Article pubs.acs.org/IECR
Isothermal Thermogravimetric Analysis of Soybean Oil Oxidation Correlated to Thin Film Micro-Oxidation Test Methods Mert Arca,†,‡ Brajendra K. Sharma,§ Joseph M. Perez,‡ and Kenneth M. Doll*,† †
USDA/NCAUR/ARS, Bio-Oils Research, 1815 North University Street, Peoria, Illinois 61604, United States Department of Chemical Engineering, Pennsylvania State University, University Park, Pennsylvania 16802, United States § Illinois Sustainable Technology Center, University of Illinois, Urbana−Champaign, 1 Hazelwood Drive, Champaign, Illinois 61820, United States ‡
ABSTRACT: An isothermal thermogravimetric analysis (TGA) method was used to generate a model capable of predicting the results of the Thin Film Micro-Oxidation (TFMO) test in a soybean oil system. Utilizing a series of pseudorate constants and “activation energies”, weight loss data from the TGA method can predict oxidation, polymerization, evaporation, and the formation of high molecular weight deposits, normally observed in the TFMO test. These parameters can be used in simple computer simulations which can calculate the weight of each component of a lubricant as it decomposes. This allows evaluation of lubricants without the difficult to obtain TFMO equipment, instead utilizing equipment available in most laboratories.
■
mol−1 as the activation energy of the oxidation of soybean oil and notes the change which occurs when the temperature is taken above 150 °C. The second, from 2001, reports the effectiveness of antioxidants in ethyl linoleate.24 A more recent report, from 2009, is a study relating to two types of biodiesel with petroleum diesel.25 According to this paper, diesel is less stable than biodiesel, which is in contrast literature. Additionally, the activation energies, 73.86 kJ mol−1 for biodiesel and 21.62 kJ mol−1 for petroleum based diesel, differ by greater than 50 kJ mol−1. This unlikely value was perhaps caused by the volatility difference in the material. Suffice it to say there are no isothermal TGA studies on the oxidation of vegetable oil at temperatures over 150 °C. Although the extension of this system to other biobased systems is of significant interest, this report will focus on the degradation of soybean oil under oxygen at temperatures ranging from 320 to 430 °C. These conditions are close to relevant temperatures yet give experiments of appropriate time. Also, the study of a system without added antioxidants was done in order to simplify the results. Further work on those systems is of significant interest. The weight curve is analyzed using a proposed kinetic model which estimates primary oxidation, further oxidation, the formation of high molecular weight components and insoluble deposits, and the evaporation of all of these components. The effective use of isothermal TGA is also shown to be an effective alternative to the less available TFMO test.
INTRODUCTION The synthesis of vegetable oil based industrial products has been of recent interest over the last century.1,2 In the lubrication area, the search for high performance lubricants that are based on renewable materials3,4 is always growing. Oxidation has been an ongoing weakness of vegetable oil based materials5,6 where several instrumental methods have proven valuable. One such method is the Thin-Film Micro-Oxidation (TFMO) test.7−9 With this test, a sample is placed in a pan and heated under oxygen flow, and the volatility and deposit formation are measured gravimetrically. However, some drawbacks include the following: Temperatures above 250° cannot be exceeded; the experiment has many steps which render it slow and can decrease accuracy of the test; only one data point for volatiles and deposits can be obtained from each sample; and in order to quantify the amount of deposit formation, the oxidized oil has to be washed with tetrahydrofuran which has a negative environmental impact. The test instrument is also of limited availability and its use is not widespread. More common methods for the study of vegetable oil based materials are the Rancimat method10−12 and Pressurized Differential Scanning Calorimetry12−14 and the Rotary Pressure Vessel Oxidation Test.15 However, the modern instrumental method which is most like TFMO in concept is thermogravimetic analysis (TGA) used in isothermal mode. The use of differential scanning calorimetry (DSC)16−18 and TGA19−21 have been used to produce kinetic rates and activation energies. These methods rely on several assumptions as pointed out by the original authors; most vitally, that the rate of conversion is directly related to the observable weight change. Also, the need for a range of temperatures and the ruling out of diffusion effects is specified. While there are numerous reports available, since the 1980s,22 that involve TGA of vegetable oils, only three reports primarily utilize the isothermal variety of this method. The first report is a well done study from 198523 which reports ∼88 kJ © 2012 American Chemical Society
■
MATERIALS AND METHODS Soybean oil (RBD grade, KIC Chemicals, New Paltz, NY) and tetrahydrofuran (ACS Reagent 99+%, Sigma-Aldrich, St. Louis, MO) were used as received. TGA was performed on a TA Received: Revised: Accepted: Published: 3550
August 4, 2011 January 31, 2012 February 11, 2012 February 13, 2012 dx.doi.org/10.1021/ie201696w | Ind. Eng. Chem. Res. 2012, 51, 3550−3555
Industrial & Engineering Chemistry Research
Article
reports,26 higher temperatures and longer reaction times gave higher levels of deposits. Isothermal Thermogravimetric Analysis. The TGA analysis of soybean oil was performed at temperatures from 200 to 430 °C under oxygen for 100 min. The sample pans showed high similarity to those in the TFMO experiment (Figure 4). Kinetic Model. A simple kinetic model was developed in order to use the observable information generated from the TGA experiment. This model is analogous to a similar model developed for the TFMO system,27 utilizing a series of pseudorate constants which account for oxidation, polymerization, and evaporation. A series of seven pseudoelementary reaction rate constants are used to connect four different species (Scheme 1). The original oil (RH) can evaporate, with a pseudorate constant k4, or oxidize, with a pseudorate constant k1. Primary oxidation products (Q), polymeric products (P), and dark insoluble deposits (D) are also accounted for in the scheme. The pseudorate constants can be put into kinetic equations (eqs 1-4) for this model
Instruments (New Castle, DE) Q-500 Thermogravimetric Analyzer. The instrument was set to approach isothermal temperate at 20 °C min−1, which the instrument could achieve in a repeatable manner, without overheating the sample. Flow rates of 20 mL min−1 were used, with dry air in the sample chamber and nitrogen in the balance chamber. Sample sizes of 0.5−9 mg were tested where it was determined that utilization of 2−3 mg samples gave the most consistent results. The samples were placed in a DSC pan which was then placed in the platinum TGA pan (Figure 1). DSC pans made of different
Figure 1. The use of a differential scanning calorimetry (DSC) pan placed in the thermogravimetric analysis (TGA) sample pan, the method used here.
dMRH = −k1MRH − k4MRH dt
metals were used, where the oxidation rates were found to vary, steel > copper > aluminum. The data in this report were generated using aluminum pans. Thin Film Micro-Oxidation (TFMO) was performed on a Walco (State College, PA), model 51000 instrument as described in the literature.8,9 Sample pans (Walco) are washed with THF prior to testing. The sample pan is weighed, ∼20 μL of oil is added, and the samples are weighed again. They are then placed in the glass microreactor cell (Figure 2), and the
dMQ dt
(2)
dMF = k2MQ − k3MP − k6MP dt
(3)
dMD = k3MP − k 7MD dt
(4)
The pseudo-kinetic rate equations for the kinetic model are described in Scheme 1, with corresponding weights obtained from the isothermal TGA curve as shown in Figure 5. The corresponding weights, which are the only observables in the experiment, correspond to the original weight (W0), to the weight at maximum weight loss corresponding to the highest concentration primary oxidation product (W1), to the weight corresponding to the highest concentration of polymeric product (W2), to the point where weight curve starts to flatten indicating conversion of deposits (W3), and to the final weight of the deposits (W4). Next, from the weight curves, the appropriate information was extracted. At this point, it is worth mentioning that we are working with weights and not concentrations. Therefore, the pseudorate constants obtained here are not truly rate constants but only proportional to actual rate constants through this assumption. The varied parts of the curve and their representative weights are shown (Figure 5). Starting with W0, the weight of the oil placed in the DSC pan at t1, t2, and t3 are the points on the curve related to change in behavior. Initially, RH oxidizes with a pseudorate constant of k1 but also evaporates at a pseudorate of k4. The observable weight loss for this evaporation is evident from the weight curve. The next part of the analysis takes advantage of the expected high volatility of Q, hence most of the weight loss should occur where the concentration of Q is at its highest point. This point, t1, is easily identified by a first derivative plot of the weight loss vs time (dW/dt), that is, the maximum of the first derivative plot. After t1, the weight loss rate suddenly decreases, and t2, the point where Q is converted into high molecular weight products (P), can be identified from the dW/dt plot. It is at the intersection of
Figure 2. A diagram of the cell used in the thin film micro-oxidation (TFMO) test.
desired test temperature is set. Flow rate of oxygen is adjusted to 20 mL min−1. Sample pans are held in the reactor for time intervals from 30 to 300 min after temperature is achieved, and the pans are washed with THF and weighed again to determine the amount of insoluble deposits. As in the TGA experiment, steel pans were found to induce greater oxidation than aluminum pans which were used in this study.
■
= k1MRH − k2MQ − k5MQ
(1)
RESULTS AND DISCUSSION
Thin Film Micro-Oxidation. The TFMO test was performed on soybean oil (Figure 3) at temperatures from 175 to 250 °C. As expected, in agreement with previous 3551
dx.doi.org/10.1021/ie201696w | Ind. Eng. Chem. Res. 2012, 51, 3550−3555
Industrial & Engineering Chemistry Research
Article
Figure 3. The sample pans from a series of TFMO runs.
Figure 4. The sample pans from TGA analysis of soybean oil under oxygen at 200 °C (left)−430 °C (right).
two tangents, the first at the highest slope of the dW/dt plot, which occurs near t1, and the lowest slope which is essentially zero. This point is t3 where the dark insoluble deposits only undergo evaporation, a first order reaction, shown by the flat dW/dt graph. The t4 point corresponds to the end of the 100 min isothermal time. From these time points, and their corresponding weights, a series of pseudokinetic equations can be developed (eqs 5-11), which are analogous to those used in the TFMO system.7,28,29
Scheme 1. Scheme Devised for Analysis of TGA Data in This Experimenta
a
Original oil (RH) reacts with oxygen at a rate of k1 forming primary oxidation product (Q). The primary oxidation products then form high molecular weight products (P) at a rate of k2. Finally, high molecular weight products form insoluble deposits (D) at a rate of k3. In addition to oxidation, each step also includes evaporation at rates represented by k4 to k7.
r1 = k1·W1 = 3552
W1 1 ⇒ k1 = t1 − t0 t1
(5)
dx.doi.org/10.1021/ie201696w | Ind. Eng. Chem. Res. 2012, 51, 3550−3555
Industrial & Engineering Chemistry Research
Article
Figure 5. The use of weight curves to generate the weight and time data used in this model.
Table 1. Energies and Constants Calculated Using This Kinetic Model pseudorate constants effective rate (g min−1)
k1
k2
k3
k4
k5
k6
k7
“activation energy” of process “pre-exponential factor” of process correlation coefficient (r2)
17561 1.81 0.86
56548 8678 0.79
84707 613345 0.97
37375 20.77 0.94
77467 183725 0.92
101018 4724851 0.96
58473 277.4 0.83
r2 = k2·W2 =
W2 1 ⇒ k2 = t2 − t1 t2 − t1
(6)
r3 = k3·W3 =
W3 1 ⇒ k3 = t3 − t2 t3 − t2
(7)
r4 = k4·W0 =
W1 − W0 W1 − W0 ⇒ k4 = t1 − t0 (t1 − t0) ·W0
(8)
r5 = k5·W1 =
W2 − W1 W2 − W1 ⇒ k5 = t2 − t1 (t2 − t1) ·W1
(9)
r6 = k6·W2 =
W3 − W2 W3 − W2 ⇒ k6 = t3 − t2 (t3 − t2) ·W2
(10)
r7 = k 7·W3 =
W4 − W3 W4 − W3 ⇒ k7 = t4 − t3 (t4 − t3) ·W3
(11)
Figure 6. An Arrhenius type relationship from the effective rates of weight loss (Table 1), which allows the calculation of “activation energy”.
could be used in a similar manner to activation energies, where pseudorate constants can be calculated at any temperature. Utilizing eqs 1-4, and the calculated “activation energy” and “pre-exponential factors”, a computer simulation (Matlab) accurately determined the weights of each component of the mixture at any temperature and time. An example plot (Figure 7) shows a simulation of an isothermal TGA run which is in high agreement with experimental values.
From these equations, the effective rate of weight loss was calculated for the study at all of the temperatures (Table 1), and an Arrhenius type plot can be made (Figure 6). From the slope of the graph, an “activation energy” and “pre-exponential factor” can be calculated. It is worth mentioning here that these are not true activation energies, a mistake commonly made in the literature of methods of this type. However, these values 3553
dx.doi.org/10.1021/ie201696w | Ind. Eng. Chem. Res. 2012, 51, 3550−3555
Industrial & Engineering Chemistry Research
Article
Figure 7. The Matlab simulation of the isothermal TGA of soybean oil ran at 300 °C.
Figure 9. A TFMO experiment conducted at 250 °C for 300 min and the results of our kinetic model.
Model Values with TFMO. Data obtained from the TGA model could be used to predict the results from TFMO at any temperature and time. Results of various 120 min TFMO experiments (Figure 8) at a range of temperatures (175−250 °C) were fit by the model. An extended (300 min) TFMO experiment at temperature of 250 °C (Figure 9) also gave results that could be accurately modeled. The experimental values are average of 3 trials and error bars are calculated from the standard deviation of the 3 trials. Figures 8 and 9 indicate agreement between our TGA kinetic model and TFMO experiments. In a soybean oil test case, TGA data can be used in a manner similar to that obtained by the more laborious multistep TFMO test. Additionally, this methodology may prove more general in
its application to other lubricant systems and conditions such as on steel surfaces or with antioxidant packages.
■
AUTHOR INFORMATION
Corresponding Author
*Phone: 309-681-6103. Fax: 309-681-6524. E-mail: Kenneth.
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS The authors would like to thank Dr. Robert O. Dunn for comments during the preparation of this manuscript. This
Figure 8. The experimental TFMO tests at 175 °C (a), 200 °C (b), 225 °C (c), and 250 °C (d) for 120 min, and the results of the kinetic model using TGA. 3554
dx.doi.org/10.1021/ie201696w | Ind. Eng. Chem. Res. 2012, 51, 3550−3555
Industrial & Engineering Chemistry Research
Article
(21) Flynn, J. H.; Wall, L. A. A quick, direct method for the determination of activation energy from thermogravimetric data. J. Polym. Sci., Part B: Polym. Lett. 1966, 4 (5), 323−328. (22) Takaoka, K.; Kobayashi, K. Study of thermal oxidation of thin film of trilinolein by thermogravimetric analysis. J. Am. Oil Chem. Soc. 1986, 63 (11), 1447−1451. (23) Mikula, M.; Khayat, A. Reaction conditions for measuring oxidative stability of oils by thermogravimetric analysis. J. Am. Oil Chem. Soc. 1985, 62 (12), 1694−1698. (24) Litwinienko, G.; Dabrowska, M. Thermogravimetric investigation of antioxidant activity of selected compounds in lipid oxidation. J. Therm. Anal. 2001, 65 (2), 411−417. (25) Rodriguez, R. P.; Sierens, R.; Verhelst, S. Thermal and kinetic evaluation of biodiesel derived from soybean oil and higuereta oil. J. Therm. Anal. 2009, 96 (3), 897−901. (26) Erhan, S. Z.; Sharma, B. K.; Liu, Z.; Adhvaryu, A. Lubricant Base Stock Potential of Chemically Modified Vegetable Oils. J. Agric. Food Chem. 2008, 56 (19), 8919−8925. (27) Hsu, S. M.; Chen, C.-I. A Chemical Kinetics Model to Predict Diesel Engine Performance. Part II. Bench-Test Procedures. Tribol. Lett. 2003, 14 (2), 91−97. (28) Chen, C.-I.; Hsu, S. M. A Chemical Kinetics Model to Predict Lubricant Performance in a Diesel Engine. Part I: Simulation Methodology. Tribol. Lett. 2003, 14 (2), 83−90. (29) Naidu, S.; Klaus, E.; Duda, J. Kinetic Model for HighTemperature Oxidation of Lubricants. Ind. Eng. Chem. Prod. Res. Dev. 1986, 25 (4), 596−603.
research was part of a joint effort by the Agricultural Research Service of the United States Department of Agriculture, BioOils Research Group, Peoria, IL, and the Tribology Group, Chemical Engineering Department of the Pennsylvania State University, University Park, PA. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.
■
REFERENCES
(1) Voith, M. Start-ups reveal biobased chemicals. Chem. Eng. News 2010, 88 (07), 28. (2) Tullo, A. H. Catalyzing Biobased Chemicals. Chem. Eng. News 2010, 88 (38), 15−17. (3) Hill, K. Industrial development and application of biobased oleochemicals. In Biorefineries- Industrial processes and products, status quo and future directions; Kamm, B., Gruber, P. R., Kamm, M., Eds.; Wiley VCH: Weinheim, 2006; Vol. 2, pp 291−314. (4) Hwang, H.-S.; Erhan, S. Z. Lubricant Base Stocks from Modified Soybean Oil. In Biobased Industrial Fluids and Lubricants; Erhan, S. Z., Perez, J. M., Eds.; AOCS Press: Champaign, IL, 2002; pp 20−34. (5) Cosgrove, J. P.; Church, D. F.; Pryor, W. A. The kinetics of the autoxidation of polyunsaturated fatty acids. Lipids 1987, 22 (5), 299− 304. (6) Sharma, B. K.; Perez, J. M.; Erhan, S. Z. Soybean Oil-Based Lubricants: A Search for Synergistic Antioxidants. Energy Fuels 2007, 21 (4), 2408−2414. (7) Naidu, S. K.; Klaus, E. E.; Duda, J. L. Evaluation of liquid phase oxidation products of ester and mineral oil lubricants. Ind. Eng. Chem. Prod. Res. Dev 1984, 23 (4), 613−619. (8) Sharma, B. K.; Doll, K. M.; Erhan, S. Z. Ester hydroxy derivatives of methyl oleate: Tribological, oxidation and low temperature properties. Bioresour. Technol. 2008, 99 (15), 7333−7340. (9) Sharma, B. K.; Doll, K. M.; Erhan, S. Z. Oxidation, friction reducing, and low temperature properties of epoxy fatty acid methyl esters. Green Chem. 2007, 9 (5), 469−474. (10) Polavka, J.; Paligová, J.; Cvengros, J.; Simon, P. Oxidation stability of methyl esters studied by differential thermal analysis and rancimat. J. Am. Oil Chem. Soc. 2005, 82 (7), 519−524. (11) Gertz, C.; Klostermann, S.; Kochhar, S. P. Testing and comparing oxidative stability of vegetable oils and fats at frying temperature. Eur. J. Lipid Sci. Technol. 2000, 102 (8−9), 543−551. (12) Dunn, R. O. Antioxidants for improving storage stability of biodiesel. Biofuels, Bioprod. Biorefin. 2008, 2 (4), 304−318. (13) Sharma, B. K.; Stipanovic, A. J. Development of a new oxidation stability test method for lubricating oils using high-pressure differential scanning calorimetry. Thermochim. Acta 2003, 402 (1−2), 1−18. (14) Dunn, R. O. Effect of antioxidants on the oxidative stability of methyl soyate (biodiesel). Fuel Process. Technol. 2005, 86 (10), 1071− 1085. (15) Cermak, S.; Biresaw, G.; Isbell, T. Comparison of a New Estolide Oxidative Stability Package. J. Am. Oil Chem. Soc. 2008, 85 (9), 879−885. (16) Ozawa, T. Kinetic analysis of derivative curves in thermal analysis. J. Therm. Anal. 1970, 2 (3), 301−324. (17) Flynn, J. H. The T ̀ emperature Integral’ -- Its use and abuse. Thermochim. Acta 1997, 300 (1−2), 83−92. (18) Adhvaryu, A.; Erhan, S. Z.; Liu, Z. S.; Perez, J. M. Oxidation kinetic studies of oils derived from unmodified and genetically modified vegetables using pressurized differential scanning calorimetry and nuclear magnetic resonance spectroscopy. Thermochim. Acta 2000, 364 (1−2), 87−97. (19) Ozawa, T. A New Method of Analyzing Thermogravimetric Data. Bull. Chem. Soc. Jpn. 1965, 38 (11), 1881−1886. (20) Ozawa, T. Applicability of Friedman plot. J. Therm. Anal. 1986, 31 (3), 547−551. 3555
dx.doi.org/10.1021/ie201696w | Ind. Eng. Chem. Res. 2012, 51, 3550−3555