ABSTRACT
Microwave technology was assessed and compared with traditional kiln drying to analyze the moisture content of yerba mate leaves/sticks. The moisture content was also measured by near-infrared spectroscopy (NIR). The samples were stored under different conditions of temperature, relative humidity (RH), and time, until 180 days. The microwave method was similar to the official kiln drying method when 900 W/4.3 min was applied. The use of microwave drying showed similar results to the kiln method in the moisture range of 0–10 g/100 g raw material. The partial least square (PLS) models obtained from NIR spectroscopy showed low predictive power, due to the natural heterogeneity of the samples.
Introduction
Yerba mate (Ilex paraguariensis) production is particularly favored in southern Brazil, northwest Argentina, and eastern Paraguay, where its cultivation and processing are important economic activities, not only to serve the South American market, but also for international trade. The species has high nutritional value, particularly antioxidant properties, which is important to maintain human health.[Citation1–Citation4]
Color, flavor, and fragrance are the principal quality attributes of yerba mate,[Citation5] all of which are influenced by moisture content. Therefore, moisture control and monitoring are necessary in the production chain to prevent the growth of spoilage microorganisms and lengthen the storage period.[Citation6] According to Brazil’s National Health Surveillance Agency (ANVISA), the maximum moisture content for secure storage of this product is around 10%.[Citation7]
An interesting alternative for monitoring the moisture of yerba mate is microwave drying, which can save time compared with the official method of kiln drying. Microwave drying has been effectively applied to evaluate the moisture content in several agricultural products, such as okra,[Citation8] olives,[Citation9] bamboo shoots,[Citation10] and wheat flour,[Citation11] among others. Because of the large time savings, this type of monitoring could allow the industry to make rapid decisions that depend on the moisture content of the samples.
Another alternative for moisture analysis is near-infrared spectroscopy (NIR), an analytical technique that measures – by optical absorption – the spectral properties of organic molecules, caused by the harmonic combination of bands in the region of 12,500–4000 cm−1. The advantages of NIR are that it is nondestructive and rapid, making it possible to monitor the characteristics of materials on production lines in a matter of minutes or seconds. This technique has been applied to monitor the moisture content in unroasted coffee,[Citation12] black beans,[Citation13] green tea,[Citation14] and yerba mate powder.[Citation15]
This study is original and innovative because as yet there are no studies based on microwave and NIR, with multivariate data analysis, of yerba mate samples. Within this context, the objective of this study was to develop and test a rapid and economical method to detect moisture content in yerba mate during storage. To achieve this goal, some points were also analyzed: i) microwave efficiency (time and power) in comparison with the official kiln-drying method, to evaluate the moisture content in yerba mate samples composed of leaves and sticks and only leaves; ii) development of partial least square (PLS) models based on NIR; iii) NIR’s potential to predict moisture content in relation to microwave and kiln drying; and iv) moisture behavior in relation to different industrial and laboratory storage conditions during 180 days.
Material and methods
Sample preparation and storage conditions
Samples of yerba mate leaves harvested and industrially processed were collected from companies located in the southeastern Paraná state, Brazil, with geographic coordinates 25° 52’ 36” South and 50° 23’ 3” West. After processing, two groups of samples were analyzed: group 1, composed of leaves and sticks (stems and branches) – 89% leaves and 11% sticks, and group 2, composed only of leaves (100%). The plant material containing sticks is sold mainly for infusion in gourd-like containers called chimarrão in Brazil, while the leaf material (without sticks) is used in tea bags or for preprepared commercial beverages. The samples’ nomenclature and storage conditions are summarized next.
Normal commercial storage
Samples were packed in raffia bags (85 × 120 cm) holding 40 kg, kept in piles of 600 bags in a brick warehouse, at 18°C ± 2°C and relative humidity (RH) of 80% ± 2%. Every 30 days, a small sample weighing about 3 g was collected from each bag for analysis (total of 2000 g).
Accelerated commercial storage
Samples were packed in raffia bags (85 × 120 cm) holding 40 kg, in piles of 200 bags in a kiln, in controlled conditions of temperature (45°C ± 2°C) and RH (31% ± 2%). Every 30 days, a sample was collected from each bag, as described above.
Accelerated laboratory storage
Samples of 15,000 g of leaves/sticks and 15,000 g of leaves, received directly from the company, were fractioned in small sacks (12 × 15 cm). The samples were stored in desiccators with a saturated salt solution at different temperatures. Then, the desiccators were placed in incubators under different temperatures. Every 30 days, a small sample was collected from each sack for analysis (total of 150 g).
The experiment had a 22 factorial design with three central points. The temperature range was 32°C–48°C and the RH range was 10%–50%, where the central points were 40°C and 30%, respectively. As recommended by Bentley,[Citation16] lithium chloride, magnesium chloride, and magnesium nitrate were used to attain 10%, 30%, and 50% RH, respectively. The samples were stored for a period of 180 days and analyzed every 30 days. Only the samples consisting exclusively of leaves were stored in laboratory conditions.
Moisture content by the official method
Samples were dried in the kiln at 105°C until constant weight, in accordance with method 925.09 of the Official Methods of Analysis of AOAC International.[Citation17] The analyses were performed in triplicate on 109 samples, for a total of 327 analyses.
Moisture content by microwave drying
The heating source consisted of an Electrolux Me45x household microwave oven with power control from 100 to 1000 W and frequency of 2450 MHz. To find adequate operational parameters for moisture analysis by microwave, a pretest was carried out, relating the power setting and processing time to the moisture values obtained by the kiln procedure. The pretest was evaluated by the surface response method. Three power settings (500, 700, and 900 W) and three processing times (4.3, 6.3, and 8.3 min) were tested. Samples were received with moisture content between 4.5 and 8.7 g/100 g. A second-order polynomial equation was applied to describe the behavior of the experimental data. The weight of each sample was established as 4 g with 0.1 mg precision in Petri dishes (oven dried at 105°C and weighed). In each stage, six dishes with samples were positioned in a circle as close to the turntable center in the microwave oven as possible. After processing in the established conditions, the dishes were removed, cooled in desiccators, and weighed. Three repetitions for each experiment were evaluated.
Power and temperature conditions, defined in the pretest with microwave heating that best correlated with moisture content obtained in the kiln, were used to assess the moisture content of the samples under all storage conditions. The results after microwave drying were compared to those from the official method by ANOVA and Pearson correlation.
NIR spectroscopy
The NIR spectra were obtained in a Bruker Tensor 37 Fourier-transform spectrophotometer (Bruker Optics, Ettlingen, Germany) with integrating sphere and indium gallium arsenide (InGaAs) detector. Spectra were collected in triplicate, for a total of 327 spectra, in the region from 4000 to 10,000 cm−1, operating in diffuse reflectance mode with 64 scans, with a resolution of 4 cm−1 and a temperature of 20°C. Reflectance data were transformed into log (1/R), where R is reflectance.
The region between 4000 and 8500 cm−1 was selected for building models for PLS analysis. In this region, it was possible to observe homogeneity in all spectra, highlighting water bands, with high absorption intensity in regions 8.333, 6.849, and 5.181 cm−1.[Citation18,Citation19]
Calibrated model for PLS analysis
To quantify the moisture content in yerba mate samples based on NIR spectra, PLS analysis was applied using the Unscrambler® software, version 10.1 (CAMO software). The 327 samples were divided into two groups: 218 samples (80%) for calibration and internal validation of the model, and 109 samples (20%) for external prediction, selected in function of moisture content difference (0.85–9 g/100 g, results of kiln drying analysis of samples) as a means of improving the representativeness.
To obtain multivariate models with good prediction capacity, different pretreatments were evaluated to correct for the effects of light dispersion caused by the non-homogeneity of the samples. The PLS models were evaluated using root mean square error of calibration (RMSEC), root mean square error of cross-validation (RMSECV), root mean square error of prediction (RMSEP), and their respective regression coefficients (R).
Results and discussion
Operational parameters for microwave drying
The results of operational parameters, relating time, and operating power are described by Eq. (1). For this polynomial model, the parameter t2 (time squared) was not significant.
where MM (g/100 g) is the moisture (g/100 g) of yerba mate samples, P is the microwave power (W), and t is the processing time (min). The equation presented a coefficient of determination of 0.81 and error in parameters (MM (g/100 g) = a + b. P + c. P2 + d.P.t) of a ± 1.7952, b ± 0.0047, c ± 3.18 × 10−6 and d ± 0.0002.
The results (Eq. 1) showed that the longer the time and the higher the power applied, the greater was the moisture response value. However, in order to minimize the operational time and define the closest value in comparison with the official AOAC method,[Citation17] the operating power of 900 W was fixed. This was the highest temperature used in the experiments that avoided sample burning. By applying this power in Eq. (1), the total operation time was defined as 4.8 min. However, because of the non-homogeneity of the samples and random errors inherent to the process, in the preliminary tests (not shown) we established that 4.3 min provided the best response and the smallest error fluctuations when determining the moisture content by microwave compared with the results obtained by the official method.
A factorial analysis (ANOVA) at 95% confidence level was applied to verify the effects of the storage conditions on moisture determination after microwave drying and the official method. The combination of factors evaluated involved all parameters, i.e., analytic method, RH, time, and temperature. We observed significant differences between moisture measurements in function of the storage parameters ( and ).
Table 1. ANOVA for leaves/sticks. Dependent variable: moisture (g/100 g).
Table 2. ANOVA for leaves. Dependent variable: moisture (g/100 g).
The results for leaves/sticks showed no significant effect for the interactions between method and temperature, method and RH, and method, time, and RH. For leaves, no significant effect for the interactions between method and temperature, method, time, and temperature, and method, temperature, and RH was observed. This observation indicates there was a statistically significant variation (95% confidence level) in the samples’ moisture based on different storage times, regardless of the analytic method.
To check for the existence of a statistical difference (p < 0.05) between methods for the moisture determination of yerba mate leaves/sticks and leaves, the Tukey test for post hoc comparison was applied to pairs of analysis methods (kiln and microwave) with fixed parameters of time, temperature, and moisture. We observed a statistically significant difference between methods in leaves/sticks only in the analysis of 150 days, 48°C, and 10% RH, where some experimental errors or non-homogeneity of the samples must have occurred.
In the test involving only leaves, five points presented statistically significant differences: 1. [(Time = 180 days), (T = 32°C) and (RH = 50%)]; 2. [(Time = 30 days), (T = 32°C) and (RH = 10%)]; 3. [(Time = 30 days), (T = 48°C) and (RH = 10%); 4. [(Time = 60 days), (T = 32°C) and (RH = 10%)]; and 5. [(Time = 60 days), (T = 48°C) and (RH = 10%)]. For point 1 we attributed the same errors found in leaf/stick samples. On the other hand, points 2–5 corresponded to the same storage time (30 and 60 days) and lower RH (10%). In this case, it is possible that the difference between methods, for these points, occurred in the function of non-homogeneity of water in the samples, since in this storage condition a drying process initiated as a consequence of the low RH. Therefore, in the first month of storage, yerba mate in contact with air might have lost more water volume compared with that inside the packages, since the drying process did not reach equilibrium. Section 3.4 of this article reports the changes in the total amount of water as a function of storage conditions.
Due to the differences between the moisture content obtained by the methods analyzed, data obtained from microwave drying were compared with data obtained from kiln drying to verify the size and behavior of this difference. Some authors applied linear correlations between moisture content from different methods.[Citation20] This representation can be verified as data dispersion (), which visually is moderately close to a linear trend. This was confirmed by Pearson correlations at 95% confidence, which resulted in R = 0.80 and R = 0.81 for leaves/sticks and leaves, respectively. The linear correlation coefficients between responses are described by Eqs. (2) and (3).
Figure 1. Pearson correlation between moisture content of yerba mate determined by the official method (kiln) and microwave for samples of leaves/sticks (solid line, ▲) and leaves (dotted line, ●).
where MM (g/100 g)w and MM (g/100 g)w/o are the moisture content (g/100 g) of yerba mate leaves/sticks and leaves, respectively, and MC (g/100 g) is the moisture content analyzed by kiln drying. Despite some correlation between methods, it is not possible to assert that the equations that represent them are really accurate. The equation for predicting the moisture content in microwave-dried samples represents the means of the sampling range. In function of deviations in the correlation between methods, the mean error obtained by microwave should increase from a determined point in the studied range. In this case, we observed that the equations accurately predicted values of 4.20 g/100 g for leaves/sticks and 3.97 g/100 g for leaves. In other words, these points were moisture values that were equal for both equations, i.e., for both microwave and kiln methods. Moisture values obtained by kiln below these points tended to present, on average, positive deviations in relation to the microwave method, while values above these points tended to present negative deviations. Considering the range between 0 and 10 g/100 g moisture content in yerba mate, the maximum mean absolute error presented by the models was 0.89 g/100 g and 1.23 g/100 g for leaves/sticks, and 0.96 g/100 g and 1.46 g/100 g for leaves, respectively, representing the lower and upper limits of this moisture range. The absolute errors between the values of moisture content obtained experimentally (within the moisture range studied) were 0.03–1.34 g/100 g (leaves/sticks) and 0.04–2.50 g/100 g (leaves), resulting in averages of 0.66 g/100 g and 0.83 g/100 g, respectively. From a practical point of view, these variations can be considered small and thus can be disregarded for the moisture range evaluated. However, with respect to a reliable method for moisture prediction in a wide range, further study should be performed to assess the parameters involved in the differences found between the methods.
According to Holtz et al.,[Citation21] there are some limitations in the drying of plant materials using microwave ovens, such as the heterogeneous structure of the material being studied and the heterogeneous heating that may result. The sum of these problems can cause overheating in some regions of the food sample, which can explain some of the results obtained in this study.
In fact, the use of microwave drying could be extended not only to assess the moisture content, but also for the process itself. Ceni et al.,[Citation5] for instance, investigated the influence of microwave energy on the oxidase activity and moisture content of yerba mate leaves cultivated under high and low light intensity. According to the authors, polyphenol oxidase was inactive after 30 s of microwave treatment, for samples exposed to high and low light intensity. In samples exposed to low intensity, the peroxidase activity was reduced to 60% after 120 s. The exposure of yerba mate to microwave energy during 220 s resulted in the moisture content required for industrial processing. Also worth mentioning is the work of Esmelindro et al.,[Citation22] who observed an increase in the chemical compounds (caffeine, theobromine, phytol, and vitamin E) of microwave-dried yerba mate leaves. Higher values of catechin and polyphenol in green tea were also found when using the microwave technique compared with kiln drying.[Citation23]
NIR spectroscopy
Prediction of moisture content based on NIR
PLS models for moisture content quantification of the samples (leaves/sticks and leaves) were developed separately, correlating the spectral data with the reference values obtained by the official method (kiln) and the values obtained by microwave. The best models were evaluated based on different pretreatments, number of latent variables, and on the verification of anomalous samples. reports the relevant values for calibration and internal validation.
Table 3. PLS models for moisture content prediction in yerba mate in comparison with the official method and microwave.
The results of calibration and cross validation were obtained based on five latent variables, for both models, presenting good precision, with similar values for the regression coefficient (R2), RMSEC, and RMSECV. First derivative and the multiplicative signal correction (MSC) pretreatment were applied for model calibration, by correlating the samples analyzed by microwave (900 W, 4.3 min).
Microwave heating is a function of the interaction of electromagnetic waves with the electric dipole of the material being analyzed. It can change compounds in different ways in comparison to kiln heating, where heat is transferred from the surface to the interior by convection,[Citation24] thus interfering in the analysis and requiring pretreatment of spectra.
When the models were applied for the external prediction of 109 samples (samples not applied in model calibration) evaluated in kiln drying, we obtained a regression coefficient of 0.54, a mean relative error of 0.75%, and RMSEP of 0.97 (). For the correlation with microwave values, a regression coefficient of 0.42, a mean relative error of 0.66%, and RMSEP of 1.08 were obtained (). The greatest disagreement between the predicted and experimental values was obtained from the microwave drying results, observed in , where the blue line is the real prediction and the black line is the ideal prediction (model) for correlating samples with reference values.
Figure 2. PLS prediction of moisture content in yerba mate (leaves/sticks and leaves) by kiln drying (a) and microwave (b).
These results differ from those found in a previous study,[Citation25] where yerba mate powder with moisture content ranging from 6 to 10 g/100 g was oven-dried and subjected to a multivariate model with MSC pretreatment and three latent variables. In that work, an average prediction error of 2.5% was found. Since the authors used powdered samples, better control was attained due to the homogeneity of the heat gradient, caused by the size reduction and improved plant material distribution. Thus, the main reason for the low moisture prediction achieved is related to non-homogeneity of the yerba mate samples (dried and fragmented), leaves/sticks and leaves, and conditions of storage (stored for a long time at different temperatures and RH). Thus, the current model is not ready for implementation in production lines yet, since these variations are inherent in ordinary yerba mate industries.
Evaluation of moisture content after different storage conditions
Yerba mate cannot be stored under ambient conditions during long periods without considerable loss in quality, a consequence of humidity and oxygen deterioration. The fast oxidation of chlorophyll produces a yellow product not accepted by Brazilian consumers, but this is a requirement of the main importers from Brazil: Uruguay, Argentina, and Paraguay.[Citation26] Because of this, exporters typically store their mate products for periods long enough (normal commercial storage (NCC) or accelerated commercial storage (ACC)) to obtain a sufficiently dried product for export. shows the moisture behavior of samples stored in different conditions (kiln method) versus storage time. NCC samples presented similar behavior in relation to moisture content, which remained from 6 to 6.8 g/100 g during 180 days of storage. The ACC samples presented higher moisture content in the first 30 days of storage, but after 60 days they presented lower moisture than the NCC samples, ranging from 4 to 5 g/100 g from 90 to 180 days of storage.
Figure 3. Moisture content during storage time of yerba mate leaves/stick (a) in conditions of NCC (18°C/81 RH (■)) and ACC (45°C/31 RH (▲)). Moisture content during accelerated storage in laboratory: (b) yerba mate leaves/sticks and (c) leaves: 32°C/50 % RH (■), 48°C/50 % RH (●), 32°C/10 % RH (▲), 48°C/10 % RH (▼), and 40°C/30 % RH (×).
Yerba mate leaves/sticks () at higher RH (32°C/50% RH) continued to have moisture content near 6 g/100 g. All other conditions resulted in a decrease, more accentuated in the condition of 48°C/10% RH. In the first 30 days, a reduction from 6 to 3.5 g/100 g occurred, and the final moisture content was near 1 g/100 g. This low value is related to the low water activity (aw) = 0.1 for RH = 10%, where the product underwent greater alteration caused by the oxidation of compounds.[Citation27]
In yerba mate leaves (), an accentuated and more homogeneous decline in moisture content occurred, which showed the same behavior during storage, while yerba mate leaves/sticks presented higher variation in the moisture profile. These changes can be attributed to the different cellular structure of the sticks, as well as their size, form, chemical composition, and the physical disposition of cells.[Citation28] This factor can be responsible for the smoothness of taste in processed yerba mate. Although storage in controlled conditions is a crucial factor to maintain quality and increase the conservation period of processed yerba mate, data from the moisture content of this raw material as a result of storage conditions are scarce in the literature, the reason why this study is important.
Conclusion
Based on the results obtained, the use of microwave drying is a promising technique for the rapid prediction of moisture content in processed yerba mate for the moisture range studied (4.5–8.7 g/100 g). The microwave method saves substantial time for moisture measurement: 4.3 min in comparison to 6 h for kiln drying when 900 W was used. There was no significant difference between the mean moisture values obtained by both methods, which showed linear correlation above 0.80. We suggest a more careful study to improve the prediction of moisture content of fragmented samples of yerba mate by the NIR technique, considering the effect of heterogeneity of samples. The moisture content after the storage time (180 days) in industry (natural and accelerated) was 4.5–6.5 g/100 g, while in the laboratory (accelerated) it was 1–6 g/100 g. Some of the storage parameters applied can have a drying effect, which probably contributed to the non-homogeneity of the moisture content of these samples.
Acknowledgments
We are grateful to Ervateira Baldo S.A. for supplying the yerba mate and to the Department of Industrial Wood Engineering/UFPR for the technical support regarding the sample anatomy and identification.
Funding
We thank CAPES (Office to Improve University Personnel) for the scholarships granted to Cátia Nara Tobaldini Frizon and João Luiz Andreotti Dagostin.
Additional information
Funding
Related Research Data
References
- Arçari, D.P.; Bartchewsky, W.; dos Santos, T.W.; Oliveira, K.A.; Funck, A.; Pedrazzoli, J.; de Souza, M.F.; Saad, M.J.; Bastos, D.H.; Gambero, A.; Carvalho, P.de O.; Ribeiro, M.L. Antiobesity Effects of Yerba Mate Extract (Ilex paraguariensis) in High-Fat Diet-Induced Obese Mice. Obesity 2009, 17, 2127–2233.
- Martins, F.; Suzan, A.J.; Cerutti, S.M.; Arçari, D.P.; Ribeiro, M.L.; Bastos, D.H.; Carvalho, P. de O. Consumption of Mate Tea (Ilex paraguariensis) Decreases the Oxidation of Unsaturated Fatty Acids in Mouse Liver. British Journal of Nutrition 2009, 101, 527–32.
- Filip, R.; Davicino, R.; Anesini, C. Antifungal Activity of the Aqueous Extract of Ilex paraguariensis against Malassezia furfur. Phytotherapy Research 2010, 24, 715–719.
- Puangpraphant, S.; Berhow, M.A.; de Mejia, E.G. Mate (Ilex paraguariensis St. Hilaire) Saponins Induce Caspase-3-Dependent Apoptosis in Human Colon Cancer Cells in vitro. Food Chemistry 2011, 125, 1171–1178.
- Ceni, G.C.; Baldissera, E.M.; Primo, M.S.; Antunes, O.A.C.; Dariva, C.; Oliveira, J.V.; Oliveira, D. Influence of Application of Microwave Energy on Quality Parameters of Mate Tea Leaves (Ilex paraguariensis St. Hil.). Food Technology and Biotechnology 2009, 47, 221–226.
- Alibas-Ozkan, I.; Akbudak, B.; Akbudak, N. Microwave Drying Characteristics of Spinach. Journal of Food Engineering 2007, 78, 577–583.
- Agência Nacional de Vigilância Sanitária. Resolução (ANVISA) - RDC nº 303, de 07 de novembro de 2002. Aprova o Regulamento Técnico para Fixação de Identidade e Qualidade do Composto de Erva-Mate. http://portal.anvisa.gov.br/documents/33916/394219/RDC_303_2002.pdf/6acf6086-1086-4bfc-af80-1b74a213a346.
- Dadali, G.; Apar, D.K.; Özbek, B. Microwave Drying Kinetics of Okra. Drying Technology 2007, 25, 917–924.
- Veillet, S.; Tomao, V.; Visinoni, F.; Chemat, F. New and Rapid Analytical Procedure for Water Content Determination: Microwave Accelerated Dean–Stark. Analytica Chimica Acta 2009, 632, 203–207.
- Bal, L.M.; Kar, A.; Satya, S.; Naik, S.N. Drying Kinetics and Effective Moisture Diffusivity of Bamboo Shoot Slices Undergoing Microwave Drying. International Journal of Food Science and Technology 2011, 45, 2321–2328.
- Garcia, L.G.C.; Vendruscolo, F.; Silva, F.A. Determinação do teor de água em farinhas por micro-ondas. Revista Brasileira de Produtos Agroindustriais 2014, 16, 17–25.
- Morgano, M.A.; Faria, C.G.; Ferrão, M.F.; Bragagnolo, N.; Ferreira, M.M.C. Determination of Moisture in Raw Coffee by Near Infra-red Reflectance Spectroscopy and Multivariate Regression. Ciência e Tecnologia de Alimentos 2008, 28, 12–17.
- El-Sayd, N.I.; Makawy, M.M. Comparison of Methods for Determination of Moisture in Food. Research Journal of Agriculture and Biological Sciences 2010, 6, 906–911.
- Sinija, V.R.; Mishra, H.N. FTNIR Spectroscopic Method for Determination of Moisture Content in Green Tea Granules. Food Bioprocess Technology 2011, 4, 136–141.
- Mazur, L.; Peralta-Zamora, P.G.; Demczuk B., Jr.; Ribani, R.H. Application of Multivariate Calibration and NIR Spectroscopy for the Quantification of Methylxanthines in Yerba Mate (Ilex paraguariensis). Journal of Food Composition and Analysis 2014, 35, 55–60.
- Bentley, R.E. Handbook of Temperature Measurement; Springer: New York, 1998.
- Horwitz, W.;Latimer Jr., G.W., Official Methods of Analysis of AOAC International, 18th ed.; AOAC INTERNATIONAL: Gaithersburg, MD, USA, 2005.
- Murray, I. The NIR Spectra of Homologous Series of Organic Compounds. In: Proceedings of the International NIR/NIT Conference, Budapest, 1986, pp. 13–28.
- Miller, C.H.E. Chemical Principles of Near Infrared Technology. In: Near Infrared Technology in the Agricultural and Food Industries, American Association of Cereal Chemist: USA, 2001, 19–39.
- Chen, C. Evaluation of Air Oven Moisture Content Determination Methods for Rough Rice. Biosystems Engineering 2003, 86, 447–457.
- Holtz, E.; Ahrné, L.; Rittenauer, M.; Rasmuson, A. Influence of Dielectric and Sorption Properties on Drying Behaviour and Energy Efficiency during Microwave Convective Drying of Selected Food and Non-food Inorganic Materials. Journal of Food Engineering 2010, 97, 144–153.
- Esmelindro, A.A.; Girardi, J.S.; Mossi, A.; Jacques, R.A.; Dariva, C. Influence of Agronomic Variables on the Composition of Mate Tea Leaves (Ilex paraguariensis) Extracts Obtained from CO2 Extraction at 30°C and 175 bar. Journal of Agricultural and Food Chemistry 2004, 52, 1990–1995.
- Gulati, A.; Rawat, R.; Singh, B.; Ravindranath, S.D. Application of Microwave Energy in the Manufacture of Enhanced-quality Green Tea. Journal of Agricultural and Food Chemistry, 2003, 51, 4764–4768.
- Rosini, F.; Nascentes, C.C.; Nóbrega, J.A. Experimentos Didáticos Envolvendo Radiação Microondas. Química Nova 2004, 27, 1012–1015.
- Mazur, L.; de Oliveira, G.A.; Bicudo, M.O.P.; Ribani, R.H.; Nagata, N.; Peralta-Zamora, P. Multivariate Calibration and Moisture Control in Yerba Mate by Near Infrared Spectroscopy. Acta Scientiarum Technology 2014, 36, 369–374.
- Texeira Neto, R.O. Alterações na qualidade de frutas e hortaliças desidratadas durante a estocagem; Desidratação de Frutas e Hortaliças, Campinas-São Paulo: ITAL. 2002, 8.1–8.9.
- Labuza, T.P.; Tannernbaum, S.R.; Karel, M. Water Content and Stability of Low Moisture and Intermediate Moisture Foods. Food Technology 1970, 24, 543–550.
- Tamasi, O.P.; Filip, R.; Ferraro, G.; Calviño, A.M. Total Polyphenol Content and Perceived Astringency of Yerba Mate “Ilex paraguariensis” Infusions. Journal of Sensory Studies 2007, 22, 653–664.