are sidikalang, lintong and gayo coffee beans were used.The samples used were 110 samples for each component.The measurement was conducted by putting the coffee beans sample into 4 layers in a petri dish (± 96 grams) then measure it using NIRFLex N-500 (BUCHI Labortechnic AG Switzerland) in 3 different points in NIR region wavelength of 1000 to 2500 nm.After that, the measured samples will be destructed into a smaller size to measure the lipid and carbohydrate content inside by the chemical method using LCMS.After that, both the spectra and chemical data will be analysed using unscrambler software by calibrating it using partial least square (PLS) method, but before that, several data pretreatment were utilized such as the first and second derivative, standard normal variate, normalization and multiple scatter correction and their combination then evaluate it by statistic [12].

Result and Discussion
The difference in reflectance of coffee beans depend on variation in particle size and composition of coffee beans.There are varieties of percentage in major chemical composition which obtained by the chemical analysis.Percentages of major chemical composition are shown in Table 1.The best calibration and validation result of each major chemical component is shown in Table 2.The correlation between both carbohydrate and lipid composition by 6 factors of PLS are quite similar; however, in the carbohydrate determination the data treatment used is more simple which is only using the 1st derivative.This condition in ifluence by the percentation of the chemical content inside the beans, since the carbohyrate contain higher percentage, then a simple method could be used to build a robust PLS model determination.Moreover, the result showed a better model compared to [5], it is because in building a NIR model, the variation, range and the number of the data will be include as a defining factors.In this case, by a bigger range of data as shown in Table 1, a better NIR model determination for both carbohydrate and lipid could be achieved.
Result shows that for both chemical compositions in Table 2 shows a robust model since the r value is higher than 90%, RPD is higher than 5 and CV is lower than 2.Moreover, the stability of the model also could be seen by the consistency that higher than 80% which mean the determination model could be used [12].

Conclusion
Both carbohydrate and lipid content of Indonesian coffee beans are predicted well by NIRS method with r > 0.98 and CV < 2%.Results indicated that this method could be used to build a nondestructive carbohydrate and lipid determination in Indonesian coffee beans.

Table 1 .
Major Chemical Composition obtained by proximate analysis

Table 2 .
Result of The Best Calibration and Validation